Heart Failure

Relationship between sex, ejection fraction, and B-type

natriuretic peptide levels in patients hospitalized with
heart failure and associations with inhospital
outcomes: Findings from the Get With The
GuidelineHeart Failure Registry
Eileen M. Hsich, MD, a Maria V. Grau-Sepulveda, MD, MPH, b Adrian F. Hernandez, MD, MHS, b
Zubin J. Eapen, MD, b Ying Xian, BM, PhD, b Lee H. Schwamm, MD, c Deepak L. Bhatt, MD, MPH, d and
Gregg C. Fonarow, MD e Cleveland, OH; Durham, NC; Boston, MA; and Los Angeles, CA

Background In heart failure (HF), there are known differences in plasma B-type natriuretic peptide (BNP) levels between
reduced and preserved ejection fraction (EF), but few HF studies have explored sex differences. We sought to evaluate the
relationship between sex, EF, and BNP in HF patients and determine prognostic significance of BNP as it relates to sex and EF.
Methods

We included hospitals in Get With The GuidelinesHeart Failure that admitted 99,930 HF patients with
reduced (EF b40%), borderline (EF 40%-49%), or preserved (EF 50%) EF. The primary end point was inhospital mortality.
Multivariate models were used to compute odds ratios while accounting for hospital clustering.

Results

There were 47,025 patients with reduced (37% female), 13,950 with borderline (48% female), and 38,955 with
preserved (65% female) EF. Women compared with men had higher admission median BNP levels with the greatest difference
among reduced EF and smallest difference among preserved EF (median BNP in women vs men: EF reduced 1,259 vs 1,113
pg/mL, borderline 821 vs 732 pg/mL, and preserved 559 vs 540 pg/mL; P b .001 all comparisons). Ejection fraction and sex
were independently associated with BNP. Inhospital mortality was 2.7%, and patients above the median BNP level had higher
mortality than those below. After adjusting for over 20 clinical variables, the ability of BNP to predict inhospital mortality was
similar among all subgroups (P for heterogeneity = .47).

Conclusions In a large registry, we found that despite sex/EF differences in BNP values, there was no significant
difference in the ability of BNP to predict inhospital mortality among these subgroups. (Am Heart J 2013;166:1063-1071.e3.)

Levels of B-type natriuretic peptide (BNP) and Nterminal (NT) pro-BNP have been shown to be elevated
in patients hospitalized with heart failure (HF) and to be
predictive of mortality. 1-6 To date, there are few studies
evaluating the relationship between sex, left ventricular

From the aCollege of Medicine, Cleveland Clinic, Cleveland, OH, bDuke Clinical Research
Center, Durham, NC, cDepartment of Neurology, Massachusetts General Hospital, Boston,
MA, dDivision of Cardiology VA Boston Healthcare System, Brigham and Women's
Hospital, and Harvard Medical School, Boston, MA, and eDivision of Cardiology,
University of California, Los Angeles, CA.
Jerome L. Fleg, MD, served as guest editor for this article.
Submitted June 12, 2013; accepted August 30, 2013.
Reprint requests: Eileen Hsich, MD, Kaufman Center for Heart Failure, Heart and Vascular
Institute, Cleveland Clinic, J3-4, 9500 Euclid Ave, Cleveland, OH 44195.
E-mail: Hsiche@ccf.org
0002-8703/$ - see front matter
2013, Mosby, Inc. All rights reserved.
http://dx.doi.org/10.1016/j.ahj.2013.08.029

ejection fraction (EF) (LVEF), and plasma BNP in patients

hospitalized with HF. For the normal population, BNP
levels are higher in women than men. 7,8 In HF, very little
data are available to evaluate sex differences, but 1
modestly sized study noted that a BNP level N500 pg/mL
in patients with shortness of breath was a more powerful
predictor of death in women compared with men. 9 As for
the relationship between EF and plasma BNP levels, many
studies have shown that BNP levels are lower in patients
with preserved EF compared with reduced EF, 2,10,11
making it imperative that any evaluation for sex
differences in HF takes into account differences in EF.
To further explore sex differences, we used Get With
The GuidelinesHeart Failure (GWTG-HF) registry to (1)
determine if there were sex differences in admission
plasma BNP for patients hospitalized with HF with
reduced, preserved, and borderline EF and (2) determine
the prognostic significance of admission BNP for

1064 Hsich et al

inhospital mortality, hospital length of stay (LOS), and

discharge location as it relates to sex and EF.

Methods
Data source
We used the American Heart Association GWTG-HF registry,
which has been previously described. 12-14 Hospitals participating in the registry collect data for patients admitted with HF as
the primary diagnosis to receive recommendations for qualitative improvement in medical management. These hospitals
included large tertiary medical centers as well as small
community hospitals across the United States. Data collected
for each HF patient included demographics, medical/surgical
history, admission medications, physical examination, serum
laboratory tests, inhospital outcomes, and discharge information. Trained hospital personnel entered the data using a Webbased patient management tool (Patient Management Tool,
Outcome, A Quintiles Company). All participating hospitals
were required to submit the GWTG protocol to their
institutional review board. Sites were granted a waiver of
informed consent under the common rule because data
collected were used for qualitative hospital improvement.
Data collected were reviewed for completeness and accuracy.
Only fully participating hospital sites were included in the
analyses. Data collection was performed by Outcome, A
Quintiles Company, and the Duke Clinical Research Institute
(Durham, NC) served as the data analysis center presenting data
in aggregate with deidentified information. Funding for this
study was provided by the American Heart Association.

Study population
From January 1, 2005, to June 29, 2012, there were 196,302
patients from 320 hospitals fully participating in GWTG-HF who
were admitted with HF. Patients were stratified based on their
EF into 3 categories: (1) reduced EF (EF b40%), (2) borderline EF
(EF 40%-49%), and those with preserved EF (EF 50%), as
previously described. 15 We excluded patients with no documentation of sex (n = 4,384), EF (n = 21,400), and BNP (n =
65,618). We also excluded patients who were either transferred
or missing discharge status information (n = 4,970). The final
study population consisted of 99,930 HF patients from 277 fully
participating hospitals (Figure 1). Patients without BNP had
similar characteristics and inhospital mortality to those with
BNP measured and included in the final cohort (online
Appendix Supplementary Table I).

Outcome measures
The primary end point was inhospital mortality based on EF,
sex, and median BNP level. The secondary end points were
hospital LOS and percentage of patients discharged to facilities
other than home.

Statistical analysis
We compared sex-specific baseline characteristics for hospitalized HF patients admitted with reduced, borderline, and
preserved EF. Continuous variables were expressed as medians
with 25th and 75th interquartile ranges (IQR), and categorical
variables were expressed as frequencies. Statistical significance
was evaluated using Pearson 2 test for categorical variables and

American Heart Journal

December 2013

nonparametric Kruskal-Wallis test for continuous. To determine

if sex and EF were independently associated with admission
BNP levels and log BNP, we used multivariable linear models
with the generalized estimated equation method to account for
hospital clustering of patients. The 6 groups analyzed were
women with reduced EF, men with reduced EF, women with
borderline EF, men with borderline EF, women with preserved
EF, and men with preserved EF. Men with reduced EF were
used as the reference. To further assess the relationship
between inhospital mortality and BNP, we calculated inhospital mortality for each of the 6 subgroups and stratified the
data by those above and below the median BNP value.
Unadjusted and mortality adjusted figures were created. The
multivariable adjusted mortality rate for each EF-Sex-BNP
subgroup was calculated as the observed mortality rate divided
by the expected mortality rate in each subgroup based on a
validated risk score for inhospital mortality in patients with HF
from the American Heart Association GWTG-HF program. 16
This value was then multiplied by the overall observed
mortality rate. Observed/expected ratios were then compared
across subgroups using a Wald test by treating the observed
number of deaths as binomial and the expected number of
deaths as fixed.
Multivariable logistic regression modeling was used to assess
the relationship of log-transformed BNP with inhospital
mortality for HF patients in each of the 6 sex/EF groups.
Because certain laboratory values were not available in all
patients, we developed 2 models for each sex/EF group. Model 1
included demographic and clinical variables with no or b15%
missing values. Variables included age, systolic blood pressure
(SBP), heart rate, body mass index (BMI), race, anemia, ischemic
history, diabetes mellitus, hyperlipidemia, hypertension, smoking, chronic obstructive pulmonary disease/asthma, peripheral
vascular disease, renal failure/dialysis, depression, valve disease,
cardiac resynchronization therapy defibrillator (CRT-D), atrial
fibrillation/flutter, and cerebral vascular accident/transient
ischemic attack. Missing continuous values in model 1 were
imputed to the median for each sex/EF group and to the most
common category for categorical variables. Hospital characteristics (region, rural location, no. of beds, academic center, and
heart transplant center) were also included in model 1, but
missing values were not imputed. Model 2 was similar to model
1 but limited to only patients with complete data and included
laboratory variables (sodium, hemoglobin, abnormal troponin,
creatinine, and blood urea nitrogen [BUN]). Cubic spline plots
were developed to explore the fit of continuous variables in
the models. Subsequently, they were coded as follows: age,
SBP, and BUN linear continuous; heart rate splines with knots
at 75 and 105 ppm; BMI splines with knot at 30; sodium
splines with knot at 140 mmol/L; hemoglobin truncated at 12
g/dL; and serum creatinine truncated at 1 and 3.5 mg/dL.
P values for the 2 test of heterogeneity were determined to
assess for differences in the ability of log BNP to predict the
outcome among all sex/EF subgroups, including a multiplicative interaction term (log BNP * sex/LVEF groups) in each
model. Secondary outcomes to assess the relationship of logBNP to LOS and discharge to facility other than home were
performed with multivariable logistic regression modeling.
Models created were similar to 1 and 2 above.
To graphically assess relationship between inhospital death
rates and BNP as a continuous variable, we used logistic

American Heart Journal

Volume 166, Number 6

Hsich et al 1065

Figure 1

Flow diagram showing study design. Flow diagram showing original GWTG-HF cohort, number of patients excluded, and final cohort. The final
cohort was then divided into 3 groups based on EF (LVEF b40%, LVEF 40%-49%, and LVEF N50%). These groups were further stratified by sex.

regression models stratified by EF and sex, unadjusted and

adjusted, with BNP included as restricted cubic spline to avoid
any assumption of linearity, and computed predicted probabilities for each value in the observed range of BNP.
Unadjusted and multivariable adjusted plots were created.
Variables in the adjusted model were the same as in model 1
above. The adjusted predicted probabilities were calculated for
the average patient in each EF-Sex subgroup, who represented
a patient with the mean value for each variable in the model in
that sex/EF subgroup.
All analyses were performed using SAS software (version 9.2;
SAS Institute, Cary, NC). Probability values were 2 sided with P
b .05 considered statistically significant.
The authors are solely responsible for the design and conduct
of this study, all study analyses, the drafting and editing of the
manuscript, and its final contents.

Results
The study cohort of 99,930 HF patients consisted of
47,025 patients admitted with HF and reduced EF (37%
female, 63% male), 13,950 HF patients with borderline EF
(48% female, 52% male), and 38,955 HF patients admitted
with preserved EF (65% female, 35% male) from 277 fully
participating hospital GWTG-HF hospitals. The median
BNP was 816 pg/mL for the cohort with an IQR of 380 to
1,670 pg/mL. Table I shows the baseline characteristics of
the cohort according to sex and EF. Median BNP levels on
admission were slightly to moderately higher in women
compared with men for all subgroups. In the reduced EF
group, the median BNP level for women was 1,259 pg/mL
compared with 1,113 pg/mL for men (P b .0001). In the
borderline EF group, the median BNP level for women

American Heart Journal

December 2013

1066 Hsich et al

Table I. Baseline characteristics for HF patients stratified by sex and LVEF

Characteristics

All patients
n = 99930

Age, median (IQR), y

74 (62-83)
Race
White, n (%)
66463 (67)
Black, n (%)
21336 (21)
Asian, n (%)
1320 (1)
Hispanic, n (%)
7344 (7)
Other, n (%)
2656 (3)
Prior HF, n (%)
61567 (62)
Atrial fibrillation, n (%)
33673 (34)
COPD or asthma, n (%)
31215 (32)
CVA or TIA, n (%)
15185 (15)
Depression, n (%)
10357 (10)
NITDM, n (%)
22728 (23)
ITDM, n (%)
21328 (22)
Renal insufficiency, n (%)
21381 (22)
Dialysis, n (%)
3647 (4)
Anemia, n (%)
19874 (20)
Hyperlipidemia, n (%)
46724 (47)
Hypertension, n (%)
78146 (79)
CAD, n (%)
49431 (50)
ICM, n (%)
56337 (57)
Valvular disease, n (%)
13922 (14)
PVD, n (%)
12909 (13)
Smoking, n (%)
17388 (17)
LVEF (%), median (IQR)
40 (25-55)
HR, beat/min,
83 (71-98)
median (IQR)
SBP, mm Hg, median (IQR) 139 (119-160)
BMI, median (IQR)
28 (24-34)
Medications on admission,
n (%)
ACE I
35086 (38)
ARB
13285 (14)
-Blocker
48046 (52)
Aldosterone antagonist
9974 (11)
Digoxin
15115 (16)
Hydralazine
7163 (7)+
Nitrate
16819 (18)
Diuretics
60531 (65)
Anticoagulant
22801 (25)
Antiarrhythmic
8518 (9)
Aspirin
43032 (46)
CCB
17381 (19)
Serum values at admission,
median (IQR)
Sodium (mEq/L)
138 (135-141)
Hgb (g/dL)
12 (11-13)
BNP (pg/mL)
816
(380-1670)
Cr (mg/dL)
1.3 (1.0-1.9)
BUN (mg/dL)

25 (17-38)

EF b40%

EF b40%

EF 40%-49%

EF 40%-49%

EF 50%

EF 50%

Female

Male

Female

Male

Female

Male

n = 17447

n = 29578

n = 6752

n = 7198

n = 25244

n = 13711

74 (62-83)

69 (57-79)

78 (67-85)

74 (63-83)

79 (69-86)

74 (63-83)

10508 (60)
4875 (28)
194 (1)
1213 (7)
481 (3)
11145 (64)
4788 (28)
5111 (30)
2667 (15)
2011 (12)
3907 (23)
3537 (20)
3314 (19)
492 (3)
3231 (19)
7705 (45)
13331 (77)
8341 (48)
9640 (56)
2527 (15)
1935(11)
2897 (17)
25 (20-31)
87 (74 -102)

18290 (62)
7424 (25)
389 (1)
2407 (8)
798 (3)
19210 (66)
9466 (32)
8335 (28)
4152 (14)
2223 (8)
6621 (23)
5534 (19)
7058 (24)
854 (3)
4273 (15)
14065 (48)
21799 (74)
16297 (56)
18769 (64)
3462 (12)
3955 (14)
7398 (25)
25 (19-30)
86 (73-100)

4681 (69)
1238 (18)
88 (1)
514 (8)
173 (3)
4100 (61)
2438 (36)
2124 (32)
1109 (17)
857 (13)
1560 (23)
1538 (23)
1485 (22)
263 (4)
1663 (25)
3201 (48)
5455 (81
3444 (51)
3914 (58)
1093 (16)
851 (13)
825 (12)
44 (40-45)
83 (72-98)

5127 (71)
1162 (16)
103 (1)
539 (7)
209 (3)
4295 (60)
2641 (37)
2304 (32)
1104 (15)
588 (8)
1719 (24)
1686 (24)
1981 (28)
331 (5)
1422 (20)
3546 (50
5578 (78)
4177 (58)
4743 (66)
949 (13)
1131 (16)
1266 (18)
43 (40-45)
80 (70-95)

18060 (72)
4373 (17)
347 (1)
1685 (7)
628 (2)
15016 (60)
9230 (37)
8516 (34)
4178 (17)
3403 (14)
5616 (22)
5670 (23)
5207 (21)
957 (4)
6317 (2)
11586 (46)
20980 (84)
10284 (41)
11559 (46)
4061 (16)
2984 (12)
2688 (11)
60 (55-65)
80 (69-94)

9797 (71)
2264 (17)
199 (1)
986 (7)
367 (3)
7801 (57)
5110 (38)
4825 (35)
1975 (15)
1275 (9)
3305 (24)
3362 (25)
3842 (28)
750 (6)
2968 (22)
6621 (49)
11003 (81)
6888 (51)
7712 (57)
1830 (13)
2053 (15)
2314 (17)
56 (53-60)
79 (68-92)

135 (116-155)
27 (23-33)

130 (112-150)
28 (24-33)

144 (125-165)
28 (24-34)

142(122-162)
29 (25-34)

146 (126-168)
29 (24 -37)

144 (124-166)
30 (25-36)

6562 (41)
2402 (15)
8345 (52)
2396 (15)
3250 (20)
1152 (7)
3095 (19)
10756 (67)
3420 (21)
1343 (8)
7546 (47)
1909 (12)

11863 (44)
2994 (11)
14560 (54)
4331 (16)
5997 (22)
2187 (7)
5082 (19)
18137 (67)
7007 (26)
3271 (12)
13348 (49)
2626 (10)

2196 (35)
1045 (17)
3251 (51)
495 (8)
864 (14)
473 (7)
1337 (21)
4094 (65)
1546 (24)
503 (8)
2912 (46)
1310 (21)

2564 (38)
878 (13)
3507 (52)
580 (9)
921 (14)
515 (7)
1244 (19)
4211 (63)
1786 (27)
620 (9)
3296 (49)
1249 (19)

7374 (31)
4230 (18)
11806 (50)
1389 (6)
2694 (11)
1736 (7)
4064 (17)
15366 (65)
5781 (24)
1757 (7)
10034 (42)
6897 (29)

4527 (35)
1736 (14)
6577 (51)
783 (6)
1389 (11)
1100 (8)
1997 (16)
7967 (62)
3261 (26)
1024 (8)
5896 (46)
3390 (27)

138 (135-140)
12 (11-13)
1259
(606-2413)
1.20 (0.901.70)
24 (16-36)

138 (135-140)
13 (11-14)
1113
(535-2130)
1.40 (1.101.90)
26 (18-39)

138 (135-141)
11 (10-13)
821
(421-1574)
1.20(0.901.70)
24 (17-37)

138 (136-141)
12 (10-13)
732
(366-1420)
1.40(1.102.10)
26 (18-400

138 (135-141)
11 (10-13)
559
(279-1075)
1.20(0.901.70)
24 (17-36)

138 (136-141)
12 (10-13)
540
(253-1064)
1.50(1.102.10)
27 (18-41)

NITDM, Noninsulin-dependent diabetes mellitus; ITDM, insulin-dependent diabetes mellitus; ICM, ischemic cardiomyopathy; CAD, coronary artery disease; PVD, peripheral vascular
disease; SBP, systolic blood pressure; ACE I, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; CCB, calcium-channel blocker; Hgb, hemoglobin.

American Heart Journal

Volume 166, Number 6

Figure 2

Inhospital mortality based on sex, median BNP level, and EF. Sex
differences in inhospital mortality for patients admitted with acute
decompensated HF in the GWTG-HF registry with reduced, borderline, and preserved EF according to serum laboratory tests at time of
admission above or below the median BNP values. Reduced EF was
defined as an EF b40%, borderline EF was defined as EF 40% to 49%,
and preserved EF was defined as EF 50%. Panel A, Unadjusted
mortality data. Panel B, Multivariable adjusted mortality data The
multivariable adjusted mortality rate for each EF-Sex-BNP subgroup
was calculated as the observed mortality rate divided by the expected
mortality rate in each subgroup based on a validated risk score for
inhospital mortality in patients with HF from the American Heart
Association GWTG-HF program. 16 This value was then multiplied by
the overall observed mortality rate. Observed/expected ratios were
compared across subgroups using a Wald test by treating the
observed number of deaths as binomial and the expected number of
deaths as fixed.

was 821 pg/mL compared with 732 pg/mL for men (P b

.0001). In the preserved EF group, the median BNP level
for women was 559 pg/mL compared with 540 pg/mL for
men (P b .0004).
Heart failure patients with reduced EF compared with
preserved EF were younger, had more black patients, and
fewer women (Table I). Those with borderline EF were

Hsich et al 1067

similar in age and racial distribution to preserved EF but

with almost equal number of women and men. Patients
with borderline EF had BNP levels between the median
values obtained for those with reduced and preserved EF.
Among patients with reduced EF, women were more
likely than men to be older and have hypertension,
anemia, depression, or valve heart disease and less likely
to have coronary artery disease, ischemic cardiomyopathy, atrial fibrillation, or tobacco usage. Similar sex
differences were notable for HF patients with borderline
and preserved EF except for prior atrial fibrillation, which
was similarly frequent among women and men. Medications from home recorded on admission were similar
between women and men. Serum laboratory tests were
notable for women having slightly lower serum creatinine levels.
Online Appendix Supplementary Table I summarizes
the relative difference between BNP levels among the
sex/EF subgroups using males with LVEF b40% as the
reference group. After adjusting for N20 possible
confounders including age, body mass index, and renal
function, there was still a significant difference between
admission BNP levels among the 6 subgroups.
The inhospital mortality rate for the overall population
was 2.7%, and there were no sex differences among
patients with reduced, borderline, and preserved EF.
When patients were further stratified based on BNP,
unadjusted and multivariable adjusted data showed that
there was a higher inhospital mortality rate among
women and men with BNP above the median levels
compared with BNP below the median levels. The
mortality rate among patients with BNP above the median
levels was slightly higher among women and men with EF
N50% compared with patients with EF b50%(Figure 2A
and B). We also assessed the relationship between
inhospital mortality rates and BNP as a continuous variable
by EF and sex graphically. Unadjusted (Figure 3A) and
multivariable adjusted (Figure 3B) data are presented as
predictive probability plots. For HF patients with reduced
EF, men had a higher mortality rate than women for any
given BNP level between the values 500 to 3,000 pg/mL.
The relationship between BNP and inhospital mortality
was nearly linear with higher mortality rates for both
women and men with higher BNP levels. The predictive
probability plots for women and men with preserved EF
were notable for higher mortality rates for any given BNP
value compared with patients with reduced EF. Sex
differences were also present for BNP and outcome
among patients with preserved EF, but the higher
mortality seen among men with BNP levels between 500
and 3,000 pg/mL was only present in the unadjusted data.
Unadjusted and adjusted inhospital mortality risks for
log-BNP values among women and men admitted with
acute decompensated HF were assessed according to EF
(Table II). Unadjusted log-BNP values were associated
with a higher inhospital mortality rate among women

1068 Hsich et al

American Heart Journal

December 2013

Figure 3

Predictive probability plots of BNP levels and inhospital mortality for women and men with acute decompensated HF. Predictive probability plots of
BNP levels and inhospital mortality risk according to sex and EF. Panel A, Plot of BNP levels and unadjusted inhospital mortality among women
and men with HF and EF b40% and EF 50%. Panel B, Plot of BNP levels and adjusted inhospital mortality among women and men with HF and EF
50%. The adjusted plot computes the predicted probability for each value of BNP in the average patient adjusting for the following variables:
chronic obstructive lung disease, hypertension, hyperlipidemia, peripheral vascular disease, cerebral vascular accident/transient ischemic attack,
anemia, renal failure/dialysis, depression, valve disease, smoking, diabetes mellitus, atrial fibrillation/flutter, ischemic history, ICD-CRT-D, race,
age, BMI, heart rate, SBP, and hospital characteristics (region, rural location, no. of beds, academic center and heart transplant center).

than men with reduced, borderline, and preserved EF.

After adjusting for over 20 clinical variables including
age, BMI, and renal insufficiency, the prognostic value
of BNP to predict inhospital mortality was similar
among all 6 subgroups (model 1: limited to hospitals
with few missing data, P for heterogeneity = .47, model
2: limited to observations with no missing data, P for
heterogeneity = .07).
The median hospital LOS in the population was 4 days.
The mean LOS was similar among all sex/EF subgroups

and ranged from 5.74 to 5.85 days. Patients with a higher

than median BNP level compared with those with lower
median BNP level had slightly more prevalent LOS N4 days
regardless of EF and with no clinically relevant sex
differences (online Appendix Supplementary Table II). Btype natriuretic peptide higher than the median level was
also associated with more prevalent discharges to facilities
other than home (online Appendix Supplementary Table
II). When comparing patients with reduced, borderline,
and preserved EF, the frequency of hospital discharges to

American Heart Journal

Volume 166, Number 6

Hsich et al 1069

Table II. Adjusted odds ratio for log BNP and inhospital mortality among HF patients stratified by EF/sex
Model 1, adjusted odds ratio (95% CI), P value

Odds ratio

All patients
(n = 98579)

EF b40%
female
(n = 17262)

EF b40%
male
(n = 29171)

EF 40%-49%
female
(n = 6666)

EF 40%-49%
male
(n = 7085)

EF 50%
female
(n = 24907)

EF 50%
male
(n = 13488)

Unadjusted

1.45 (1.32-1.58), 1.51 (1.23-1.86), 1.48 (1.27-1.74), 1.35 (1.04-1.75), 1.28 (1.05-1.56), 1.53 (1.38-1.71), 1.32 (1.16-1.52),
P b .0001
P = .0001
P b .0001
P = .0265
P = .0163
P b .0001
P b .001
Multivariable 1.34 (1.14-1.58), 1.25 (0.97-1.61), 1.28 (1.05-1.56), 1.20 (0.84-1.71), 1.18 (0.92-1.52), 1.52 (1.34-1.71), 1.25 (1.02-1.52),
adjusted
P b .0004
P = .0876
P = .0161
P = .3231
P = .1857
P b .0001
P = .0291
Model 2, adjusted odds ratio (95% CI), P value

Odds ratio

All patients
(n = 74516)

EF b40%
female
(n = 11798)

EF b40%
male
(n = 20054)

EF 40%-49%
female
(n = 4697)

EF 40%-49%
male
(n = 4881)

EF 50%
female
(n = 17243)

EF 50%
male
(n = 9088)

1.45
(1.31-1.61),
P b .0001
1.20
(1.07-1.35),
P = .0025

1.49
(1.17-1.90),
P = .001
1.08
(0.97-1.20),
P = .1802

1.47
(1.23-1.76),
P b .0001
1.12
(1.00-1.25),
P = .0515

1.34
(1.00-1.79),
P = .0494
1.05
(0.88-1.24),
P = .6056

1.29
(1.03-1.60),
P = .0262
1.09
(0.97-1.21),
P = .1413

1.61
(1.41-1.84),
P b .0001
1.42
(1.22-1.64),
P b .0001

1.32
(1.14-1.54),
P = .0003
1.15
(0.99-1.33),
P = .0589

Unadjusted

Multivariable
adjusted

Model 1 included demographic, clinical, and hospital variables with no or b15% missing values. Variables included age, SBP, heart rate, BMI, race, anemia, ischemic history, diabetes
mellitus, hyperlipidemia, hypertension, smoking, chronic obstructive pulmonary disease/asthma, peripheral vascular disease, renal failure/dialysis, depression, valve disease, ICDCRT-D, atrial fibrillation/flutter and cerebral vascular accident/transient ischemic attack, region, hospital teaching status, hospital location rural, number of beds, and heart transplant
capability. P for heterogeneity between log BNP and the 6 sex/ejection fraction subgroups is .4709.
Model 2 was similar to model 1 but limited to only patients with complete data and included laboratory variables (sodium, hemoglobin, abnormal troponin, creatinine, and BUN). P for
heterogeneity between log BNP and the 6 sex/ejection fraction subgroups is .0717.

a facility other than home was more common in women

than men and greater in the preserved EF cohort
compared with the reduced EF cohort.
Unadjusted log-BNP values were associated with a
higher rate of LOS N4 days among women and men
with reduced, borderline, and preserved EF (online
Appendix Supplementary Table III). After adjusting for
over 20 clinical variables including age, BMI, and renal
insufficiency, the prognostic value of log BNP to predict
LOS N4 days was similar among all 6 subgroups (model
1: limited to hospitals with few missing data, model 2:
limited to observations with no missing data). Unadjusted log-BNP values were also associated with a
higher rate of discharge other than home among
women and men with reduced, borderline, and
preserved EF (online Appendix Supplementary Table
IV). After adjusting for over 20 clinical variables, the
prognostic value of BNP to predict discharge other than
home was similar among all 6 subgroups.

Discussion
In a large, multicenter, national HF registry, we found
women to have higher median BNP levels than men upon
hospital admission for acute decompensated HF with
reduced, borderline, and preserved EF. The median BNP
levels were highest among HF patients with reduced EF
and lowest among those with preserved EF. Admission

BNP levels were independently associated with sex and

EF. Inhospital mortality was 2.7% and similar among
patients with reduced, borderline, and preserved EF.
When patients were further stratified by BNP, patients
above the median level had higher mortality than those
below. After adjusting for over 20 clinical variables
including age, BMI, and renal insufficiency, the ability of
BNP to predict inhospital mortality was similar among the
6 subgroups.
The present study reinforces the existing literature
noting that BNP levels are lower in HF patients with
preserved EF compared with reduced EF 2,10,11 and
expands on our current knowledge regarding sex
differences. Several publications have shown for the
normal population, BNP levels are higher in women than
men, 7,8 but the literature has not been consistent with
decompensated HF. 17-19 This likely is due to the fact that
the general population is a homogenous group with
respect to EF, whereas HF patients consist of those with
preserved, borderline, or reduced EF. Because more
women than men have HF with preserved EF, 20 HF
studies that do not separate the cohort by EF would be
expected to have lower median BNP levels in women
than men. This assumption is consistent with existing HF
literature demonstrating women with lower BNP levels
than men when the cohort combines all EF 17,18 and
higher BNP levels in women when the data are stratified
by EF. 19

American Heart Journal

December 2013

1070 Hsich et al

The ability of BNP to predict prognosis for HF patients

has been well established with higher levels portending a
worse prognosis. Many studies involving hospitalized HF
patients have shown the use of BNP to predict mortality,
myocardial infarction, and HF readmission. 1-4 In the
Acute Decompensated HF National Registry (ADHERE),
higher BNP levels were associated with higher inhospital
mortality for both patients with reduced and preserved
EF. The ADHERE study analyzed N45,000 acute decompensated HF admissions and stratified patients based on
EF and BNP quartile. 2 A subgroup of the ADHERE registry
consisting of 11,679 patients admitted with HF age 65
years and older who did not have atrial fibrillation were
followed up for up to a year postdischarge to assess the
ability of BNP to predict all-cause mortality, thromboembolic events, myocardial infarction, and ischemic stroke.
In that study, higher BNP levels on hospital admission
were associated with increased risk of all-cause mortality
and myocardial infarction but not associated with stroke
or other thromboembolic events. 3 Our study with nearly
100,000 HF admissions has similar findings to the
ADHERE 2 with higher inhospital mortality rates among
HF patients with BNP values above the median level
compared with those below the median BNP for both
women and men with preserved, borderline, and
reduced EF. Despite sex/EF differences in median BNP
levels upon admission, there was no significant difference in inhospital survival among groups. Furthermore,
after adjusting for over 20 clinical variables, the ability of
BNP to predict inhospital mortality was similar among all
6 subgroups.
Our study supports the use of BNP to predict inhospital
survival and adds to the literature by demonstrating that
admission BNP levels are dependent on sex and EF.
Current guidelines recognize the importance of BNP for
diagnosis and prognosis of HF. 21,21 However, the utility
of this biomarker has been limited by intraindividual/
interindividual variability. 22 Until now, variability
appeared random, and significance of 2 patients with
different BNP levels admitted for acute decompensated
HF remained unclear. We have found with a very large
cohort that women and men have different BNP levels,
and these differences are also dependent on whether
they have reduced, borderline, or preserved EF. For
instance, in our study, men with preserved EF with
median BNP levels above 540 pg/mL have similar risk of
inhospital mortality as men with reduced EF with median
BNP above 1,113 pg/mL. Therefore, sex/EF differences
should be taken into account to best use BNP to predict
inhospital survival in patients admitted with decompensated HF.

Study limitations
The registry is dependent on accuracy of data and
completeness of data abstraction from medical charts.
There were 96,372 patients excluded because of missing

sex, EF, BNP, or incomplete discharge information,

which could impact generalizability of the findings,
although patients with and without BNP measured
were similar in characteristics and inhospital mortality.
It is also important to mention that our findings were
based on results from various commercially available BNP
assays rather than from a single central core laboratory. In
addition, only admission BNP levels were studied and
whether there are sex and EF difference in serial or
discharge BNP levels requires further study. We limited
our study to BNP instead of NT pro-BNP because most of
the participating centers used BNP testing for measurement of natriuretic peptides. The prognostic abilities of
BNP and NT pro-BNP appear similar based on prior
literature. 22 Residual measured or unmeasured confounding may account for some or all of these findings. Finally,
postdischarge data, including mortality and readmissions,
were not available because our cohort is a registry of
hospitalized patients. The ability of hospital admission
BNP to further predict long-term survival has been shown
in the ADHERE registry and other studies, but sex
differences were not investigated. 3

Conclusions
In a large, multicenter registry, we found women to
have higher median BNP levels than men upon hospital
admission for HF with reduced, borderline, and preserved EF. Despite sex/EF differences in baseline BNP
values, there were no significant differences in the ability
of BNP to predict inhospital mortality among these
subgroups. These findings expand the knowledge base
regarding sex-based similarities and differences among
hospitalized HF patients.

Disclosures
Funding and relationship with industry: The Get With
The GuidelinesHeart Failure program is provided by the
American Heart Association. The Get With The GuidelinesHeart Failure program has been funded in the past
through support from Medtronic, GlaxoSmithKline,
Ortho-McNeil, and the American Heart Association
Pharmaceutical Roundtable. Adrian Hernandez receives
funding from Johnson & Johnson, and Amylin and has
received honorarium from Amgen and Corthera. Dr
Deepak L. Bhatt discloses the following relationships
Advisory Board: Medscape Cardiology; Board of Directors: Boston VA Research Institute, Society of Chest Pain
Centers; Chair: American Heart Association Get With The
Guidelines Science Subcommittee; Honoraria: American
College of Cardiology (Editor, Clinical Trials, Cardiosource), Duke Clinical Research Institute (clinical trial
steering committees), Slack Publications (Chief Medical
Editor, Cardiology Today Intervention), WebMD (CME
steering committees); Other: Senior Associate Editor,

American Heart Journal

Volume 166, Number 6

Journal of Invasive Cardiology; Research Grants:

Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, Sanofi Aventis, The Medicines Company; Unfunded Research: FlowCo, PLx Pharma, Takeda.
Gregg Fonarow is a consultant for Novartis, Gambro,
and Medtronic.

References
1. Du JB, Da CH, Zhao Y, et al. The role of brain natriuretic peptide and
serum triiodothyronine in the diagnosis and prognosis of chronic
heart failure. Acta Cardiol 2012;67(3):291-6.
2. Fonarow GC, Peacock WF, Phillips CO, et al. Admission B-type
natriuretic peptide levels and in-hospital mortality in acute decompensated heart failure. J Am Coll Cardiol 2007;49(19):1943-50.
3. Kociol RD, Greiner MA, Hammill BG, et al. B-type natriuretic peptide
level and postdischarge thrombotic events in older patients hospitalized with heart failure: insights from the Acute Decompensated Heart
Failure National Registry. Am Heart J 2012;163(6):994-1001.
4. Logeart D, Thabut G, Jourdain P, et al. Predischarge B-type natriuretic
peptide assay for identifying patients at high risk of re-admission
after decompensated heart failure. J Am Coll Cardiol 2004;43(4):
635-41.
5. Cohen-Solal A, Logeart D, Huang B, et al. Lowered B-type natriuretic
peptide in response to levosimendan or dobutamine treatment is
associated with improved survival in patients with severe acutely
decompensated heart failure. J Am Coll Cardiol 2009;53(25):
2343-8.
6. Waldo SW, Beede J, Isakson S, et al. Pro-B-type natriuretic peptide
levels in acute decompensated heart failure. J Am Coll Cardiol 2008;
51(19):1874-82.
7. Redfield MM, Rodeheffer RJ, Jacobsen SJ, et al. Plasma brain
natriuretic peptide concentration: impact of age and gender. J Am
Coll Cardiol 2002;40(5):976-82.
8. Wang TJ, Larson MG, Levy D, et al. Impact of age and sex on plasma
natriuretic peptide levels in healthy adults. Am J Cardiol 2002;90(3):
254-8.
9. Christ M, Laule-Kilian K, Hochholzer W, et al. Gender-specific risk
stratification with B-type natriuretic peptide levels in patients with
acute dyspnea: insights from the B-type natriuretic peptide for acute
shortness of breath evaluation study. J Am Coll Cardiol 2006;48(9):
1808-12.
10. Fonarow GC, Peacock WF, Horwich TB, et al. Usefulness of B-type
natriuretic peptide and cardiac troponin levels to predict in-hospital
mortality from ADHERE. Am J Cardiol 2008;101(2):231-7.

Hsich et al 1071

11. Bursi F, Weston SA, Redfield MM, et al. Systolic and diastolic heart
failure in the community. Jama 2006;296(18):2209-16.
12. Hernandez AF, Fonarow GC, Liang L, et al. Sex and racial differences
in the use of implantable cardioverter-defibrillators among patients
hospitalized with heart failure. JAMA 2007;298(13):1525-32.
13. Hong Y, LaBresh KA. Overview of the American Heart Association
Get with the Guidelines programs: coronary heart disease, stroke,
and heart failure. Crit Pathw Cardiol 2006;5(4):179-86.
14. LaBresh KA, Gliklich R, Liljestrand J, et al. Using get with the
guidelines to improve cardiovascular secondary prevention. Jt
Comm J Qual Saf 2003;29(10):539-50.
15. Steinberg BA, Zhao X, Heidenreich PA, et al. Trends in patients
hospitalized with heart failure and preserved left ventricular ejection
fraction: prevalence, therapies, and outcomes. Circulation 2012;
126(1):65-75.
16. Peterson PN, Rumsfeld JS, Liang L, et al. A validated risk score for inhospital mortality in patients with heart failure from the American
Heart Association get with the guidelines program. Circ Cardiovasc
Qual Outcomes 2010;3(1):25-32.
17. Knudsen CW, Riis JS, Finsen AV, et al. Diagnostic value of a rapid test
for B-type natriuretic peptide in patients presenting with acute
dyspnoe: effect of age and gender. Eur J Heart Fail 2004;6(1):55-62.
18. Tasevska-Dinevska G, Kennedy LM, Cline-Iwarson A, et al. Gender
differences in variables related to B-natriuretic peptide, left ventricular
ejection fraction and mass, and peak oxygen consumption, in
patients with heart failure. Int J Cardiol 2011;149(3):364-71.
19. Hsich EM, Grau-Sepulveda MV, Hernandez AF, et al. Sex differences
in in-hospital mortality in acute decompensated heart failure with
reduced and preserved ejection fraction. Am Heart J 2012;163(3):
430-7, 437 e1-3.
20. Hsich EM, Pina IL. Heart failure in women: a need for prospective
data. J Am Coll Cardiol 2009;54(6):491-8.
21. Jessup M, Abraham WT, Casey DE, et al. 2009 focused update:
ACCF/AHA
Guidelines for the Diagnosis and Management of Heart Failure in
Adults: a report of the American College of Cardiology
Foundation/American Heart Association Task Force on Practice
Guidelines: developed in collaboration with the International Society
for Heart and Lung Transplantation. Circulation 2009;119(14):
1977-2016.
22. Tang WH, Francis GS, Morrow DA, et al. National Academy of
Clinical Biochemistry Laboratory Medicine practice guidelines:
clinical utilization of cardiac
biomarker testing in heart failure. Circulation
2007;116(5):e99-109.

American Heart Journal

Volume 166, Number 6

Hsich et al 1071.e1

Appendix
Supplementary Table I. Estimates for BNP levels by sex/EF adjusting for potential confounders (reference group, male with LVEF b40%)

Variable

EF b40%
Male
(n = 29578)
-coefficient
(95% CI),
P value

EF b40%
Female
(n = 17447)
-coefficient
(95% CI),
P value

EF 40%-49%
Male
(n = 7198)
-coefficient
(95% CI),
P value

EF 40%-49%
Female
(n = 6752)
-coefficient
(95% CI),
P value

EF 50%
Male
(n = 13711)
-coefficient
(95% CI),
P value

EF 50%
Female
(n = 25244)
-coefficient
(95% CI),
P value

0.4246
(0.4578 to
0.3914),
P b .001
431.6
(486.7 to
376.4),
P b .001
0.4446
(0.4864 to
0.4027),
P b .001
438.3
(495.7-381.0),
P b .001

0.2902
(0.3249 to
0.2554),
P b .001
328.2
(379.7 to
276.8),
P b .001
0.2805
(0.3131 to
0.2480),
P b .001
277.5
(327.4-227.6),
P b .001

0.7371
(0.7720 to
0.7023),
P b .001
655.9
(728.4 to
583.4),
P b .001
0.7069
(0.7358 to
0.6780),
P b .001
622.9
699.8-545.9),
P b .001

0.6930
(0.7253 to
0.6607),
P b .001
659.4
(728.7 to
590.0),
P b .001
0.6040
(0.6296 to
0.5784),
P b .001
540.2
(607.6-472.7),
P b .001

Log BNP,
unadjusted

Reference

0.0928
(0.0718-0.1139),
P b .001

BNP,
unadjusted

Reference

Log BNP,
adjusted

Reference

153.5
(122.8184.2),
P b .001
0.0930
(0.0708-0.1152),
P b .001

BNP,
adjusted

Reference

169.2
(139.1-199.4),
P b .001

Variables in the model: age, gender, body mass index, blood urea nitrogen, serum creatinine, sodium, abnormal troponin, hemoglobin, SBP, heart rate, ischemic history, history of
anemia, stroke/transient ischemic attack, diabetes mellitus, hyperlipidemia, hypertension, chronic obstructive pulmonary disease/asthma, peripheral vascular disease, renal failure/
dialysis, depression, valve disease, ICD-CRT-D, atrial fibrillation/flutter, smoking, geographic region, teaching hospital, number of beds, rural location, and heart transplant center.

Supplementary Table II. Inhospital LOS N4 days and hospital discharges other than home for heart failure patients according to sex, BNP,
and EF

LOS N4 d
Below BNP
median, n (%)
Above BNP
median, n (%)

LVEF b40%

LVEF b40%

LVEF 40%-49%

LVEF 40%-49%

LVEF 50%

LVEF 50%

Female

Male

Female

Male

Female

Male

All patients
n = 91906

n = 15996

n = 27135

n = 6218

n = 6651

n = 23204

n = 12702

20626 (45)

2451 (46)

4348 (42)

1404 (45)

1566 (43)

7086 (47)

3771 (45)

22851 (50)

5273 (50)

8270 (49)

1574 (51)

1472 (49)

4164 (52)

2098 (49)

LVEF b40%

LVEF b40%

LVEF 40%-49%

LVEF 40%-49%

LVEF 50%

LVEF 50%

All patients

Female

Male

Female

Male

Female

n = 97231

n = 16971

n = 28707

n = 6585

n = 7019

n = 24613

n = 13336

1043 (18)

1351 (12)

790 (24)

640 (17)

4304 (27)

1685 (19)

2867 (26)

3207 (18)

988 (30)

677 (21)

2850 (34)

1071 (24)

Hospital discharge other than home

Below BNP
9813 (20)
median, n (%)
Above BNP
11660 (24)
median, n (%)

P b .0001 when comparing all groups above and below the BNP median.

Male

American Heart Journal

December 2013

1071.e2 Hsich et al

Supplementary Table III. Adjusted odds ratio for Log BNP and LOS N4 days among heart failure patients stratified by ejection fraction/sex
Model 1

Odd ratio
Unadjusted

Multivariable
adjusted

All
patients
(n = 90745)

EF b40%
female
(n = 15834)

EF b40%
male
(n = 26786)

EF 40%-49%
female
(n = 6147)

EF 40%-49%
male
(n = 6544)

EF 50%
female
(n = 22931)

EF 50%
male
(n = 12503)

1.14
(1.12-1.16)
P b .0001
1.20
(1.16-1.24)
P b .0001

1.13
(1.08-1.18)
P b .0001
1.17
(1.10-1.25)
P b .0001

1.20
(1.15-1.25)
P b .0001
1.23
(1.16-1.29)
P b .0001

1.14
(1.08-1.19)
P b .0001
1.19
(1.13-1.26)
P b .0001

1.12
(1.07-1.18)
P b .0001
1.19
(1.11-1.27)
P b .0001

1.14
(1.11-1.17)
P b .0001
1.19
(1.14-1.24)
P b .0001

1.08
(1.04-1.12)
P b .0001
1.14
(1.10-1.18)
P b .0001

All
patients
(n = 68771)

EF b40%
female
(n = 12020)

EF b40%
male
(n = 20218)

EF 40%-49%
female
(n = 4776)

EF 40%-49%
male
(n = 4951)

EF 50%
female
(n = 17579)

EF 50%
male
(n = 9227)

1.13
(1.11-1.15)
P b .0001
1.15
(1.11-1.18)
P b .0001

1.10
(1.05-1.16)
P = .0001
1.11
(1.04-1.18)
P = .0012

1.18
(1.14-1.24)
P b .0001
1.16
(1.10-1.22)
P b .0001

1.13
(1.07-1.19)
P b .0001
1.15
(1.09-1.22)
P b .0001

1.11
(1.06-1.17)
P b .0001
1.14
(1.07-1.21)
P b .0001

1.14
(1.11-1.17)
P b .0001
1.15
(1.10-1.20)
P b .0001

1.08
(1.04-1.12)
P b .0001
1.10
(1.05-1.15)
P b .0001

Model 2

Odd ratio
Unadjusted

Multivariable
adjusted

Model 1 included demographic, clinical, and hospital variables with no or b15% missing values: These variables were the following: age, SBP, heart rate, BMI, race, anemia, ischemic
history, diabetes mellitus, hyperlipidemia, hypertension, smoking, chronic obstructive pulmonary disease/asthma, peripheral vascular disease, renal failure/dialysis, depression,
valve disease, ICD-CRT-D, atrial fibrillation/flutter and cerebral vascular accident/transient ischemic attack, and hospital characteristics (region, rural location, no. of beds, academic
center and heart transplant center).
Adjusted odds ratio (95% CI), P value.
Model 2 was similar to model one but limited to only patients with complete data and included laboratory variables (sodium, hemoglobin, abnormal troponin, creatinine, and BUN).
Adjusted odds ratio (95% CI), P value.

American Heart Journal

Volume 166, Number 6

Hsich et al 1071.e3

Supplementary Table IV. Adjusted odds ratio for log BNP and hospital discharge other than home among heart failure patients stratified by
ejection fraction/sex
Model 1

Odds ratio
Unadjusted

Multivariable
adjusted

All
patients
(n = 95914)

EF b40%
female
(n = 16789)

EF b40%
male
(n = 28308)

EF 40%-49%
female
(n = 6910)

EF 40%-49%
male
(n = 6501)

EF 50%
female
(n = 13120)

EF 50%
male
(n = 24286)

1.20
(1.17-1.22)
P b .0001
1.24
(1.15-1.34)
P b .0001

1.34
(1.26-1.43)
P b .0001
1.26
(1.11-1.43)
P = .0004

1.40
(1.29-1.52)
P b .0001
1.26
(1.09-1.46)
P = .0021

1.24
(1.15-1.33)
P b .0001
1.22
(1.09-1.37)
P = .0006

1.23
(1.15-1.32)
P b .0001
1.22
(1.12-1.33)
P b .0001

1.25
(1.21-1.29)
P b .0001
1.23
(1.17-1.29)
P b .0001

1.15
(1.11-1.19)
P b .0001
1.14
(1.07-1.21)
P b .0001

All
patients
(n = 72469)

EF b40%
female
(n = 12717)

EF b40%
male
(n = 21272)

EF 40%-49%
female
(n = 5015)

EF 40%-49%
male
(n = 5217)

EF 50%
female
(n = 18601)

EF 50%
male
(n = 9647)

1.19
(1.16-1.21)
P b .0001
1.18
(1.10-1.27)
P b .0001

1.31
(1.22-1.40)
P b .0001
1.18
(1.05-1.33)
P = .0068

1.39
(1.26-1.54)
P b .0001
1.18
(1.04-1.35)
P = .0094

1.24
(1.14-1.34)
P b .0001
1.19
(1.05-1.35)
P = .0072

1.23
(1.15-1.31)
P b .0001
1.19
(1.09-1.29)
P = .0001

1.24
(1.19-1.28)
P b .0001
1.18
(1.11-1.24)
P b .0001

1.15
(1.10-1.20)
P b .0001
1.10
(1.06-1.15)
P b .0001

Model 2

Odds ratio
Unadjusted

Multivariable
adjusted

Model 1 included demographic, clinical and hospital variables with no or b15% missing values. These variables were the following: age, SBP, heart rate, BMI, race, anemia, ischemic
history, diabetes mellitus, hyperlipidemia, hypertension, smoking, chronic obstructive pulmonary disease/asthma, peripheral vascular disease, renal failure/dialysis, depression,
valve disease, ICD-CRT-D, atrial fibrillation/flutter and cerebral vascular accident/transient ischemic attack, and hospital characteristics (region, rural location, number of beds,
academic center and heart transplant center).
Adjusted odds ratio (95% CI), P value.
Model 2 was similar to model one but limited to only patients with complete data and included laboratory variables (sodium, hemoglobin, abnormal troponin, creatinine, and BUN).
Adjusted odds ratio (95% CI), P value.