MODERN TRENDS

Edward E. Wallach, M.D.

Associate Editor

Evidence-based investigations and treatments of

recurrent pregnancy loss
Ole B. Christiansen, M.D., Ph.D.,a Anne-Marie Nybo Andersen, Ph.D.,b Ernesto Bosch, M.D.,c
Salim Daya, M.B.,d Peter J. Delves, Ph.D.,e Thomas V. Hviid, Ph.D.,f William H. Kutteh, Ph.D.,g
Susan M. Laird, Ph.D.,h Tin-Chiu Li, Ph.D.,i and Katrin van der Ven, M.D.j
a

Fertility Clinic 4071 and f Department of Clinical Biochemistry, Rigshospitalet, Copenhagen; b Institute of Social Medicine,
University of Copenhagen, Copenhagen, Denmark; c Instituto Valenciano de Infertilidad, Valencia, Spain; d Department of
Obstetrics and Gynecology, McMaster University, Hamilton, Ontario, Canada; e Department of Immunology and Molecular
Pathology, Division of Infection and Immunity, University College London; h Division of Biomedical Sciences, Sheffield
Hallam University, Sheffield; i Biomedical Research Unit, Jessop Hospital for Women, Sheffield, United Kingdom; g Division
of Reproductive Endocrinology and Immunology, Department of Obstetrics and Gynecology, University of Tennessee,
Memphis, Tennessee; and j IVF Clinic, Department of Obstetrics and Gynaecology, University of Bonn, Bonn, Germany

Objective: To give an overview of currently used investigations and treatments offered to women with recurrent
pregnancy loss (RPL) and, from an evidence-based point of view, to evaluate the usefulness of these interventions.
Design: Ten experts on epidemiologic, genetic, anatomic, endocrinologic, thrombophilic, immunologic, and
immunogenetic aspects of RPL discussed methodologic problems threatening the validity of research in RPL
during and after an international workshop on the evidence-based management of RPL.
Conclusion(s): Most RPL patients have several risk factors for miscarriage, and an extensive investigation for all
major factors should always be undertaken. There is an urgent need for agreement concerning the thresholds for
detecting what is normal and abnormal, irrespective of whether laboratory tests or uterine abnormalities are
concerned. A series of lifestyle factors should be reported in future studies of RPL because they might modify
the effect of laboratory or anatomic risk factors. More and larger randomized controlled trials, including trials of
surgical procedures, are urgently needed, and to achieve this objective multiple centers have to collaborate.
Current meta-analyses evaluating the efficacy of treatments of RPL are generally pooling very heterogeneous
patient populations and treatments. It is recommended that future meta-analyses look at subsets of patients and
treatment protocols that are more combinable. (Fertil Steril 2005;83:82139. 2005 by American Society for
Reproductive Medicine.)
Key Words: Abortion, anticardiolipin, HLA-G, recurrent miscarriage, uterine fibroids, recurrent pregnancy loss

There is no doubt that the introduction of a series of new

assisted reproduction technologies (ART) has recently
greatly improved the treatment options available for infertile
couples. Most of these couples will now be offered treatments that are generally accepted and often evidence-based.
Unfortunately, the situation is much less clear regarding
couples with recurrent pregnancy loss (RPL), who are often
treated as second-class infertility patients in the health care
system. Accurate prevalence figures are not available, but it
has been estimated that 2%5% of women have RPL, deReceived August 23, 2004; revised and accepted December 14, 2004.
The workshop inspiring this article was supported by grants from the
European Society for Human Reproduction and Embryology and the
Danish National Research Agency.
Reprint requests: Ole B. Christiansen, M.D., Ph.D., Fertility Clinic 4071,
Rigshospitalet, Blegdamsvej 9, Copenhagen DK-2100, Denmark. (FAX:
45-35-45-49-46; E-mail: obc@pregnancyloss.dk).

0015-0282/05/$30.00
doi:10.1016/j.fertnstert.2004.12.018

fined as three or more consecutive losses of intrauterine

pregnancies before the 28th gestational week (1, 2).
Although the array of diagnostic tests and possible therapeutic interventions in the management of RPL have grown
significantly, in 1992 a U.S. study (3) showed that the total
live birth rate had not increased in a cohort of RPL patients
from 1987 to 1991 as compared with a similar group from
1968 to 1977.
The motivation for organizing a workshop under the auspices of the European Society for Human Reproduction and
Embryology with the title Evidence-Based Investigations
and Treatments of Recurrent Pregnancy Loss was the fact
that many of the large number of diagnostic tests that have
become available for the investigation of patients with RPL
are probably of doubtful value and need proper evaluation
and standardization. Furthermore, many treatments also need

Fertility and Sterility Vol. 83, No. 4, April 2005

Copyright 2005 American Society for Reproductive Medicine, Published by Elsevier Inc.

821

proper evaluation concerning their therapeutic value and

possible risks.
One reason for the slow progress in RPL research might
be related to the fact that RPL is a complex area in which
information from many disciplines, such as gynecology,
genetics, epidemiology, occupational medicine, immunology, hematology, and endocrinology, are to be integrated to
ensure that the research is valid. This collaboration among
many disciplines unfortunately has only been established
infrequently, with the result that most investigations in this
area have been narrowly focused, lacking an integrated approach to the subject.
A key element in this workshop was to let experts from the
different disciplines present their specific views on the
causes and treatments of RPL and engage the audience in
discussion to attempt a clarification on what is the state-ofthe-art knowledge and where consensus can be reached.
Gynecologists, obstetricians, and fertility specialists from
18 countries participated in the 3-day workshop held in
Denmark in 2002. The following review summarizes the
conclusions from the discussions undertaken between the
participants during the workshop and the extensive follow-up
discussions between the authors that have continued until now.
RESEARCH METHODOLOGY AND EPIDEMIOLOGY OF
RELEVANCE IN RPL STUDIES
Readers of the extensive literature about RPL often become
confused owing to the contradictory and ever-changing
views and results that are being published. Many of these
controversies are caused by the very different estimates of
the frequencies of RPL risk factors in patients and controls,
of the effect of these risk factors on pregnancy outcome, and
of the efficacy of various treatments. An important reason for
the controversy in this area is the apparent lack of appreciation of the many methodologic pitfalls threatening valid
research in the area of RPL. Following is a review of some
of the pitfalls inherent in this area of research with different
study designs.
CaseControl Studies
In case control studies, flaws can occur during the sampling
of cases and controls. Patients can be incorrectly sampled
because of an incorrect RPL diagnosis or ascertainment bias.
Furthermore, flaws can occur in relation to the tests carried out.

struations in the past as early miscarriages. More miscarriages might also be reported owing to the womans wish to
be offered investigations and treatment. Only 71% of miscarriages reported by non-RPL women in a questionnaire
could be verified in hospital records (4).
Biochemical pregnancies (pregnancies documented only
by a positive urine or serum hCG test) constitute a considerable proportion of some RPL patients previous pregnancy
history. Some of these pregnancies might be spontaneous
resorbed ectopic pregnancies or very early implantation failures due to genetically abnormal embryos, according to
currently available tools. The etiologies of recurrent biochemical pregnancies might thus be different from those of
clinical pregnancy losses. Thus, inclusion of patients with a
large proportion of biochemical pregnancies in clinical studies of RPL would be expected to diminish the estimate of a
maternal risk factor in case control studies and the treatment
effect in controlled clinical trials.
Many studies have included women with only two previous miscarriages, which often might be a chance phenomenon caused by de novo fetal chromosome abnormalities
rather than a recurrent maternal factor (5). Including women
with only two early miscarriages in the study will in most
cases dilute the estimate of the risk factor (in case control
and cohort studies) or the treatment effect in controlled
clinical trials. This is supported by findings that the frequency of many immunologic risk factors (6, 7) and the
possible effect of immunotherapy increases (8) and the frequency of chromosomally abnormal abortions declines (9)
with the number of previous pregnancy losses.
Ascertainment Bias. Ascertainment bias occurs when patients referred to clinics with special interests are deliberately or unconsciously selected because of that clinical feature on which the clinics interest is focused. Such patients
therefore are not representative of the general RPL population. For example, RPL patients investigated in clinics with
expertise in coagulation and antiphospholipid antibodies
might comprise an excess of women with antiphospholipid
antibodies (10) because referring centers preferentially refer
patients who in addition to obstetric problems also have
suffered thromboembolic episodes or expressed lupus-like
symptoms.

Incorrect RPL Diagnosis. Women can be diagnosed erroneously as having RPL owing to faulty recall of the pregnancy
history, classification of biochemical pregnancies as miscarriages, and the investigators failure to adhere to the generally accepted criteria for RPL.

Sampling of Controls. In case control studies, the quality of

the control group is just as important as that of the case
group. Sampling of controls is subject to confounding, mismatch with regard to pregnancy-related variables, and ascertainment bias. Age is typically an important confounding
factor because it is associated with both the risk of developing RPL and the occurrence of many serologic abnormalities
(e.g., autoantibodies).

An example of information (recall) bias is that, owing to

their increased attention on pregnancy and miscarriage,
women with only two previously confirmed miscarriages
might be more prone to interpret and report delayed men-

Blood parameters are often investigated at different stages

of pregnancies in patients and controls but still compared by
statistical methods. Many immunologic (1114) and coagulation factors change during pregnancy and after a pregnancy

822

Christiansen et al.

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

that progressed beyond the first trimester. Therefore, unreliable results can be produced if blood samples from pregnant
RPL patients are compared with samples from women who
are not pregnant or in a different stage of pregnancy.
Ascertainment bias often occurs when controls are sampled. Normal women are difficult to sample; therefore,
individuals consulting the hospital for reasons other than
RPL are often used as controls. Ascertainment of controls by
this method is often biased or subject to confounding. The
controls might differ from the RPL patients by the occurrence of a disorder that can affect the variable being evaluated, or they might differ by social class or by potential
hazardous exposures (tobacco, alcohol, caffeine, medicine).
Testing Procedures. Testing is often done during and immediately after miscarriage. Miscarriage might induce an
inflammatory reaction and endocrine changes that locally or
systemically can affect immunologic variables, such as cytokines (15, 16) or antibodies. Abnormal findings after miscarriage therefore might be the result of miscarriage rather
than its cause (17). It is therefore important to be careful in
establishing causality from abnormal findings in samples
taken during and just after miscarriage. The best evidence for
causality comes from samples taken remote from miscarriage.
In many studies, numerous statistical comparisons are
carried out without information being given concerning the
prior hypothesis being tested. Defining statistical significance as a P value .05 will result in every 20th statistical
comparison being significant by chance. Without a clearly
defined prior hypothesis, statistically significant post hoc
findings should not be overstated. Multiple statistical testing
should use appropriate multiple comparison methods to
avoid incorrect inferences from being drawn.
Cohort Studies: Differential Misclassification
Differential misclassification is an uneven intensity of monitoring of exposure to risk factors or outcomes in the different cohortsa problem often reported in occupational medicine.
Biased intensity of monitoring is also a problem in research in human reproduction. For example, owing to the
standard procedures in most ART clinics of measuring serum -hCG 14 days after ET, more biochemical pregnancies
will be registered in cohorts of ART patients compared with
women who conceive spontaneously.
An example of the impact of biased intensity of monitoring of pregnancy outcomes comes from the comparison of
two cohorts of RPL patients who were found suitable and
agreed to participate in placebo-controlled trials of treatment
with intravenous Ig in Sweden (18) and Denmark (19). In the
first trial, the patients were included in the trial in the next
pregnancy when fetal heart action could be demonstrated by
ultrasound in gestational weeks 6 8. In the second trial, the
patients had to call the trial center as soon as a pregnancy test
Fertility and Sterility

was positive, and they were included immediately in the trial

within 5 days after the missed menstrual period. The two
cohorts behaved very differently. In the Swedish study 51%
of the patients were classified as not achieving pregnancy,
and only 7% had pre-embryonal miscarriages before week 6.
In the Danish trial only 15% did not report pregnancy in the
study period, and 22% of the pregnancy losses happened
before week 6, prior to the possibility of demonstrating fetal
echoes by ultrasound. This very big difference in the frequencies of nonconception and pre-embryonal pregnancies
between the two trials probably reflects the different intensity of monitoring pregnancy outcomes due to the different
times of inclusion.
Treatment Trials
Unfortunately, there is a paucity of good-quality treatment
trials in RPL, and this limits the ability to make confident
recommendations about care in pregnancy. There are many
issues that have to be addressed in therapeutic trials before
inferences can be made that are reliable. Some of these
issues will be discussed below.
Design Issues. The importance of secure randomization,
concealment of treatment allocation, adequate sample size,
blinding, completeness of follow-up, stratification for important covariates, and adequate assessment of outcome are well
known to clinical trial methodology and do not require
in-depth discussion in this review. It is only by ensuring that
a study is valid that one can make confident inferences from
the results obtained.
To make studies mutually comparable, it is necessary to
use the same terminology for the description of patients and
their miscarriages.
Definition of Miscarriage. The term miscarriage (or abortion) is used to describe a pregnancy that fails to progress,
resulting in the death and expulsion of the embryo or fetus.
The generally accepted definition stipulates that the fetus or
embryo should weigh 500 g (20), a stage that corresponds
to a gestational age of 20 weeks. Unfortunately, this definition is not used consistently, and pregnancy losses at
higher gestational ages are also, in some studies, classified as
abortion instead of stillbirth or preterm neonatal death. Thus,
from a definition perspective, it is important to characterize
the population being studied so that comparisons across
therapeutic trials can be made more appropriately and reliably. Consensus on this issue is urgently required.
Effect of Male Partner. It is well known that some women
might have RPL with one male partner and not with another.
Therefore, it is important when evaluating treatment efficacy
to ensure that the sample is homogeneous from this perspective (i.e., consecutive miscarriages with the same partner
should be stipulated as an inclusion criterion).
Subgroups of RPL. The pregnancy history in women with
RPL might include pregnancies that have ended in live birth.
Thus, three different groups can be identified that should be
823

assessed separately because the risk of subsequent miscarriage among these groups varies (1): [1] the primary RPL
group consists of women with three or more consecutive
miscarriages with no pregnancy progressing beyond 20
weeks gestation, [2] the secondary RPL group consists of
women who have had three or more miscarriages following
a pregnancy that progressed beyond 20 weeks gestation,
which might have ended in live birth, stillbirth, or neonatal
death, and [3] the tertiary RPL group, which has not been
well characterized or studied, consists of women who have
had at least three miscarriages that are not consecutive but
are interspersed with pregnancies that have progressed beyond 20 weeks gestation (and might have ended in live
birth, stillbirth, or neonatal death).
From these three different groups, it is evident that the
study population being evaluated should be clearly specified
because the prognosis for a successful outcome undoubtedly
will be influenced by the group being selected. The current
approach of lumping all three groups together will not allow
the effect of the experimental intervention to be detected
easily. In the Cochrane meta-analysis of the effect of immunotherapy in RPL (21), studies that are very heterogenous
are combined in a common meta-analysis including patients
with both primary and secondary RPL. Because there are
strong indications that allogeneic lymphocyte immunization
might only have a beneficial effect in women with primary
RPL, and because intravenous Ig displays its strongest effect
in women with secondary RPL, it is important that metaanalyses of the efficacy of the respective treatments be
undertaken in the relevant patient subgroups.
Untreated Control Group. In many trials of treatments in
RPL, pregnancy outcome in the treated group is compared
with the outcome in a group of historical controls. Such
controls might comprise untreated patients from studies undertaken decades ago in another setting. This method is
likely to produce flawed results because the ascertainment
and the pregnancy-monitoring procedures of the controls
might be very different from those of the treated patients.
Another variation of the use of historical controls is the
comparison of pregnancy outcome of the patients subjected
to the new treatment with the outcome in their previous
untreated pregnancies. This method is often used in studies
of surgical procedures to prevent RPL. However, women
with RPL were sampled because of their previous pregnancy
losses, and some of them might have suffered them merely
by chance. They will still have a very favorable chance of
successful pregnancy without any treatment, owing to a
phenomenon called regression to the mean. In women
included in a trial because of RPL, the accumulated past
pregnancy success rate will often be 10%. Comparing this
low pretreatment success rate with an 85% posttreatment
success rate (e.g., after cerclage) will not require many
patients for obtaining statistical significance (22). However,
by using historical controls, most interventions can be
824

Christiansen et al.

proved to be highly efficient in the treatment of RPL (23).

This apparent improvement of outcome can partly be attributed to the effect of regression to the mean.
In conclusion, proper evaluation of interventions in RPL
can only be accomplished in double-blinded, randomized
controlled trials in which the allocation groups are similar
with regard to all important prognostic determinants except
for the intervention that is tested.
Prognostic Determinants: Number of Previous Miscarriages.
For many years, the mathematic estimates of miscarriage
rates were used as control rates against which the efficacy of
various therapeutic regimens introduced to prevent miscarriage were assessed. The reliability of these rates was challenged after evidence from a number of clinical studies
suggested that the miscarriage rate after three consecutive
miscarriages was substantially lower than had been predicted
by the earlier models (1). Nevertheless, despite the varied
methods of ascertainment, the results of the studies showed
remarkable consistency in finding an increasing risk of miscarriage as the number of previous miscarriages increases.
The effect of prior losses on the subsequent probability of
live birth was confirmed with the data from the placebo arm
of studies in unexplained RPL and provided a quantitative
estimate of the risk (24). It is clear from this evidence that
the number of previous miscarriages is an important covariate, which has to be taken into account when planning
therapeutic trials.
The ideal trial should have stratification for the number of
previous miscarriages, with randomization between control
and experimental treatments within each stratum. To date,
such a study has not been undertaken. It is quite likely that
by stratifying the sample by number of previous miscarriages that the effect of the experimental intervention will
become more easy to demonstrate in those women with
higher numbers of previous miscarriage than in those with
fewer previous miscarriages because the control event rate is
so much lower in the former group (8, 24).
Prognostic Determinant: Female Age. Miscarriage-prone
women have more pregnancies and have their pregnancies
occur at a later age than successful reproducers. Because
gravidity is closely linked to female age, it is possible that
the increased risk of miscarriage with gravidity can be ascribed, in part, to the effect of maternal age, particularly in
view of the fact that chromosomal anomalies are associated
with advancing maternal age. It is well established that the
risk of miscarriage resulting from trisomic conceptuses increases with maternal age, especially after 35 years. Thus,
clinical trials of treatment efficacy must take female age into
consideration during the design of the trial by using stratification for this covariate.
Issues Relating to Treatment Protocols and the Evaluation
of Effect. Of crucial importance for the evaluation of the
results of a treatment trial are the timing, doses, and the
numbers of treatments given, possible termination of the trial

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

prematurely after interim analyses, the possible inclusion of

several pregnancies from the same patient, and the results of
karyotypic analysis of products of conception.

planned, the stopping rules should be very specific and

should only be used when there is clear evidence of benefit
or harm.

There is no standardization in many of the treatment

protocols with respect to the onset of treatment. For example, intravenous Ig has been administered preconceptionally
in some studies, whereas in other studies treatment is commenced only after confirmation of the pregnancy (25).
Sometimes treatment is instituted only after fetal cardiac
activity has been demonstrated, as has been observed in
some of the thrombophilia treatment trials. The likelihood of
a successful outcome without treatment once fetal cardiac
activity has been demonstrated is relatively high and will
result in efficacy studies failing to detect a sizable treatment
effect with the experimental intervention.

In some case series and controlled trials, several pregnancies from the same patients were included. The commonly
used statistical methods, such as the 2 test and t-test, require
that all measurements are independent. Pregnancy outcomes
in the same women are not independent variables; therefore,
inclusions of several pregnancies from the same women
require suitable sophisticated tests.

Another methodologic concern in efficacy trials is the

practice of limiting the number of cycles of preconceptional
treatment patients might undergo before they are withdrawn
from the study, despite the fact that they have been randomized into one of the two comparable groups. The overall
result becomes biased by excluding the patients who have
failed to conceive within a specified (usually short) period of
time after randomization. Women with RPL, compared with
those with sporadic miscarriage, have a longer interpregnancy conception interval (i.e., length of time taken for
conception to occur in women attempting pregnancy after
the miscarriage) (26, 27). The pathologic mechanism for this
observation is not clear. One possible hypothesis is that fear
of miscarriage in a subsequent pregnancy induces significant
stress that might adversely influence the hypothalamus and
result in subtle ovulatory dysfunction (1). Thus, it is clear
that the evaluation of treatment commenced preconceptionally will require many cycles of observation before pregnancy can be achieved. For this reason and for the methodologic reasons discussed, women enrolled into such
randomized trials should not be withdrawn just because they
fail to conceive within a specified period.
Interim analyses are sometimes carried out during a placebo-controlled trial, and the trial is stopped when a significant result for or against a treatment effect is reached.
However, if premature termination of the study after interim
analyses occurs, there is a great risk that the trial will be
concluded at a point at which the difference between the
allocation groups has reached an extreme fluctuation. The
result at this stopping point is not likely to be typical for the
result if the study were to be conducted to the end. The trials
by Jablonowska et al. (18) and Ober et al. (28) might be
affected by this problem because the very high success rate
of the placebo group in the former study and the low success
rate of the lymphocyte-immunized group (46%) in the latter
trial might reflect a short extreme fluctuation resulting in a
decision to terminate the study after interim analysis. In
conclusion, interim analyses and premature termination of
trials should be avoided, and such trials should be clearly
identified in any meta-analyses. If interim analyses are
Fertility and Sterility

The possibility that a miscarriage after treatment is the

result of aneuploidy must be investigated, particularly in
efficacy trials. Without this information it is impossible to
ascertain whether the pregnancy loss is the result of treatment failure or a de novo chromosomal anomaly. The magnitude of the size of the treatment effect will be affected
without correction for the aneuploidy factor. The improvement in ultrasonography technology now provides images
with better resolution, thereby allowing the diagnosis of
pregnancy failure to be made much earlier, a process that is
assisted with hormone assays. Thus, it is possible to have
access to viable and noninfected fetal and trophoblast tissue
that can be submitted for karyotypic analyses in a way that
does not jeopardize the cell culture. Furthermore, improved
techniques in cytogenetics have permitted more accurate and
reliable assessments of the products of conception. Given
these improvements in our diagnostic ability, it is even more
important that every effort be made to study the products of
conception in every case of miscarriage in therapeutic trials
so that a more valid evaluation can be made regarding the
efficacy of the experimental treatment.
Intention-to-treat analysis is a necessary first step in the
analysis of randomized controlled trials, but an explanatory
analysis is an important second step. In this step we can
exclude aneuploid pregnancies. Calculations of statistical
power and sample size can incorporate an estimated aneuploidy rate, but this is generally not done. It might be ideal,
but it will increase the sample size.
Summary. The most important methodologic issues considered to affect and potentially flaw studies of RPL are listed
in Table 1. Apart from the obvious study design issues, it is
important to select the sample appropriately by clearly defining the population through strict inclusion criteria. In
treatment trials, stratification for relevant covariates (especially number of previous miscarriages) must be undertaken
ad hoc. Onset of treatment (preconceptionally or postimplantation) should be supported by a biological rationale. Sufficient time should be provided for the intervention to have
been adequately tested. Postrandomization withdrawals
should be avoided. The outcome should be clearly documented, and every effort should be made to submit products
of conception for karyotypic analysis. Only through such a
comprehensive approach can the resulting evidence be used
appropriately to direct care of couples with RPL.
825

TABLE 1
Methodologic factors to evaluate in studies of recurrent pregnancy loss.
Factor to evaluate

Effect on study outcome

Definition of RPL as 2 miscarriages

Decreases difference in frequency of factor

studied in CCS or treatment effect in RCT
Increases prevalence of factor studied
Increases difference in prevalence of factor
studied in CCS or treatment effect in RCT
Increases recurrence of RPL
Increases difference in prevalence of factor
Increases treatment effect in treatment trials
Renders comparisons between studies and
meta-analyses difficult
Decreases or increases treatment effect in RCT
Decreases treatment effect in RCT
Decreases treatment effect in RCT

Ascertainment bias
Selection of controls
Advanced age of patients
Uneven monitoring of two cohorts studied
Historical controls
Poor characterization of miscarriage and
subgroups of RPL
Premature termination after interim analysis
Inclusion after detection of fetal heart action
No exclusion of aneuploid abortuses

Note: RPL recurrent pregnancy loss; CCS case control studies; RCT randomized control trials.
Christiansen. Management of recurrent pregnancy loss. Fertil Steril 2005.

CHROMOSOMAL ABNORMALITIES IN RPL

Only 4.7% of couples with two or more abortions include a
carrier of a balanced reciprocal or Robertsonian structural
abnormality (29), but chromosomal analysis of RPL couples
must still be considered a part of the routine investigation
(Table 2).
It is well documented that fetal de novo chromosomal
abnormalities are a major cause of sporadic first-trimester
spontaneous abortions, and some cases of RPL thus might be
caused by repeated fetal chromosome abnormality. Cytogenetic evaluations of specimens of sporadic abortions have

revealed an overall incidence of chromosomal abnormalities

of 50%70% (30 34). The most common cause of spontaneous abortions is de novo numerical abnormalities, in particular autosomal trisomies for chromosomes 13, 14, 15, 16,
21, and 22, followed by monosomy X (34 36). On the other
hand, autosomal monosomies are rarely found in spontaneous abortions and considered to be responsible for preclinical abortions (30, 31, 34). This mechanism of natural selection
might also operate during preimplantation embryogenesis, with
a progressive loss of abnormal embryos at specific stages in
early development through developmental arrest and degeneration of abnormal embryos.

TABLE 2
Potential etiologic factors in the causation of recurrent pregnancy loss.
Factor
Parental genetics
Uterine abormalities
Uncontrolled thyroid disease
Uncontrolled diabetes
Polycystic ovary syndrome
Antithyroid antibodies
Antiphospholipid antibodies
Factor V Leiden mutation
Th1 cytokine bias
Increased NK cell cytotoxicity
Maternal HLA alleles
Parental HLA sharing

Association with RPL

Causation of RPL

Definite
Definite
Probable
Probable
Definite
Doubtful
Definite
Definite
Probable
Probable
Probable
Doubtful

Definite
Probable
Probable
Probable
Probable
Doubtful
Probable
Probable
Probable
Probable
Probable
Doubtful

Note: RPL recurrent pregnancy loss; Th1 T helper 1; NK natural killer; HLA human leukocyte antigen.
Christiansen. Management of recurrent pregnancy loss. Fertil Steril 2005.

826

Christiansen et al.

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

With IVF and preimplantational genetic diagnosis (PGD),

the in vitro developmental ability of human embryos at these
stages can be observed, enabling us to learn about their
behavior and perhaps about the mechanisms involved in the
genetic causes of RPL (37). A study involving PGD on
embryos from 106 couples with RPL classified women into
two groups according to their age: 37 years or 37 years.
A control group of 28 couples undergoing PGD because of
the risk of sex-linked diseases, and without other infertility
problems, was included to compare clinical results and the
incidence of abnormalities for each chromosome with the
RPL group. The fluorescence in situ hybridization and PGD
protocols have been previously described (37, 38).
In the RPL group, all embryos were diagnosed as abnormal in 33 cycles, and no ET was performed, and in the other
95 cycles, at least one embryo was transferred, resulting in
31 pregnancies (32.6%), with an implantation rate of 25.3%
and a miscarriage rate of 12.9%. In the control group, 35
cycles were included and ET was performed in 31. Nine
clinical pregnancies were achieved, and no miscarriages
were observed.
In the RPL group, 532 of 764 embryos were abnormal
(69.6%), whereas in the control group 97 of 215 embryos
(45.1%) were abnormal (P.0001). Results were also compared separately in the two age subgroups and showed an
increased incidence in chromosomally abnormal embryos in
both (more evident, however, in younger patients [P.0001]
than in the aged women [P.046]). All the abortion specimens were cytogenetically analyzed: in three of the cases, a
normal male karyotype was identified, and in one case a
trisomy for chromosome 15 was observed.
These results confirm preliminary findings that couples
with RPL seem to produce chromosomally abnormal embryos in significantly higher percentages than those not having this reproductive problem (38, 39). Moreover, in 25.8%
of the cases, the incidence of chromosomal aberrations affects all the embryos, and the percentage of abnormal embryos is quite similar in subsequent attempts. Thus, the use
of PGD in these couples, after appropriate workup has ruled
out other cases of RPL, might be advisable not only as a
therapeutic but also as a diagnostic tool, as long as an
acceptable implantation rate per embryo replaced is reached,
providing evidence that PGD would not damage the embryos.
An open question is why the frequency of chromosomal
abnormality was not increased with increased age in the RPL
patients. In controls, the frequency of chromosomal abnormalities increased, as expected, with increased age. It is
possible that the basic frequency of chromosomal abnormalities was so high in RPL patients that no age differences
could be detected.
It cannot be ruled out that the high frequency of chromosomally abnormal embryos in the study might be a result of
the IVF procedure. The gonadotropin-stimulated superovuFertility and Sterility

lation might stimulate a cohort of abnormal eggs that would

not be ovulated in a spontaneous cycle. However, this would
be difficult to prove because it will be almost impossible to
get eggs from normal cycles.
It can be claimed that PGD in RPL is not cost effective.
After four previous miscarriages, the spontaneous birth rate
in RPL patients is still 50%; although the selection of embryos by PGD might decrease the miscarriage rate, only 33%
become pregnant after each PGD cycle. This might be acceptable in countries with high birth rates, where the women
are willing to attempt many pregnancies. In countries with
low birth rates, such as Spain, PGD for improving the
prognosis in RPL might be cost effective.
NONCHROMOSOMAL GENETICS OF RPL
Few studies have investigated the genetics of RPL in families of RPL couples with normal chromosomes. Three relevant studies have been published, all of which found that the
RPL prevalence in first-degree relatives of women with
unexplained RPL was significantly higher than in controls
(40 42). In all the studies, the risk of RPL in first-degree
relatives was approximately six times higher than the risk in
the background population. It can be calculated (43) that the
observed RPL prevalence in first-degree relatives is in closest agreement with a polygenic mode of inheritance with
RPL, occurring when a patient has inherited a sufficient
number of risk factors to exceed a threshold for disease
development (Fig. 1). There is evidence that one of the
factors determining the increased risk of RPL among sisters
of RPL patients is linked to the human leukocyte antigen
(HLA) system (44), but now a series of other relevant
congenital defects in coagulation or immune function have
been discovered that might explain the increased miscarriage
risk among relatives.
UTERINE ANATOMIC ANOMALIES IN RPL
Several uterine factors have often been reported as associated with RPL, although there is not always sufficient proof
of a causative role.
Congenital Uterine Anomaly
Of the various congenital structural uterine anomalies, the
septate uterus is the most common. There is sufficient observational data to suggest that a septate uterus is associated
with an increased risk of miscarriage due to impairment of
implantation (45). The septum might be removed hysteroscopically with the use of scissors or electrical or laser
energy. Uncontrolled studies have reported a better pregnancy performance after surgery than before surgery in the
same patients, but no prospective trial has been carried out in
which appropriate randomization to surgery and no surgery
was undertaken.
827

FIGURE 1
Figures are illustrating the traditional mono-etiological thinking (pie) and the recommended multifactorial
thinking (column) regarding the pathogenesis of recurrent pregnancy loss. NK-cells natural killer cells;
Th1 cytokine T helper cell type 1 cytokines; MBL mannan-binding lectin; HLA human leukocyte
antigen.

Christiansen. Management of recurrent pregnancy loss. Fertil Steril 2005.

Ashermans Syndrome/Endometrial Fibrosis

Ashermans syndrome is an acquired condition, which is due
to the presence of posttraumatic intrauterine adhesions partly
or completely obliterating the uterine cavity. Endometrial
responsiveness to the steroid hormones is reduced in areas
affected by intrauterine adhesions or fibrosis. The prognosis
depends on the extent of endometrial damage. Although the
adhesions might be divided hysteroscopically, endometrial
function might be permanently damaged in the presence of
extensive, dense fibrosis, which is associated with a poor
prognosis.
Uterine Fibroid
There is convincing observational data from several IVF
series (46 51) to suggest that reproductive outcome is significantly compromised by submucosal fibroids, modestly
compromised by intramural fibroids, and possibly compromised by subserosal fibroids. There is an association between advanced age and uterine fibroids, thus age might be
828

Christiansen et al.

a confounding factor when the real impact of fibroids on

subsequent reproductive performance is to be assessed. From
nonrandomized studies, there seems to be some evidence to
suggest that removal of submucosal fibroids reduces the
miscarriage rate and some evidence that removal of intramural fibroids also reduces miscarriage rates (52). For gynecologists who are not very experienced in laparoscopic
surgery, it is advised to remove fibroids by laparotomy
because the uterine scar should be carefully and appropriately closed to avoid adhesions and bleeding. However, it is
important to bear in mind that no prospective study has been
carried out in which random allocation to surgical removal
of fibroids or no surgery was done.
SYSTEMIC AND LOCALIZED ENDOCRINE
ABNORMALITIES IN RPL
Several endocrinologic factors might be associated with
RPL.

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

Thyroid Disease
For some years, hyper- or hypothyroidism has been thought
to be associated with RPL. Current evidence seems to suggest that treated thyroid dysfunction is not associated with
RPL (53). A recent study suggested that 2% of women with
a midtrimester loss were found to be hypothyroid (54).
Because measuring TSH is a sensitive and inexpensive tool
to assess thyroid function, it should be considered in the
evaluation of RPL for women with any signs or symptoms of
thyroid dysfunction.
Diabetes Mellitus
Current evidence suggests that well-controlled diabetes mellitus is not associated with RPL (53). In a recent study,
investigators looked at the frequency of insulin resistance in
unselected women who presented to a clinic with a diagnosis
of RPL. Insulin resistance was identified significantly more
frequently in women with RPL when compared with age-,
race-, and body mass indexmatched controls (55).
Hypersecretion of LH
Initial reports suggested that a high level (10 IU/L) of LH
on cycle day 8 was associated with an increased risk of
miscarriage (56). Later studies, however, could not confirm
the impact of high follicular LH on the risk of miscarriage
(57), and the suppression of LH with GnRH analogues did
not affect the miscarriage risk (58).

Metformin treatment of infertile women with polycystic

ovary syndrome (PCOS) decreases insulin resistance and
might improve the implantation rate after IVF (63, 64). It is
still uncertain whether metformin decreases the rate of miscarriage in patients with PCOS, because no properly designed placebo-controlled studies concerning this treatment
have been conducted.
High Androgen Levels
Two recent studies have shown that androgen levels in the
follicular phase are higher in women who have RPL than in
normal fertile controls (65, 66). The observation seems to be
independent of the association between PCOS and RPL.
Further studies are required to examine the relationship
between hyperandrogenism and RPL.
Hyperprolactinemia
It is debatable whether hyperprolactinemia is associated with
RPL. In a recent study (59), 3 of 122 subjects (2.5%) with
RPL had marginally elevated PRL levels, and one of them
had significantly high (1,000 IU/L) PRL. Additionally, in
a subgroup of women with RPL (n 23) who had daily
plasma PRL measured in the midfollicular phase to early
luteal phase, there was no evidence of a preovulatory rise of
PRL. The data suggested that there is no firm evidence of an
association between hyperprolactinemia and RPL.

The prevalence of high LH levels in the follicular phase of

women with recurrent miscarriage, defined as plasma concentration 10 IU/L, varies enormously, from 0 to 37%
(59). With the use of improved methodology, including daily
blood samples collected in the mid- to late follicular phase,
and with the use of newer, more precise immunometric
assays, hypersecretion of LH has been found to occur in only
8% of women (59). Studies have also suggested that a
too-low LH level in the follicular phase could affect miscarriage rates negatively in IVF cycles (60); however, there is
no evidence that this is a cause of RPL.

Luteal Phase Defects

Traditionally, low serum P in the midluteal phase or an
endometrial biopsy out of date according to Noyes criteria
(67) have been taken as evidence for a luteal phase defect
that could cause failed implantation and RPL. However,
previous studies in this area have not dated the blood samples or biopsies according to the LH peak or ultrasonic
demonstration of ovulation. Recent studies have questioned
the clinical relevance of dating of the endometrium by
Noyes criteria.

Polycystic Ovaries
In a recent study (61), the prevalence of polycystic ovaries
(PCO) among women with RPL was reported to be 40.7%
(852 of 2,199), but in another study of 102 women with RPL
who underwent transvaginal ultrasonography, only 8 (7.8%)
had the typical features of PCO (59). In the study by Rai et
al. (61), the live birth rate was similar in women with PCO
(60.9%) and in women with normal ovarian morphology
(58.5%). The above evidence leads to the conclusion that
PCO pathology is not predictive of pregnancy loss among
ovulatory women with RPL who spontaneously conceived.

Primary Endometrial Defect

In many cases, a luteal phase defect is caused by a primary
defect of P responsiveness of the endometrium or a secondary defect caused by, for example, adhesions or fibroids, in
which cases the luteal phase P level is normal. A number of
morphologic and immunohistochemical studies suggest that
a primary endometrial defect might be present in more than
20% of subjects with RPL (68).

A study on the impact of laparoscopic ovarian diathermy

on reproductive performance in patients with PCO (62)
suggests some benefit of this treatment in infertility, but
randomized studies are needed in RPL patients.
Fertility and Sterility

AUTOANTIBODIES AND COAGULATION DEFECTS IN RPL

Autoantibodies
Numerous autoantibodies have been reported to be associated with RPL. Most of the literature has dealt with antiphospholipid antibodies (APLs), which are antibodies directed
against negatively charged phospholipid molecules consti829

tuting parts of the membranes of the cell surfaces and cell

organelles. There is consensus that the APL lupus anticoagulant (LAC) and anticardiolipin antibodies (ACA) should be
looked for in RPL patients because [1] these can be found
with significantly increased prevalence in RPL patients, [2]
the pregnancy prognosis of patients with these antibodies
seems to be poorer than in those without, and [3] the chance
of live birth seems to be improved if treatment with heparin
and low-dose aspirin is given (69 71). Unfortunately, there
is no consensus regarding how to define an APL-positive
RPL patient: very different laboratory methods and cut-off
values for positivity are used, and some centers require at
least two positive tests for the diagnosis (72), whereas others
are content with one. A recent comparison of APLs evaluated by 10 different specialty laboratories was an important
step in developing a consensus on APL test results and
interpretation (73).
However, there is considerable disagreement regarding
the mechanisms of action of APLs in women with RPL: are
the antibodies acting on coagulation factors or placental
endothelial cells and thereby promoting thrombosis of placental vessels; are they acting directly on villous or extravillous trophoblast cells, inhibiting implantation processes or
cell fusion; or are they only markers of other immunologic
processes harmful to the pregnancy (e.g., a T helper [Th]1
cytokine bias)?
Another important question concerning APLs that remains
unanswered is whether other APLs, such as antiphosphatidylserine and antiphosphatidylethanolamine, should be
looked for and whether anticoagulation treatment should be
offered to patients with these antibodies. Results from one
study (74) suggested that APLs other than LAC and ACA
are associated with RPL and display impact on the prognosis. It also seems that both unfractionated and low-molecular-weight heparin inhibit binding of ACA and antiphosphatidylserine to their respective phospholipid antigens in
vitro (75).
The prevalence of positivity for ACA and lupus anticoagulant varies significantly between studies. This can partly be
explained by different definitions of a positive ACA test, the
timing of the tests, and the screening procedures used. Some
clinics are only investigating IgG ACA and not IgM ACA
because the stability and the prognostic impact of IgG ACA
seems to be higher. However, 20% of APL-positive RPL
patients will only have IgM RPL, and until further studies
have clarified the importance of this antibody isotype, it is
advisable to continue to screen patients for IgM ACA.
Another important factor explaining some of the heterogeneity between studies is the fluctuation of ACAs with
timea fluctuation that is particularly large during pregnancy. In one study (69), RPL patients were tested 6 weeks
after the last pregnancy loss and again in early pregnancy,
gestational weeks 20 and 30, and after pregnancy. In only
approximately 60% of the women the titers remained constant. There seem to be three subgroups of patients: those
830

Christiansen et al.

with constantly positive findings, both before and during

pregnancy; those with borderline titers (10 20 IgG-phospholipid/IgM-phospholipid [GPL/MPL] units), and those
who are completely negative. There is probably no need to
repeat testing or anticoagulate those who are completely
negative before pregnancy, but in those with borderline titers
before pregnancy new tests should be taken in early pregnancy, after which a decision can be made regarding treatment with heparin/aspirin or no treatment.
Screening for the lupus anticoagulant poses a special
problem. In many clinics, a patient with the lupus anticoagulant is a rare finding. This might partly be owing to the use
of activated partial thrombin time (APTT) as a screening test
for the lupus anticoagulant. The use of only standard APTT
as a screening test will miss half of the cases that are positive
for the lupus anticoagulant because it is too insensitive. This
can be overcome by the use of a more sensitive modification
of the APTT test (PTT-LAC) or by the use of two screening
tests: both the dilute Russells viper venom time and the
PTT-LAC.
Several studies have reported that a high frequency of
women with unexplained infertility or secondary infertility
after RPL are positive for APL (76, 77). However, the
chance of achieving pregnancy after a subsequent IVF attempt seems not to be related to the presence of APL (77),
and from the studies that have been undertaken it is not
justified to offer these patients anticoagulation treatment (76,
78).
A last relevant question is the significance of non-APL
autoantibodies, such as antithyroid antibodies and antinuclear antibodies, in RPL. Although these antibodies can
probably be found with increased prevalence in RPL patients, most studies have not been able to document that they
show a negative impact on subsequent pregnancy outcome.
Two prospective trials have failed to show any benefit from
intervention with low-dose prednisone (79) or any difference
in outcome if patients with normal thyroid function and
thyroid antibodies are followed (80). It is advisable that
patients with antithyroid antibodies are monitored for thyroid function during pregnancy (e.g., TSH measurements),
but no special intervention should be offered.
Acquired and Congenital Thrombophilias
Investigators in one study evaluated 80 women with three
consecutive pregnancy losses, fathered by the same partner,
and whose prior evaluation for RPL had been negative
(Kutteh et al., unpublished data). Evaluation for thrombophilias included the following: antiphospholipid antibodies
(APA), LAC, protein C activity, protein S activity, antithrombin III activity, factor V Leiden mutation, activated
protein C resistance, prothrombin (factor II) activity, and
fasting plasma homocysteine. Thirty-two of eighty women
(40%) with unexplained RPL had at least one abnormal test
result from the thrombophilic workup, compared with 9

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

(11%) in the control group (P.01). The women with unexplained RPL differed from controls, especially with regard
to hyperhomocystinemia, elevated APA or LAC, and the
presence of the factor V Leiden mutation. These results are
in concordance with another study that found that 66% of
RPL patients had at least one thrombophilic defect compared
with 28% of controls; however, in this study an accumulation of thrombophilic defects characterized RPL patients
because at least two defects were found in 21% of patients
and in 5.5% of controls (81). These data suggest that hypercoagulable states might be an important factor in RPL.
A series of 31 relevant studies judged to be of good quality
has now been subjected to meta-analysis, which concluded
that there is an increased prevalence of several acquired and
inherited thrombophilias in populations of women with unexplained RPL compared with women without a history of
adverse pregnancy outcome (82). Factor V Leiden mutation
was weakly associated with early RPL (odds ratio [OR], 2.0;
95% confidence interval [CI], 1.13.6) but strongly associated with late RPL (OR, 7.8; 95% CI, 2.8 21.7). The prothrombin G20210A mutation and protein S deficiency
seemed also to be associated with RPL, but the ORs had very
wide confidence limits owing to small studies, and more
studies are needed to document their association with RPL.
The meta-analysis could not document any association between methylenetetrahydrofolate mutation (a risk factor for
hyperhomocystinemia), protein C, and antithrombin deficiencies.
A second meta-analysis evaluated the association of factor
V Leiden and prothrombin gene mutation G20210A in
women with RPL (83). The combined ORs for the association between RPL and factor V Leiden and between RPL and
G20210A were 2.0 (95% CI, 1.52.7; P.00l) and 2.0 (95%
CI, 1.0 4.0; P.03), respectively.
An unsolved question is whether RPL patients with congenital thrombophilic factors, such as the factor V Leiden
mutation should be treated with heparin and aspirin. There
are no placebo-controlled trials documenting the efficacy of
this treatment in these women. So far, there is only evidence
for the use of heparin and aspirin in APL-positive RPL
patients.
IMMUNE CELLS AND CYTOKINES IN RPL
One of the causes of pregnancy loss is thought to be the
immunologic rejection of the fetus due to a breakdown in the
mechanisms that normally prevent the maternal immune
system from becoming activated by the paternal antigens
expressed on the developing fetus. Recent studies have investigated the role of specific immune cells and molecules in
the cause of repeated miscarriage, by comparing their expression in peripheral blood, endometrium, and decidua of
women with RPL and control women.
Fertility and Sterility

Natural Killer Cells in the Blood and Endometrium

Several studies have shown that the immune cell distribution
within the endometrium and decidua has a profile different
from that in peripheral blood, in the former tissue consisting
mainly of CD56CD16-CD3- natural killer (NK) cells (84).
Endometrial expression of these cells increases during the
secretory phase of the menstrual cycle and in early pregnancy. Macrophages and T cells make up the remaining
leucocyte population, but their numbers show little variation
in the menstrual cycle. Studies comparing the expression of
these various cell populations in control and RPL women
have shown very varied and inconsistent results. Numbers of
CD56 NK cells can be either increased or decreased in
women with RPL, depending on the compartment that is
studied (85, 86). Peripheral blood and peri-implantation endometrial CD56 NK cells in early pregnancy are decreased, whereas numbers in the nonpregnant endometrium
are increased. A murine strain without NK cells can only
reproduce if NK cells are transferred to the mice (87). This
observation challenges the notion that NK cells are harmful
for pregnancy.
It is indeed unclear whether NK cells are good or bad,
particularly in humans. In mice, NK cells were initially
regarded as a potential threat to the fetus, because studies
showed increased numbers in abortion-prone mice. However, more recent studies have shown that the placenta in
NK-deficient mice is hypertrophic and is associated with
fetal death, suggesting that NK cells play a positive role. The
situation in humans is even less clear. Furthermore, because
many T lymphocytes also have NK cell markers, evaluating
the importance of NK cells becomes more complicated.
There seem to be no differences in total T cell numbers,
but studies have suggested that expression of peri-implantation endometrial or peripheral blood T cell subsets, such as
- T cells and the ratio of CD4 to CD8 T cells might be
altered in women with RPL (88, 89).
Cytokines
Cytokines are immune molecules that control cells in both
the immune and nonimmune systems. They are divided into
groups according to their function. Some cytokines produced
by Th1 lymphocytes include interleukin (IL)2, interferon
(IFN)-, and tumor necrosis factor (TNF)- and stimulate
cell-mediated immune responses. Other cytokines produced
by T helper 2 (Th2) lymphocytes include IL4 and IL10 and
stimulate antibody-mediated responses. Experiments in animals suggest that Th1 cytokines are detrimental to pregnancy and cause pregnancy loss, whereas Th2 cytokines are
beneficial and prevent pregnancy loss (90, 91). There is
some evidence for an abnormal balance of Th1 and Th2
cytokines in women with RPL (9294). Peripheral blood
mononuclear cells taken either before pregnancy or during
the first trimester of pregnancy from women with RPL
produce IL2 and IFN- (Th1 cytokines) when stimulated
with either trophoblast antigen or the mitogen phorbol my831

ristate acetate (PMA), whereas those from control women

produce IL4 and IL10 (Th2 cytokines) (92). In addition, T
cell clones produced from CD4 cells isolated from decidua
of RPL women produced less IL4 and IL10 than similarly
prepared cells from control women (93). Another group of
cytokines are proinflammatory cytokines, such as IL1, IL6,
and leukemia inhibitory factor. Studies have shown that
expression of these in peri-implantation endometrium is decreased in women with RPL (95, 96). Thus, there is some
evidence for differences in the expression of specific immune cells and molecules in the peripheral blood, endometrium, and decidua of women with RPL.
Most studies have not been able to find that genetic
polymorphisms resulting in low cytokine production are
associated with RPL. There could still be a problem of
producing adequate levels of the right cytokines in these
patients, but the failure could be at the level of transcription.
The hypothesis of Th1 and Th2 cytokine response as
being important for pregnancy is probably too simple and
rigid. It is important to look at the totality of cytokine
function at the place where things are happening in the
uterus. Both Th1 and Th2 cells can secrete what is normally
considered a Th2 cytokine, IL10, in humans but not in mice.
Indeed, human and murine reproductive and immune systems are very different, and extrapolation of findings from
the murine to the human setting should be done very cautiously. Furthermore, there is much evidence that the neurologic, endocrine, and immune systems interact. It is very
likely that psychological stress in pregnant women with RPL
results in increased levels of stress hormones that can produce miscarriage through immunologic mechanisms. Stressinduced fetal loss exists in mice. There is also evidence that
P can interact with proinflammatory cytokines.
To advance this area of research further and to limit the
variation seen in the results of different studies, care should
be taken both with respect to the compartment (peripheral
blood, endometrium, or decidua) in which the cells and
molecules are measured and the timing of the sampling, both
with respect to the menstrual cycle and pregnancy and
whether it is at the time of or just after miscarriage.
THE ROLE OF HLA IN RPL
Prospective studies in inbred populations clearly show an
influence of HLA genes or closely linked non-HLA loci on
reproductive processes. Prospective, population-based studies in the Hutterites, a population isolate of European origin
that has settled in the United States, showed significantly
increased fetal loss rates in couples who shared alleles for
HLA-B or HLA-C. The most significant effects were observed when partners were identical for a complete 16-locus
haplotype, which covered the HLA class I, class II region
and the HLA class III genes (97). However, results obtained
in this population isolate cannot be generalized to outbred
populations in an attempt to understand the pathophysiology
of RPL.
832

Christiansen et al.

The contribution of HLA genes to the etiology of this

pregnancy complication is still under discussion. Immunologically, the fetus is a semi-allograft that has inherited one
set of maternal and one set of foreign paternal HLA
antigens. It was initially suggested that maternal recognition
of paternally derived HLA is required or might at least be
beneficial in normal pregnancy (98). As a result, early studies on HLA in RPL were based on the hypothesis that
increased HLA similarity between partners would lead to
inadequate maternal protective immune responses and to
fetal loss. Although a considerable number of studies of
HLA-sharing in couples with RPL have been performed,
convincing evidence for this hypothesis is still lacking.
Additionally, recent research on the expression of HLA
genes in early embryonic and trophoblast tissues revealed
that, with the exception of HLA-C, classic HLA class I and
II antigens are not expressed on human preimplantation
embryos or at the maternalfetal interface. Thus, maternal
immune recognition of classic HLA antigens expressed by
the fetus is unlikely to be the key mechanism that leads to
maternalfetal tolerance in pregnancy.
Alternatively, the existence of maternal high-risk alleles
for RPL has been proposed under the assumption of allelespecific differences in immune responsiveness or genetic
linkage of high-risk HLA alleles to non-HLA loci, which
predispose to hypersecretion of certain cytokines (e.g., TNF). A meta-analysis of 18 studies on HLA-DR1 and -DR3
frequencies in women with RPL documented a slightly increased risk for RPL in women carrying the HLA-DR1 allele
(OR, 1.29; 95% CI, 1.051.58) (99), and a large case
control study found that HLA-DR3 was associated with RPL
in Danish women (OR, 1.40; 95% CI, 1.11.9) (7). Thus, the
maternal genetic constitution with regard to the HLA-DR1
and -DR3 alleles or closely linked non-HLA genes might be
one of the contributing factors in RPL.
Considerable progress has been made in the elucidation of
the immunologic composition of the fetalmaternal interface. After the nonclassic HLA-G had been identified as the
dominant HLA antigen on trophoblast cells (100), coexpression of HLA-C and HLA-E, albeit at lower concentrations,
was demonstrated more recently (101). Although the exact
functions of these molecules in the placenta have not been
fully elucidated, they seem to participate in an intrinsic
network that guarantees the survival of the fetal allograft.
However, studies on polymorphisms of HLA-C and -E or
sharing of HLA-C or -E alleles in patients with RPL could
not provide evidence for a key role of those antigens in the
pathogenesis of RPL (102106).
Human leukocyte antigen G is the dominant HLA antigen
at the maternalfetal interface, and different HLA-G alleles,
including a functional null-allele, G*0105N, has been identified (107). Clinical data on HLA-G*0105N homozygotes
provided the important information that the major membrane-bound isoform of HLA-G protein, HLA-G1, is not
essential for fetal survival (108, 109). However, the serum

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

level of soluble HLA-G seems to be a predictor of pregnancy

success; decreased HLA-G expression in trophoblasts has
been described in spontaneous abortions (110 112). Furthermore, embryos expressing HLA-G seem to have a better
chance of implanting (113, 114). Those data provide some
evidence that HLA-G or some of its soluble isoforms, although not essential for fetal survival, might still influence
pregnancy development.
Several studies have investigated a potential role for
HLA-G in the pathogenesis of RPL. Recent studies based on
molecular determination of HLA-G alleles and HLA-G nucleotide polymorphisms suggest that women with RPL who
carry specific HLA-G alleles might have lower chances for a
successful pregnancy (106, 115), but the results were not
confirmed in a third study (116), and the pathophysiologic
basis of those allele associations has yet to be clarified.
However, HLA-G polymorphisms are associated with differences in the levels of expression of HLA-G messenger
RNA and protein and in the pattern of alternatively spliced
HLA-G messenger RNA isoforms, which might be involved
in the observed link between detection of HLA-G and pregnancy success, as described above (117, 118).
In summary, although HLA-mediated effects on reproduction were clearly shown in prospective studies in inbred
populations (97), the involvement of the human HLA complex in the pathophysiology of RPL has not yet been clearly
defined. In addition to maternalfetal immunologic interactions, which involve classic and nonclassic HLA loci, effects
of HLA-linked non-HLA genes on fetal development might
play an important role. Future research will have to delineate
which of the many hypotheses of HLA-mediated effects on
pregnancy outcome are valid for this special group of patients.
INDUCTION OF IMMUNOLOGIC TOLERANCE AS AN
APPROACH FOR TREATING RPL
When the immune system encounters an antigen, such as a
pathogen, vaccine, or allogeneic cells, the result is either
sensitization or tolerance (119). During pregnancy, the
mother encounters paternal antigens borne by the semiallogeneic fetus, and the potential arises for a damaging
immune response being mounted against the fetoplacental
unit. On the basis of hypotheses that RPL in humans could
be caused by an immunologic attack, treatments have been
developed aimed at resetting the immune system so that
the mother becomes tolerant of her fetus. The first substantial
trial with an immunologic approach was published by Mowbray and colleagues in 1985 (120). In that trial, immunization with paternal peripheral blood mononuclear cells
(PBMC) given intravenously, intradermally, and subcutaneously, led to successful pregnancy in 71% of patients (n
35) given paternal cells, compared with 47% of patients (n
30) given their own (i.e., maternal) cells. The successful
outcomes reported by this group led to considerable interest
in the idea that RPL can be treated by prior immunization of
Fertility and Sterility

the mother with paternal PBMC. Immunologic changes that

have been observed after such procedures include a decrease
in the mixed lymphocyte reaction, decreased soluble IL2
receptor levels, decreased NK cell activity, an increase in
blocking factors, and a decreased Th1/Th2 ratio (121
124).
A number of clinical trials have now been carried out,
with some studies showing a beneficial effect and others no
beneficial effect; in some instances such procedures seem to
have worsened the prognosis. Unfortunately, heterogeneity
in cell numbers, number of injections, type of placebo used,
and patients reproductive histories make comparisons
among the trials somewhat difficult. A recent review for the
Cochrane library of 11 trials of paternal PBMC immunization found that, overall, paternal PBMC treatment led to a
live birth rate of 62.2%, compared with 60.0% in controls
(21). This analysis concluded that paternal cell immunization, along with other immunologic treatments, such as thirdparty donor leukocytes, trophoblast membranes, and intravenous Ig, provide no significant beneficial effect over
placebo in preventing further miscarriages. However, the
results of another meta-analysis suggest that such treatment
has value in a subgroup of women with primary RPL who
have suffered a large number of miscarriages (125).
It has been argued that the current immunologic therapies
for the treatment of RPL do not have a sound scientific basis.
Are there other immunologically based options that could be
explored? Despite much research, the key mechanisms allowing a mother to tolerate her semi-allogeneic fetus remain to be established (126, 127). Recently, evidence for a
role of CD4 CD25 T regulatory cells has been put
forward (128). With respect to immune responses in general,
although some of the criteria that determine whether exposure to an antigen leads to sensitization or to immunologic
tolerance have been defined, the details again remain
sketchy. Polymeric antigen, given at a moderate dose either
SC or IM will usually lead to sensitization in the presence of
costimulatory signals, such as those provided by adjuvants.
On the other hand, monomeric antigens or antigens given at
very low or very high doses, particularly if given IV, mucosally, or intrathymically will, in the absence of additional
costimulatory signals, preferentially lead to tolerance induction. Indeed, an encounter with antigen in the absence of
costimulatory signals, such as cytokines and the cell-surface
B7 molecules, leads to either apoptosis or anergy (functional
inactivation) of the relevant lymphocytes. The encounter of
T lymphocytes with allogeneic HLA in the absence of costimulation has also been shown to result in tolerance (129).
Several strategies aimed at preventing rejection of allografts are being developed by transplantation biologists
(130). Here there are some parallels but also some fundamental differences from the situation regarding a semi-allogeneic fetus. Nonetheless, if the immune system has the
potential to destroy the fetus, any way of preventing immunologic destruction of a fully allogeneic tissue or organ graft
833

might provide some signposts for developing immunologic

treatments of RPL. Current experimental approaches that are
being investigated in transplantation biology involve exposing the recipient immune system to donor antigen in the
presence of T cell modulators, such as monoclonal antibodies, directed toward important T cell surface molecules,
including CD3, CD4, or CD154 (CD40L). Another approach
is to create mixed allogeneic chimeras by infusion of allogeneic bone marrow in combination with T cell modulators.
On the basis of such strategies, what potential therapeutic
options might open up for the treatment of RPL? Paternal
cells could be administered in the presence of immunomodulators, such as IL10, or soluble versions of cell-surface
molecules, such as CD178 (Fas-ligand). Paternal bone marrow plus T cellmodulating agents could be used to establish
chimerism. However, such therapies might carry a substantial risk to the recipient, and these types of approaches are
unlikely to be investigated in RPL until much more information is available from transplantation studies.
In conclusion, although a substantial review of the trials of
paternal lymphocyte immunization has not shown an overall
benefit (21), there is some evidence that allogenic leukocyte
immunotherapy with freshly prepared cells is beneficial for a
subset of women with primary RPL with no evidence of
antipaternal cytotoxic immunity and in whom there is no
autoimmune disease (125, 131). The current incomplete understanding of the immune mechanisms involved in the
maintenance of pregnancy makes it hard to devise rational
novel immunotherapies. Potential future directions can be
gleaned from the experience of transplantation biologists and
the experimental therapies being developed in that field. The
problem might not, in fact, be so great in the case of RPL.
After all, an allogenic kidney graft should be for life,
whereas a pregnancy is for only 9 months. Thus, for RPL, it
is not essential to permanently switch off any allo-reactive
response but rather to curtail the response for a relatively
short period.

LIFESTYLE FACTORS AND RPL

Lifestyle factors rarely, if ever, cause RPL; however, epidemiologic studies have given evidence that a series of lifestyle
factors can increase the risk of miscarriage. There is good
evidence that obesity (132, 133), daily caffeine intake 300
mg (134 137), alcohol consumption (136, 138), and use of
nonsteroidal anti-inflammatory drugs (139, 140) increase the
risk of miscarriage significantly. It is more controversial
whether smoking increases the risk of miscarriage (136, 138,
141, 142). Social class and occupation also impact the rate of
miscarriage, with the greatest risk among women exposed to
high physical or psychic stress during work (143145). Several studies now also indicate that a previous infertility
diagnosis or infertility treatment might increase the risk of
miscarriage (146, 147). It is important to be aware of these
risk factors because, in an optimal case control or random834

Christiansen et al.

ized controlled trial, an equal distribution of the factors

between the groups should be documented.
DISCUSSION
Many of these controversies in RPL research are due to the
apparent lack of appreciation of the many methodologic
pitfalls highlighted in the first part of this article. It is our
hope that the design of future RPL studies will improve so as
not to fall into the methodologic pitfalls inherent in this area
of research. When more properly designed and conducted
studies are published, we believe that consensus about the
investigations and treatments needed for RPL patients can be
reached.
There is no doubt that research in RPL during recent years
has identified a series of new factors that exhibit an association with RPL in case control studies and that also seem to
affect the pregnancy prognosis and thus might be characterized as causative factors (Table 2). So many risk factors have
now been identified that it is very common to find more than
one of them in the same patient. Both thrombophilic risk
factors (81) and immunogenetic risk factors, such as specific
HLA alleles and genes for low mannan-binding lectin serum
levels (Christiansen et al., unpublished data), seem to aggregate significantly more frequently than expected in RPL
patients, indicating that they might exhibit an additive or
multiplicative effect on RPL risk. So far, it has been common to divide the causes of RPL into single sufficient factors
like slices of a pie uterine malformations 10%, endocrine
factors 10%, antiphospholipid antibodies 15%, and so on
which together with the unexplained group make up 100%
(Fig. 1). This model is probably not adequate, owing to the
frequent finding of several risk factors in the same patients
and the general belief that disorders with a high population
prevalence (such as RPL) almost always have a multifactorial background. In internal medicine and other disciplines,
the development of many common diseases (e.g., arterial
hypertension, diabetes mellitus, schizophrenia) are thought
to be determined by a threshold model. A number of risk
factors are in themselves not sufficient to cause disease, but
when several intrinsic and extrinsic factors come together in
the same individual, the risk exceeds a threshold level, and
disease develops. Previous family studies (41, 148) found
that the RPL prevalence in siblings of RPL probands indeed
was in accordance with a multifactorial threshold model for
inheritance (43). We therefore encourage scientists and clinicians working in the area of RPL to think within the
threshold rather than the pie model.
The clinical implication is that, in principle, an RPL
patient should be screened for all potential risk factors, and
the investigation should not for economic or other reasons
stop as soon as the first risk factor has been identified. A
problem is that there is no consensus as to the importance of
all risk factors, and there is very little agreement regarding
the level of positivity of the tests. The threshold of detecting
abnormality varies greatly from study to study. This holds

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

true for both laboratory tests, such as anticardiolipin antibodies (149), and uterine abnormalities. Indeed, only studies
of the impact of various potential risk factors on the outcome
of future, untreated pregnancies in RPL patients can establish what degree of abnormality of a factor should be considered as important for RPL.
It is recommended that examinations in RPL should as a
minimum be carried out according to Table 3 (documentation A) and, in addition, factors with association and prognostic impact documented in studies of adequate size and
quality could be investigated (Table 3, documentation B). A
specific prognosis for the patient can only be given if all risk
factors are investigated. When clinicians encounter a patient
with RPL and several risk factors, it is important to inform
her about all the risk factors and then discuss with her
whether an attempt should be made to treat all risk factors at
the same time or only one of them before or during the next
pregnancy attempt. In many cases, the patient will wish to
have all factors corrected or treated before attempting a
subsequent pregnancy.
As mentioned previously, a number of lifestyle factors,
including obesity, occupation, history of alcohol use, and
caffeine consumption, are important for the risk of miscarriage. Recurrent pregnancy loss is a complex disorder for
which lifestyle factors are expected to modify the effect of
the non-lifestyle (intrinsic) factors previously discussed. The
prevalence of the most important lifestyle factors should be
reported in publications to establish whether groups studied
for the occurrence of non-lifestyle risk factors or pregnancy
outcome are comparable. Some RPL clinics are already
using a structured information sheet for each patient, which
includes a series of lifestyle factors. It is recommended that
investigators doing research in RPL should use a standardized sheet for recovering lifestyle information from patients.
It might be a task for international societies in the field of
fertility and reproduction to design such a standardized

sheet, which can be referred to in publications. However, it

is important to bear in mind that lifestyle factors such as
smoking and alcohol use are very difficult to evaluate from
a sheet. Patients might not provide the correct information
about these matters in a fertility clinic. In a recent study, no
correlation between the patients initial history regarding
tobacco use and subsequent IVF outcome was found; however, when cotinine levels (a serum marker for tobacco use)
were measured in the blood from Day 3 samples, an excellent correlation between high cotinine concentration and bad
embryo quality was found in the same patients (150).
A major problem in RPL research is the scarcity of highquality randomized controlled trials of treatments. Some
useful trials testing hormonal, anticoagulation, and immunologic therapies have been conducted. However, except for
cervical cerclage, almost none of the surgical interventionsremoval of uterine fibromas and septae, ovarian diathermia, and abdominal cerclage have been evaluated in
randomized trials. Unfortunately, surgeons are not disposed
to allocate patients to a non-treatment group. Many surgeons
are convinced that there is a narrow window of opportunity
to do a randomized study of surgical interventions beyond
a certain stage it would be too late to start because the
treatment becomes too established and the patients expect it
to be offered.
Fortunately, there are examples of large randomized trials
of surgery to improve fertility. In Scandinavia, a multicenter
trial was organized to evaluate the effect on implantation rate
and miscarriage rate of removal of hydrosalpinges in IVF
patients (151). Two hundred twenty patients were randomized to surgical treatment or no treatment. A similar trial of
removal of fibroids in women with RPL could be done.
However, to facilitate surgical trials we need a change in the
whole concept of being a good scientist. By participating in
a multicenter randomized trial, an investigator will be one of
perhaps 100 collaborators, but only a minority of these will

TABLE 3
Recommended investigations of recurrent pregnancy loss patients.
Investigation
Hysterosalpingography or hysteroscopy or sonohysterography
Karyotyping of the couple
Thyroid hormones
Androgens, LH, FSH in patients with irregular menstruations
APTT-/dRVVT/lupus anticoagulant
IgG and IgM anticardiolipin antibodies
Factor V Leiden mutation
Mannan-binding lectin
Maternal HLA-G and HLA-DR types

Documentation
A
A
A
B
A
A
A
B
B

Note: A value documented in many studies; B value suggested in few but large studies; APTT activated partial
thrombin time; dRVVT dilute Russels viper venom time.
Christiansen. Management of recurrent pregnancy loss. Fertil Steril 2005.

Fertility and Sterility

835

be accepted as authors of the resulting article. Much more

benefit should be given to the scientists participating in such
trials.
We must realize that the real placebo in randomized trials
of surgical intervention is the sham operation. This will
probably not be approved by any ethics committee today,
and therefore no real placebo-controlled trials can be done to
assess surgery, only trials comparing surgery with no surgery
or comparing two different surgical techniques.
So far, the discussion has dealt with the scientists opportunities and responsibilities to do better research. However,
the editors of the medical journals also carry some responsibility for the slow progress in our understanding and management of RPL. Higher rejection rates of reports from small
and improperly designed studies would help in the development of larger and better studies.
Meta-analyses have become increasingly important in developing a more evidence-based management of RPL, but as
in all types of research, faulty conclusions might be reached
in a meta-analysis (152). As previously discussed, the Cochrane meta-analysis on immunotherapy in RPL lumps together a series of trials that are very heterogenous with
regard to the patients and the treatments. In an optimal
meta-analysis, data from the original sources should be extracted by two or three reviewers and checked for bias and
afterwards sent to analyzers to be checked for clinical and
statistical heterogeneity. In the Cochrane database, source
data are very often not obtained. If a Cochrane meta-analysis
makes conclusions from too-small and heterogenous studies,
this can hinder rather than inspire the performance of further
randomized controlled trials in the area. If the Cochrane
database suggests that studies so far have not pointed toward
a treatment effect, it makes it very difficult to continue
investigation of this treatment option. This is a relevant
problem for initiating new and better treatment trials in the
area of RPL.
In conclusion, a series of hurdles must be overcome before
our understanding and treatment of RPL becomes optimal.
The major hurdles relate to defining RPL, recognizing the
multifactorial/polygenic nature of RPL, overcoming the difficulty in conducting large placebo-controlled trials, and
changing the practice of combining data from heterogenous
studies for meta-analysis.
REFERENCES
1. Daya S. Habitual abortion. In: Copeland LJ, Jarrell JF, eds. Textbook
of gynecology, 2nd ed. Philadelphia: W.B. Saunders, 2002:22771.
2. Mills JL, Simpson JL, Driscoll SG, Jovanic-Peterson L, Van Allen M,
Aarons JH, et al. Incidence of spontaneous abortion among normal
women and insulin-dependent women whose pregnancies were identified within 21 days of conception. N Engl J Med 1988;319:161723.
3. Plouffe L, White EW, Tho SP, Sweet CS, Layman LC, Whitman GF,
et al. Etiological factors of recurrent abortion and subsequent reproductive performance of couples: have we made any progress in the
past 10 years? Am J Obstet Gynecol 1992;167:31321.

836

Christiansen et al.

4. Axelsson G, Rylander R. Validation of questionnaire reported miscarriage, malformation and birth weight. Int J Epidemiol 1984;13:94 8.
5. Christiansen OB. A fresh look at the causes and treatment of recurrent
miscarriage, especially its immunological aspects. Hum Reprod Update 1996;2:27193.
6. Kruse C, Rosgaard A, Steffensen R, Varming K, Jensenius JC, Christiansen OB. Low serum level of mannan-binding lectin is a determinant for pregnancy outcome in women with recurrent spontaneous
abortion. Am J Obstet Gynecol 2002;187:131320.
7. Kruse C, Steffensen R, Varming K, Christiansen OB. A study of
HLA-DR and -DQ alleles in 588 patients and 562 controls confirms
that HLA-DRB*03 is associated with recurrent miscarriage. Hum
Reprod 2004;19:121521.
8. Carp H, Toder V, Torchinsky A, Portuguese S, Lipitz S, Gazit E, et al.
Allogenic leukocyte immunization after five or more miscarriages.
Recurrent Miscarriage Immunotherapy Trialists Group. Hum Reprod
1997;12:250 5.
9. Ogasawara M, Aoki K, Okada S, Suzumori K. Embryonic karyotype
of abortuses in relation to the number of previous miscarriages. Fertil
Steril 2000;73:300 4.
10. Out HJ, Bruinse HW, Christiaens GCML, van Vliet M, Meilof JF, de
Groot PG, et al. Prevalence of antiphospholipid antibodies in patients
with fetal loss. Ann Rheum Dis 1991;50:53357.
11. Birkeland SA, Kristoffersen K. The fetus as an allograft: a longitudinal study of normal human pregnancies studied with mixed lymphocyte cultures between mother-father and mother-child. Scand J Immunol 1980;11:3119.
12. Coulam CB. Immunological tests in the evaluation of reproductive
disorders: a critical review. Am J Obstet Gynecol 1992;167:1844 51.
13. Alius KT. A follow-up study of immunoglobulin levels and autoantibodies in an unselected pregnant population. Am J Reprod Immunol
1994;31:189 96.
14. Makhseed M, Raghupathy R, Azizieh F, Omu A, Al-Shamali E,
Ashkanani L. Th1 and Th2 cytokine profiles in recurrent aborters with
successful pregnancy and with subsequent abortions. Hum Reprod
2001;16:2219 26.
15. Piccinni MP, Giudizi MG, Biagiotti R, Beloni L, Giannarini L, Sampognaro S, et al. Progesterone favors the development of human T
helper cells producing Th2-type cytokines and promotes both IL-4
production and membrane CD30 expression in established Th1 cell
clones. J Immunol 1995;155:128 33.
16. Szekeres-Bartho J, Faust Z, Varga P, Szereday L, Kelemen K. The
immunological pregnancy protective effect of progesterone is manifested via controlling cytokine production. Am J Reprod Immunol
1996;35:348 51.
17. Michel M, Clark DA, Mowbray JF, Beard RW. Histologic and immunologic study of uterine biopsy tissue of women with incipient
abortion. Am J Obstet Gynecol 1989;161:409 14.
18. Jablonowska B, Selbing A, Palfi M, Ernerudh J, Kjellberg S, Lindton
B. Prevention of recurrent spontaneous abortion by intravenous immunoglobulin: a double-blind placebo-controlled study. Hum Reprod
1999;14:838 41.
19. Christiansen OB, Pedersen B, Rosgaard A, Husth M. A randomized,
double-blind, placebo-controlled trial of intravenous immunoglobulin
in the prevention of recurrent miscarriage: evidence for a therapeutic
effect in women with secondary recurrent miscarriage. Hum Reprod
2002;17:809 16.
20. World Health Organization. Recommended definitions, terminology
and format for statistical tables related to the perinatal period. Acta
Obstet Gynaecol Scand 1977;56:24753.
21. Scott JR. Immunotherapy for recurrent miscarriage [Cochrane Review]. In: The Cochrane Library, 3. Oxford: Update Software, 2003.
22. Gibb DMF, Salaria DA. Transabdominal cervicoistmic cerclage in the
management of recurrent second trimester miscarriage and preterm
delivery. Br J Obstet Gynaecol 1995;102:802 6.
23. Christiansen OB. Transabdominal cervicoistmic cerclage in the management of recurrent second trimester miscarriage and preterm delivery [comment]. Br J Obstet Gynaecol 1996;103:595 6.

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

24. Daya S. Immunotherapy for unexplained recurrent spontaneous abortion. Infertil Reprod Med Clin North Am 1997;8:6577.
25. Daya S, Gunby J, Porter F, Scott J, Clark DA. Critical analysis of
intravenous immunoglobulin therapy for recurrent miscarriage. Hum
Reprod Update 1999;5:475 82.
26. Strobino BR, Kline J, Shrout P, et al. Recurrent spontaneous abortion:
definition of a syndrome. In: Porter IH, Hook EB, eds. Human embryonic and fetal death. New York: Academic Press, 1980:31529.
27. Fitzsimmons J, Jackson D, Wapner R, Jackson L. Subsequent reproductive outcome in couples with repeated pregnancy loss. Am J Med
Genet 1983;16:5837.
28. Ober C, Karrison T, Odem RR, Barnes RB, Branch DW, Stephenson
MD, et al. Mononuclear-cell immunisation in prevention of recurrent
miscarriages: a randomised trial. Lancet 1999;354:3659.
29. De Braekeleer M, Dao TN. Cytogenetic studies in couples experiencing repeated pregnancy losses. Hum Reprod 1990;5:519 28.
30. Bou J, Bou A, Lazar P. Retrospective and prospective epidemiological studies of 1500 karyotyped spontaneous human abortions.
Teratology 1975;12:1126.
31. Hassold TJ, Matsuyama A, Newlands IM, Matsuura JS, Jacobs PA. A
cytogenetic study of abortions in Hawaii. Ann Hum Genet 1978;41:
44354.
32. Plachot M. Chromosome analysis of spontaneous abortions after IVF.
A European survey. Hum Reprod 1989;4:4259.
33. Eiben B, Bartels I, Bahr-Porch S, Borgmanns S, Gatz G, Gellert G, et
al. Cytogenetic analysis of 750 spontaneous abortions with the direct
preparation method of chorionic villi and its implications for studying
genetic causes of pregnancy wastage. Am J Hum Genet 1990;47:656
63.
34. Stephenson MD, Awartani KA, Robinson WP. Cytogenetic analysis of
miscarriages from couples with recurrent miscarriage: a case-control
study. Hum Reprod 2002;17:446 51.
35. Hassold TJ, Chen N, Funkhouser T, Jooss T, Manuel B, Matsuura J,
et al. A cytogenetic study of 1000 spontaneous abortions. Ann Hum
Genet 1980;44:15178.
36. Strom CM, Ginsberg N, Applebaum M, Bogorzi N, White M, Caffarelli M, et al. Analysis of 95 first-trimester spontaneous abortions by
chorionic villus sampling and karyotype. J Assist Reprod Genet 1992;
9:458 61.
37. Rubio C, Simon C, Vidal F, Rodrigo L, Pehlivan T, Remohi J, et al.
Chromosomal abnormalities and embryo development in recurrent
miscarriage couples. Hum Reprod 2003;18:182 8.
38. Vidal F, Gimnez C, Rubio C, Simn C, Pellicer A, Santal J, et al.
FISH preimplantation diagnosis of chromosome aneuploidy in recurrent pregnancy wastage. J Assist Reprod Genet 1998;15:309 12.
39. Simn C, Rubio C, Vidal F, Gimnez C, Moreno C, Parrilla JJ, et al.
Increased chromosome abnormalities in preimplantation embryos after in-vitro fertilization in patients with recurrent miscarriage. Reprod
Fertil Dev 1998;1:8792.
40. Alexander SA, Latinne D, Debruyere M, Dupont E, Gottlieb W,
Thomas K. Belgian experience with repeat immunization in recurrent
spontaneous abortion. In: Beard RW, Sharp F, eds. Early pregnancy
loss: mechanisms and treatment. London: Springer Verlag, 1988:355
63.
41. Christiansen OB, Mathiesen O, Lauritsen JG, Grunnet N. Idiopathic
recurrent spontaneous abortion. Evidence of a familial predisposition.
Acta Obstet Gynecol Scand 1990;69:597 601.
42. Ho H, Gill TJ, Hsieh C, Yang Y, Lee T. The prevalence of recurrent
spontaneous abortion, cancer, and congenital anomalies in the families
of couples with recurrent spontaneous abortions or gestational trophoblastic tumors. Am J Obstet Gynecol 1991;165:461 6.
43. Emery AEH. Methodology in medical genetics, 2nd rev ed. New
York: Churchill Livingstone, 1986.
44. Christiansen OB, Andersen HH, Hojbjerre M, Kruse TA, Lauritzen
SL, Grunnet N. Maternal HLA Class II allogenotypes are markers for
the predisposition to fetal losses in families of women with unexplained recurrent fetal loss. Eur J Immunogenet 1995;22:32334.

Fertility and Sterility

45. Homer HA, Li TC, Cooke ID. The septate uterus: a review of management and reproductive outcome. Fertil Steril 2000;73:114.
46. Seoud MD, Patterson R, Mausher SJ, Coddington CC. Effects of
myomas or prior myomectomy on the in vitro fertilisation (IVF)
performance. J Assist Reprod Genet 1992;9:21722.
47. Farhi J, Ashkenazi J, Feldberg D, Dicker D, Orvieto R, Ben Rafael Z.
Effect of uterine leiomyomata on the results of in-vitro fertilization
treatment. Hum Reprod 1995;10:2576 8.
48. Eldar-Geva T, Meagher S, Healy DL, MacLachlan V, Breheny S,
Wood C. Effect of intramural, subserosal and submucosal uterine
fibroids on the outcome of assisted reproductive technology treatment.
Fertil Steril 1998;70:68791.
49. Ramzy AM, Satar M, Amin Y, Mansour RT, Serour GI, Aboulghar
MA. Uterine myomata and outcome of assisted reproduction. Hum
Reprod 1998;13:198 202.
50. Stovall DW, Parrish SB, Van Voorhis BJ, Hahn BJ, Sparks AE, Syrop
CH. Uterine leiomyomas reduce the efficacy of assisted reproduction
cycles: results of a matched follow-up study. Hum Reprod 1998;13:
1927.
51. Hart R, Khalaf Y, Yeong CT, Seed P, Taylor A, Braude P. A
prospective controlled study of the effect of intramural uterine fibroids
on the outcome of assisted conception. Hum Reprod 2001;16:24117.
52. Bajekal N, Li TC. Fibroids, infertility and pregnancy wastage. Hum
Reprod Update 2000;6:614 20.
53. Clifford K, Rai R, Watson H, Regan L. An informative protocol for
the investigation of recurrent miscarriage: preliminary experience of
500 consecutive cases. Hum Reprod 1994;9:1328 32.
54. Drakeley AJ, Quenby S, Farquharson RG. Mid-trimester loss: appraisal of a screening protocol. Hum Reprod 1998;13:14719.
55. Craig LB, Ke RW, Kutteh WH. Increased prevalence of insulin
resistance in women with a history of recurrent pregnancy loss. Fertil
Steril 2002;78:48790.
56. Regan L, Owen EJ, Jacobs HS. Hypersecretion of luteinising hormone, infertility, and miscarriage. Lancet 1990;336:1141 4.
57. Nardo LG, Rai R, Backos M, El-Gaddal S, Regan L. High serum
luteinizing hormone and testosterone concentrations do not predict
pregnancy outcome in women with recurrent miscarriage. Fertil Steril
2002;77:348 52.
58. Clifford K, Rai R, Watson H, Franks S, Regan L. Does suppressing
luteinising hormone secretion reduce the miscarriage rate? Results of
randomized controlled trial. BMJ 1996;312:1508 11.
59. Li TC, Spuijbroek MDEH, Tuckerman E, Anstie B, Loxley M, Laird
SM. Endocrinological and endometrial factors in recurrent miscarriage. Br J Obstet Gynaecol 2000;107:1975 80.
60. Westergaard LG, Laursen SB, Andersen CY. Increased risk of early
pregnancy loss by profound suppression of luteinizing hormone during ovarian stimulation in normogonadotrophic women undergoing
assisted reproduction. Hum Reprod 2000;15:1003 8.
61. Rai R, Backos M, Rushworth F, Regan L. Polycystic ovaries and
recurrent miscarriage: a reappraisal. Hum Reprod 2000;15:6125.
62. Amer SAKA, Copolan V, Li TC, Cooke ID. Long term follow-up of
patients with polycystic ovarian syndrome after laparoscopic ovarian
drilling: clinical outcome. Hum Reprod 2002;17:2035 42.
63. Stadtmauer LA, Wong BC, Oehninger S. Should patients with polycystic ovary syndrome be treated with metformin? Benefits of insulin
sensitizing drugs in polycystic ovary syndrome beyond ovulation
induction. Hum Reprod 2002;17:3016 26.
64. Costello MF, Eden JA. A systematic review of the reproductive
system effects of metformin in patients with polycystic ovary syndrome. Fertil Steril 2003;79:113.
65. Okon MA, Laird SM, Tuckerman E, Li TC. Serum androgen levels in
women who suffer recurrent miscarriage and their correlation with
markers of endometrial function. Fertil Steril 1998;69:68290.
66. Bussen S, Sutterlin M, Steck T. Endocrine abnormalities during the
follicular phase in women with recurrent spontaneous abortions. Hum
Reprod 1999;14:18 20.
67. Noyes RW, Hertig AT, Rock J. Dating the endometrial biopsy. Fertil
Steril 1950;1:325.

837

68. Li TC, Tuckerman E, Laird SM. Endometrial factors in recurrent

miscarriage. Hum Reprod Update 2002;8:4352.
69. Kutteh WH. Antiphospholipid antibody-associated recurrent pregnancy loss: treatment with heparin and low-dose aspirin is superior to
low-dose aspirin alone. Am J Obstet Gynecol 1996;174:1584 9.
70. Rai R, Cohen H, Dave M, Regan L. Randomised controlled trial of
aspirin and aspirin plus heparin in pregnant women with recurrent
miscarriage associated with phospholipid antibodies (or antiphospholipid antibodies). BMJ 1997;314:2537.
71. Empson M, Lassere M, Craig JC, Scott JR. Recurrent pregnancy loss with
antiphospholipid antibody: a systematic review of therapeutic trials. Obstet Gynecol 2002;99:1335 44.
72. Wilson WA, Gharavi AE, Koike T, Lockshin MD, Branch DW, Piette
JC, et al. International consensus statement on preliminary classification criteria for definite antiphospholipid syndrome: report of an
international workshop. Arthritis Rheum 1999;42:309 11.
73. Kutteh WH, Franklin RD. Assessing the variation in antiphospholipid
antibody (APA) assays: comparison of results from 10 centers. Am J
Obstet Gynecol. 2004;191:440 8.
74. Franklin RD, Kutteh WH. Antiphospholipid antibodies (APA) and
recurrent pregnancy loss: treating a unique APA positive population.
Hum Reprod 2002;17:29815.
75. Franklin RD, Kutteh WH. Effects of unfractionated and low molecular
weight heparin on antiphospholipid antibody binding in vitro. Obstet
Gynecol 2003;101:455 62.
76. Kutteh WH, Yetman DL, Chantilis SJ, Crain J. Effect of antiphospholipid antibodies in women undergoing in-vitro fertilization: role of
heparin and aspirin. Hum Reprod 1997;12:11715.
77. Chilcott IT, Margara R, Cohen H, Rai R, Skull J, Pickering W, et al.
Pregnancy outcome is not affected by antiphospholipid antibody status
in women referred for in vitro fertilization. Fertil Steril 2000;73:526
30.
78. Stern C, Chamley L, Norris H, Hale L, Baker HW. A randomized
double-blind, placebo-controlled trial of heparin and aspirin for
women with in vitro fertilization implantation failure and antiphospholipid or antinuclear antibodies. Fertil Steril 2003;80:376 83.
79. Laskin CA, Bombardier C, Hannah ME, Mandel FP, Ritchie JW,
Farewell V, et al. Prednisone and aspirin in women with autoantibodies and unexplained recurrent pregnancy loss. N Engl J Med 1997;
337:148 53.
80. Rushworth FH, Backos M, Rai R, Chilcott IT, Baxter N, Regan L.
Prospective pregnancy outcome in untreated recurrent miscarriers
with thyroid autoantibodies. Hum Reprod 2000;15:16379.
81. Sarig G, Younis JS, Hoffman R, Lanir N, Blumenfeld Z, Brenner B.
Thrombophilia is common in women with idiopathic pregnancy loss
and is associated with late pregnancy wastage. Fertil Steril 2002;77:
3427.
82. Rey E, Kahn SR, David M, Shrier I. Thrombophilic disorders and fetal
loss: a meta-analysis. Lancet 2003;361:901 8.
83. Kovalsky G, Garcia CR, Berlin JA, Sammel MD, Barnhart KT.
Evaluation of the association between hereditary thrombophilias and
recurrent pregnancy loss: a meta-analysis. Arch Intern Med 2004;164:
558 63.
84. Bulmer JN, Morrison L, Longfellow M, Ritson A, Pace D. Granulated
lymphocytes in human endometrium: histochemical and immunohistochemical studies. Hum Reprod 1991;6:791 8.
85. Lachapelle MH, Miron P, Hemmings R, Roy DC. Endometrial T, B,
and NK cells in patients with recurrent spontaneous abortion. Altered
profile and pregnancy outcome. J Immunol 1996;156:402734.
86. Yamada H, Kato EH, Kobashi G, Ebina Y, Shimada S, Morikawa M,
et al. High NK cell activity in early pregnancy correlates with subsequent abortion with normal chromosomes in women with recurrent
abortion. Am J Reprod Immunol 2001;46:132 6.
87. Guimond MJ, Wang B, Croy BA. Engraftment of bone marrow from
severe combined immunodeficient (SCID) mice reverses the reproductive deficits in natural killer cell-deficient tg epsilon 26 mice. J Exp
Med 1998;187:21723.
88. Quenby S, Bates M, Doig T, Brewster J, Lewis-Jones DI, Johnson

838

Christiansen et al.

89.
90.

91.

92.

93.

94.

95.

96.

97.

98.
99.

100.

101.

102.

103.

104.

105.

106.

107.

108.

109.

PM, et al. Pre-implantation endometrial leukocytes in women with

recurrent miscarriage. Hum Reprod 1999;14:2386 91.
Barakonyi A, Polgar B, Szekeres-Bartho J. The role of / TCR
positive cells in pregnancy II. Am J Reprod Immunol 1999;42:837.
Wegmann TG, Lin H, Guilbert L, Mosmann TR. Bidirectional cytokine interactions in the maternal-fetal relationship: is successful pregnancy a TH2 phenomenon? Immunol Today 1993;14:353 6.
Chaouat G, Assal Meliani A, Martal J, Raghupathy R, Elliot J,
Mosmann T, et al. IL-10 prevents naturally occurring fetal loss in the
CBA x DBA/2 mating combination, and local defect in IL-10 production in this abortion-prone combination is corrected by in vivo injection of IFN-tau. J Immunol 1995;154:4261 8.
Hill JA. T-helper 1-type immunity to trophoblast: evidence for a new
immunological mechanism for recurrent abortion in women. Hum
Reprod 1995;10(Suppl 2):114 20.
Piccinni MP, Beloni L, Livi C, Maggi E, Scarselli G, Romagnani S.
Defective production of both leukemia factor and type 2 T-helper
cytokines by decidual T cells in unexplained recurrent abortions. Nat
Med 1998;4:1020 4.
Raghupathy R, Makhseed M, Azizieh F, Hassan N, Al-Azemi M,
Al-Shamali E. Maternal Th1- and Th2-type reactivity to placental
antigens in normal human pregnancy and unexplained recurrent spontaneous abortions. Cell Immunol 1999;196:12230.
Lim KJH, Odukoya OA, Ajjan RA, Li TC, Wheetman AP, Cooke ID.
The role of T-helper cytokines in human reproduction. Fertil Steril
2000;73:136 42.
Von Wolff M, Thaler CJ, Strowitzki T, Broome J, Stolz W, Tabibzadeh S. Regulated expression of cytokines in human endometrium
throughout the menstrual cycle: dysregulation in habitual abortion.
Mol Hum Reprod 2000;6:62734.
Ober C, Hyskop T, Elias S, Weitkamp LR, Hauck WW. Human
leukocyte antigen matching and fetal loss: results of a 10 year prospective study. Hum Reprod 1998;13:33 8.
Beer AE, Billingham RE. The embryo as a transplant. Sci Am 1974;
230:36 46.
Christiansen OB, Ring M, Rosgaard A, Grunnet N, Gluud C. Association between HLA-DR1 and -DR3 antigens and unexplained repeated miscarriage. Hum Reprod Update 1999;5:249 55.
Ellis SA, Palmer MS, McMichael AJ. Human trophoblast and the
choriocarcinoma cell line BeWo express a truncated HLA-class I
molecule. J Immunol 1990;144:7315.
King A, Boocock C, Sharkey AM, Gardner L, Beretta A, Siccardi AG,
et al. Evidence for the expression of HLA-C class I mRNA and protein
by first trimester trophoblast. J Immunol 1996;156:2068 76.
Christiansen OB, Mohapeloa HP, Steffensen R, Jersild C. HLA-C and
Bw typing of couples with unexplained recurrent miscarriages. J
Reprod Immunol 1997;37:6377.
Steffensen R, Christiansen OB, Bennett EP, Jersild C. HLA-E polymorphisms in patients with recurrent spontaneous abortion. Tissue
Antigens 1998;52:569 72.
Kanai T, Fujii T, Keicho N, Tokunaga K, Yamashita T, Hyodo H, et
al. Polymorphism of human leucocyte antigen-E gene in the Japanese
population with and without recurrent abortion. Am J Reprod Immunol 2001;45:168 73.
Imai T, Takakuwa K, Ishii K, Adachi H, Higashino M, Kurata H, et al.
HLA-class I antigens in patients with unexplained recurrent abortion.
J Perinat Med 2001;29:42732.
Pfeiffer KA, Fimmers R, Engels G, van der Ven H, van der Ven K.
The HLA-G genotype is potentially associated with idiopathic recurrent spontaneous abortion. Mol Hum Reprod 2001;7:373 8.
Suarez MB, Morales P, Castro MJ, Fernandez V, Varela P, Alvarez M,
et al. A new HLA-G allele (HLA-G*0105N) and its distribution in the
Spanish population. Immunogenetics 1997;45:464 5.
Ober C, Aldrich C, Rosinsky B, Robertson A, Walker MA, Willadsen
S, et al. HLA-G1 protein expression is not essential for fetal survival.
Placenta 1998;19:12732.
Castro MJ, Morales P, Rojo-Amigo R, Martinez-Laso J, Allende I,
Varela P, et al. Homozygous HLA-G*0105N healthy individuals

Management of recurrent pregnancy loss

Vol. 83, No. 4, April 2005

110.

111.

112.

113.

114.

115.

116.

117.

118.

119.
120.

121.

122.

123.

124.

125.

126.

127.

128.
129.

indicate that membrane-anchored HLA-G1 molecule is not necessary

for survival. Tissue Antigens 2000;56:2329.
Athanassakis I, Paflis M, Ranella A, Vassiliadis S. Detection of
soluble HLA-G levels in maternal serum can be predictive for a
successful pregnancy. Transplant Proc 1999;31:1834 7.
Pfeiffer KA, Rebmann V, Passler M, van der Ven K, van der Ven H,
Krebs D, et al. Soluble histocompatibility antigen levels in early
pregnancy after in vitro fertilization. Hum Immunol 2000;61:559 64.
Emmer PM, Steegers EA, Kerstens HM, Bulten J, Nelen WL, Boer K,
et al. Altered phenotype of HLA-G expressing trophoblast and decidual natural killer cells in pathological pregnancies. Hum Reprod
2002;17:1072 80.
Jurisicova A, Casper RF, MacLusky NJ, Librach CL. Embryonic
human leucocyte antigen-G expression: possible implications for human preimplantation development. Fertil Steril 1996;65:9971002.
Fuzzi B, Rizzo R, Criscuoli L, Noci I, Melchiorri L, Scarselli B, et al.
HLA-G expression in early embryos is a fundamental prerequisite for
the obtainment of pregnancy. Eur J Immunol 2002;32:3115.
Aldrich CL, Stephenson MD, Karrison T, Odem RR, Branch DW,
Scott JR, et al. HLA-G genotypes and pregnancy outcome in couples
with unexplained recurrent miscarriage. Mol Hum Reprod 2001;7:
116772.
Hviid TV, Hylenius S, Hoegh AM, Kruse C, Christiansen OB. HLA-G
polymorphisms in couples with recurrent spontaneous abortions. Tissue Antigens 2002;60:12232.
Rebmann V, van der Ven K, Passler M, Pfeiffer K, Krebs D, GrosseWilde H. Association of soluble HLA-G plasma levels with HLA-G
alleles. Tissue Antigens 2001;57:1521.
Hviid TV, Hylenius S, Roerbye C, Nielsen LG. HLA-G allelic variants are associated with differences in the HLA-G mRNA isoform
profile and HLA-G mRNA levels. Immunogenetics 2003;545:6379.
Kamradt T, Mitchison NA. Tolerance and autoimmunity. N Engl
J Med 2001;344:655 64.
Mowbray JF, Gibbings C, Liddell H, Reginald PW, Underwood JL,
Beard RW. Controlled trial of treatment of recurrent spontaneous
abortion by immunisation with paternal cells. Lancet 1985;1:9413.
Kilpatrick DC. Soluble interleukin-2 receptors in recurrent miscarriage and the effect of leukocyte immunotherapy. Immunol Lett 1992;
34:201 6.
Takakuwa K, Higashino M, Yasuda M, Ishii S, Ueda H, Asano K, et
al. Is an additional vaccination necessary for a successful second
pregnancy in unexplained recurrent aborters who were successfully
immunized with their husbands lymphocytes before the first pregnancy? Am J Reprod Immunol 1993;29:39 44.
Gafter U, Sredni B, Segal J, Kalechman Y. Suppressed cell-mediated
immunity and monocyte and natural killer cell activity following
allogeneic immunization of women with spontaneous recurrent abortion. J Clin Immunol 1997;17:408 19.
Hayakawa S, Karasaki-Suzuki M, Itoh T, Ishii M, Kanaeda T, Nagai
N, et al. Effects of paternal lymphocyte immunization on peripheral
Th1/Th2 balance and TCR V beta and V gamma repertoire usage of
patients with recurrent spontaneous abortions. Am J Reprod Immunol
2000;43:10715.
Daya S, Gunby J, The Recurrent Miscarriage Trialists Group. The
effectiveness of allogeneic leukocyte immunization in unexplained
primary recurrent abortion. Am J Reprod Immunol 1994;32:294 302.
Thellin O, Coumans B, Zorzi W, Igout A, Heisen E. Tolerance to the
foeto-placental graft: ten ways to support a child for nine months.
Curr Opin Immunol 2000;12:7317.
Veenstra van Nieuwenhoven AL, Heineman MJ, Faas MM. The
immunology of successful pregnancy. Hum Reprod Update 2003;9:
34757.
Aluvihare VR, Kallikourdis M, Betz AG. Regulatory T cells mediate
maternal tolerance to the fetus. Nat Immunol 2004;5:266 71.
Wood KJ, Jones ND, Bushell AR, Morris PJ. Alloantigen-induced
specific immunological unresponsiveness. Phil Trans R Soc Lond
2001;356:665 80.

Fertility and Sterility

130. Ansari MJ, Sayegh MH. Clinical transplantation tolerance: the promise and challenges. Kidney Int 2004;65:1560 3.
131. Clark DA, Coulam CB, Daya S, Chaouat G. Unexplained sporadic and
recurrent miscarriage in the new millennium: a critical analysis of
immune mechanisms and treatments. Hum Reprod Update 2001;7:
50111.
132. Fedorcsak P, Storeng R, Dale PO, Tanbo T, Abyhom T. Obesity is a
risk factor for early pregnancy loss after IVF or ICSI. Acta Obstet
Gynecol Scand 2000;79:43 8.
133. Bellver J, Rossal LP, Bosch E, Zuniga A, Corona JT, Melendez F, et
al. Obesity and the risk of spontaneous abortion after oocyte donation.
Fertil Steril 2003;79:1136 40.
134. Infante-Rivard C, Fernandez A, Gauthier R, David M, Rivard GE.
Fetal loss associated with caffeine intake before and during pregnancy.
JAMA 1993;270:2940 3.
135. Fenster L, Hubbard AE, Swan SH, Windham GC, Waller K, Hiatt RA,
et al. Caffeinated beverages, decaffeinated coffee, and spontaneous
abortion. Epidemiology 1997;8:51523.
136. Rasch V. Cigarette, alcohol, and caffeine consumption: risk factors for
spontaneous abortion. Acta Obstet Gynecol Scand 2003;82:182 8.
137. Giannelli M, Doyle P, Roman E, Pelerin M, Hermon C. The effect of
caffeine consumption and nausea on the risk of miscarriage. Paediatr
Perinat Epidemiol 2003;17:316 23.
138. Windham GC, Von Behren J, Waller K, Fenster L. Exposure to
environmental and mainstream tobacco smoke and risk of spontaneous
abortion. Am J Epidemiol 1999;149:2437.
139. Nielsen GL, Sorensen HT, Larsen H, Pedersen L. Risk of adverse
outcome and miscarriage in pregnant users of non-steroidal antiinflammatory drugs: population based observational study and casecontrol study. BMJ 2001;322:266 70.
140. Li DK, Liu L, Odouli R. Exposure to nonsteroidal anti-inflammatory
drugs during pregnancy and risk of miscarriage: population based
cohort study. BMJ 2003;327:368.
141. Coste J, Job-Spira N, Fernandez H. Risk factors for spontaneous
abortion: a case-control study in France. Hum Reprod 1991;6:13327.
142. Wisborg K, Kesmodel U, Henriksen TB, Hedegaard M, Secher NJ. A
prospective study of maternal smoking and spontaneous abortion.
Acta Obstet Gynecol Scand 2003;82:936 41.
143. Brandt LP, Nielsen CV. Job stress and adverse outcome of pregnancy:
a causal link or recall bias? Am J Epidemiol 1992;35:30211.
144. Florack EI, Zielhuis GA, Pellegrino JE, Rolland R. Occupational
physical activity and the occurrence of spontaneous abortion. Int J
Epidemiol 1993;22:878 84.
145. McDonald AD, McDonald JC, Armstrong B, Cherry NM, Cote R,
Lavoie J, et al. Fetal death and work in pregnancy. Br J Ind Med
1988;45:148 57.
146. Erza Y, Schenker JG. Abortion rate in assisted reproductiontrue
increase? Early Pregnancy 1995;1:1715.
147. Wang JX, Norman RJ, Wilcox AJ. Incidence of spontaneous abortion
among pregnancies produced by assisted reproductive technology.
Hum Reprod 2004;19:2727.
148. Christiansen OB. Epidemiological, immunogenetic and immunotherapeutic aspects of unexplained recurrent miscarriage. Dan Med Bull
1997;44:396 424.
149. Roberts JM, Macara LM, Chalmers EA, Smith GCS. Inter-assay
variation in antiphospholipid antibody testing. Br J Obstet Gynaecol
2002;109:348 9.
150. Wells MD, Kutteh WH, Ke RW. Diminished ovarian reserve is
associated with tobacco use and poor reproductive performance. Obstet Gynecol 2004;103:1245.
151. Strandell A, Lindhard A, Waldenstrom U, Thorburn J. Hydrosalpinx
and IVF outcome: cumulative results after salpingectomy in a randomized controlled trial. Hum Reprod 2001;16:240310.
152. Daya S. The challenge of meta-analysis and the need to register
clinical trials. Fertil Steril 2000;74:420 2.

839