You are on page 1of 7

REVIEW

Strength of Study Evidence Examined by the FDA


in Premarket Approval of Cardiovascular Devices
Sanket S. Dhruva, MD Context Medical devices are common in clinical practice and have important effects
Lisa A. Bero, PhD on morbidity and mortality, yet there has not been a systematic examination of evi-
Rita F. Redberg, MD, MSc dence used by the US Food and Drug Administration (FDA) for device approval.
Objectives To study premarket approval (PMA)the most stringent FDA review

C
ARDIOVASCULAR DEVICES ARE processof cardiovascular devices and to characterize the type and strength of evi-
increasing in number, use, dence on which it is based.
complexity, and cost.1,2 In Data Sources and Study Selection Systematic review of 123 summaries of safety
2008, at least 350 000 pace- and effectiveness data for 78 PMAs for high-risk cardiovascular devices that received
makers, 140 000 implantable cardio- PMA between January 2000 and December 2007.
verter-defibrillators,and 1 230 000stents Data Extraction Examination of the methodological characteristics considered es-
were implanted2 (Mike Weinstein, BS, sential to minimize confounding and bias, as well as the primary end points of the 123
J.P. Morgan Securities; written commu- studies supporting the PMAs.
nication). Although there has been re-
Results Thirty-three of 123 studies (27%) used to support recent FDA approval of
cent scrutiny of evidence used in the US cardiovascular devices were randomized and 17 of 123 (14%) were blinded. Fifty-
Food and Drug Administration (FDA) one of 78 PMAs (65%) were based on a single study. One hundred eleven of 213
drug approval process,3 less attention has primary end points (52%) were compared with controls and 34 of 111 controls (31%)
been paid to the approval process for were retrospective. One hundred eighty-seven of 213 primary end points (88%) were
medical devices. Medical devices are less surrogate measures and 122 of 157 (78%) had a discrepancy between the number of
likely than drugs to have demonstrated patients enrolled in the study and the number analyzed.
clinical safety before they are marketed,4 Conclusion Premarket approval of cardiovascular devices by the FDA is often based
and evidence shows that review perfor- on studies that lack adequate strength and may be prone to bias.
mance has begun to decline.5 JAMA. 2009;302(24):2679-2685 www.jama.com
Devices are stratified by increasing
risk for patients, as class I, II, and III, As clinical use and insurance cover- sist of randomized, double-blind stud-
with stringency of the approval pro- age often quickly follow FDA ap- ies with adequate controls, sufficient
cess corresponding to device risk.6 Class proval of a device,10 and as medical de- duration, and thorough follow-up on
III devices, the highest risk, are de- vices are increasingly being marketed prespecified primary end points with-
fined as usually those that support or directly to consumers after FDA ap- out bias.14
sustain human life, are of substantial proval,11,12 it is important to under- After a device receives PMA, the FDA
importance in preventing impairment stand thoroughly the study data on makes publicly available its approval or-
of human health, or which present a po- which such approval is based. Fur- der, labeling guidelines, and a sum-
tential, unreasonable risk of illness or ther, the US Supreme Courts Febru- mary of safety and effectiveness data
injury.7 The scientific and regulatory ary 2008 decision in Riegel v Medtronic (SSED). The SSED is intended to pre-
review process to evaluate the safety means that FDA approval of a device sent a reasoned, objective, and bal-
and effectiveness of class III medical de- preempts consumers from suing be-
vices is the premarket approval (PMA), cause of problems with the safety or ef- Author Affiliations: Department of Medicine (Drs
Dhruva and Redberg); Department of Clinical Phar-
the most stringent type of device mar- fectiveness of the device, making this macy, School of Pharmacy (Dr Bero); Institute for
keting application required by FDA.8 approval a vital consumer protection Health Policy Studies, School of Medicine (Dr Bero);
and Division of Cardiology (Dr Redberg), University
The PMAs are required for novel or safeguard.13 Given these issues, it is es- of California, San Francisco.
high risk devices.9 Class III devices ac- sential that the study evidence on which Corresponding Author: Rita F. Redberg, MD, MSc, Di-
vision of Cardiology, Department of Medicine, 505
count for 50 to 80 of the 8000 new device approval is based is of high qual- Parnassus Ave, Ste M-1180, San Francisco, CA 94143-
medical devices marketed each year.4 ity. Ideally, this evidence should con- 0124 (redberg@medicine.ucsf.edu).

2009 American Medical Association. All rights reserved. (Reprinted) JAMA, December 23/30, 2009Vol 302, No. 24 2679

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015


EVIDENCE IN PREMARKET APPROVAL OF CARDIOVASCULAR DEVICES

and the FDAs decision date. Except was 1.6 (0.9) (range, 1-5 studies). Of the
Figure. Flowchart of Number of Individual
Studies Included in PMAs when coding surrogate measures, 1 au- 78 PMAs, 51 (65%) were supported by a
thor (S.S.D.) classified all data, which single study. For the 123 studies, only 98
81 Approved cardiovascular were verified by at least 1 other author SSEDs(80%)reportedthenumberofpar-
PMAs between January 1, (L.A.B. or R.F.R. or both). ticipants enrolled (mean [SD], 308 [284]
2000, and December 31, 2007
participants) (TABLE 1). Both the num-
2 PMAs excluded (SSEDs unavailable)
Number of Studies berenrolledandnumberofsiteswerepro-
All127studieslistedundertheSummary vided for 80 studies (65%). The median
79 PMAs with SSEDs of Clinical Studies sections in the SSEDs number of patients enrolled per site was
(129 studies)
wereincludedintheanalysis.Pooledstud- 13 (interquartile range, 8-21 patients).
1 SSED excluded (duplicate SSED;
ies whose data were not presented sepa-
2 studies) rately were counted as 1 study, which Demographic Data
4 Instances of 2 pooled studies
counted as 1 study each occurred 4 times and encompassed 8 Of 123 studies cataloged in SSEDs,
studies (in P010041, P020040, P000007, mean age was stated in 87 (71%), en-
123 Studies included in the and P030047). Therefore, we coded 123 rollment by sex in 89 (72%), and en-
analysis (78 PMAs)
studies. We coded fields of data pertain- rollment by race in 11 (9%) (Table 1).
PMAs indicates premarket approvals; SSEDs, summa-
ing to essential characteristics of studies The mean (SD) age was 62.7 (11.4)
ries of safety and effectiveness data. aimedtoreducebiasandconfoundingand years, 66.9% of study participants were
several characteristics of primary end male, 87% white, 6% African Ameri-
points (BOX). can or black, 5% Hispanic or Latino, and
anced critique of the scientific evidence 3% another race or ethnicity.
which served as the basis of the deci- Data Analysis
sion to approve or deny the PMA.15 To For each category, data were tabu- Strength of Study Design
our knowledge, this study evidence has lated across PMAs, studies, and pri- Of 123 studies in SSEDs, 33 (27%) were
not been systematically examined. There- mary end points. These summary data randomized and 17 (14%) were blinded
fore, the type and quality of study evi- are presented as numbers (PMAs, stud- (Table 1). Some device groups had a
dence for devices were analyzed, focus- ies, or primary end points); percent- higher proportion of randomized and
ing on cardiovascular devices, because ages of the whole category to which they blinded studies. For example, of the 24
it was expected they would undergo the refer; and mean, standard deviation, and studies for cardiac stents, 13 (54%) were
most stringent approval process given range when applicable. In calculating randomized and 11 (46%) were blinded.
their far-reaching impact on morbidity the mean, standard deviation, and mini- One hundred ten studies were multi-
and mortality and their increasing use. mum number of end points, we ex- center, although 20 (18%) did not specify
cluded studies with no primary end the number of sites. Of studies stating the
METHODS points. In calculating the number of number of sites, the mean (SD) was 23
Data Acquisition PMAs not reporting 1 or more US cen- (17) sites (range, 1-80 sites).
On October 15, 2008, a search of the ters, we included studies that did not Follow-up time varied by type of de-
PMA database was performed16 using the state their location. Median follow-up vice; the longest median follow-up time
parameters of Advisory Committee: Car- times were calculated based on subcat- for primary end point analysis was for
diovascular and Supplement Type: Origi- egories of cardiovascular devices. intracardiac devices and endovascular
nals Only. All PMAs with a date re- grafts, both at 365 days, and the short-
ceived between January 1, 2000, and RESULTS est was for hemostasis devices at 1 day
December 31, 2007, were included and Eighty-onecardiovascularPMAswereap- (Table 1).
their SSEDs were downloaded (FIGURE). proved between January 1, 2000, and De- For SSEDs stating the number of pa-
Data were abstracted from each SSEDs cember 31, 2007. Two SSEDs (P040016, tients enrolled and the number ana-
Summary of Clinical Studies section P020035) were not available on the FDA lyzed for each study, there was a dis-
and from the adverse events section if Web site. Two PMAs had identical SSEDs crepancy for mean age in 37 of 74
it described clinical study information (P030039 and P010022) and those data studies (50%), for number by sex in 37
presented in the Summary of Clinical were included only once. The remaining of 78 studies (47%), and for number by
Studies. We recorded each PMAs 78PMAsincludedinouranalysisincluded race in 5 of 11 studies (45%).
number, the devices trade name as it 123 studies (eTable, available at http:
appears in the SSED, the applicants //www.jama.com). With the exception of Primary End Point Characteristics
name as it appears in the SSED, the ap- closure devices, all devices were either Of 123 studies, 17 (14%) did not state
plicants name as it appears on the FDA implanted or invasive during their use. a primary end point. There were a total
Web site with a link to the SSED, the The mean (SD) number of studies of 213 primary end points and a mean
date the PMA was received by the FDA, stated in SSEDs supporting each PMA (SD) of 2.0 (1.5) end points per study
2680 JAMA, December 23/30, 2009Vol 302, No. 24 (Reprinted) 2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015


EVIDENCE IN PREMARKET APPROVAL OF CARDIOVASCULAR DEVICES

(range, 1-10 end points per study). Of 213 primary end points, most dred thirteen discrepancies (93%) were
Thirty of these end points (14%) were (187, or 88%) were surrogate end that more patients were enrolled than
designated, all in cases where there was points. Examples of surrogate end analyzed; for these primary end points,
no explicitly stated primary end point points include target lesion revascular- a median of 50 patients was enrolled but
for a study. ization for a coronary stent, primary not analyzed (range, 1-604 patients).
Of the 213 primary end points, 111 patency for an endoprosthesis, and lead These 113 primary end point discrep-
(52%) were compared with controls and implant success for an electrophysiol- ancies totaled 10 351 patient exclu-
of these, 34 (31%) were retrospective ogy device. sions, 27% of the total enrolled. In 9 of
controls. Studies without controls were In the SSEDs, there were 157 pri- 122 primary end points (7%), there was
compared with objective performance mary end points for which both the a greater number analyzed than en-
criteria, which specified safety and/or ef- number enrolled and analyzed were rolled. All of these were due to retro-
ficacy targets for the device. One hun- stated. Of these, 122 (78%) had a dis- spective controls, meaning that pa-
dred nineteen primary end points (56%) crepancy between the number en- tients from a previous study were
were composites (TABLE 2). rolled and those analyzed. One hun- included in the PMA analysis, leading
to more patients being analyzed than
enrolled. For these, the median dis-
Table 1. Characteristics of Studies crepancy was 238 excess analyzed pa-
Studies tients (range, 50-848 patients).
for Which
Data
Available, Interpretation of Primary
No. (%)
Value a Range (N = 123) End Point Results
Demographic Data Of 213 primary end points in the SSEDs,
Age, mean, y 62.7 0.1-99 87 (71)
the results of 32 (15%) were noninter-
Age, SD, y b 11.4 55 (45)
pretable (Table 2). The most common
Male sex, % 66.9 28-92 89 (72)
Race, mean, %
reason was that no target goal for de-
White 87 77-94 11 (9) vice performance was stated in 25 end
African American or black 6 2-11 8 (7) points (78%), and in one instance the re-
Hispanic or Latino 5 2-6 8 (7) sults were not stated. Forty primary end
Other c 3 0-6 9 (7) points (19%) had training, lead-in, or
Studies with total enrolled patients stated, No. (%) 98 (80) roll-in patients excluded from analyses
Enrolled patients, mean (SD) 308 (284) 23-1548 and 21 (10%) had a post hoc analysis of
Enrolled patients/site, mean (SD) 40 (172) 2-1492 80 (65)
the primary end point (Table 2).
Enrolled patients/site, interquartile range 8-21
In some instances, end points were
Study Strength
Randomized studies 33/123 (27) interpreted to meet their targets when
Blinded studies 17/123 (14) they may have met only a part of them.
Double-blind 11/123 (9) In 1 PMA, for example, 107 of 226 pa-
Single-blind 6/123 (5) tients (47.3%) had chronic success in
Studies stating multicenter or single center 114/123 (93) the effectiveness analysis cohort (de-
Multicenter 110/114 (96) fined as no recurrence of clinically rel-
Multicenter but did not state No. sites 20/110 (18)
evant monomorphic ventricular tachy-
No. of sites, mean (SD) 23/123 (16.9) 1-80 94 (76)
cardia that were targeted at ablation17),
Studies with location stated 83/123 (67)
All US sites 43/83 (52)
with a 95% lower confidence bound of
Some US sites 18/83 (22)
41.7%. The target end point on that
No US sites 22/83 (27) same table in the SSED,17 however, is
Median follow-up time for primary end points by device type, d shown to be 50% chronic success with
Cardiac stents 180 1-284 a 95% lower confidence bound of 40%.
Noncardiac stents 270 1-720 The SSED explains this discrepancy as
Bridge to transplant 180 30-180 follows: The results demonstrate that
Intracardiac devices 365 180-2920
the percentage of subjects achieving
Hemostasis devices 1 1-30
chronic success (47.3%, 95% lower con-
Electrophysiology devices 90 0.02-609
Endovascular grafts 365 30-365
fidence bound of 41.7%) met the pro-
Miscellaneous/other 180 0.01-365 tocol end point for chronic success. This
a Values are number/total number (percentage) unless otherwise stated. is due to the fact that although the point
b Standard deviation is presented separately for age because it was not available for all studies for which age was available.
c Asian, American Indian or Alaskan Native, Pacific Islander or Hawaiian Native, nonwhite, other, and unknown. estimate for chronic success was lower
than the protocol end point, the 95%
2682 JAMA, December 23/30, 2009Vol 302, No. 24 (Reprinted) 2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015


EVIDENCE IN PREMARKET APPROVAL OF CARDIOVASCULAR DEVICES

(range, 1-10 end points per study). Of 213 primary end points, most dred thirteen discrepancies (93%) were
Thirty of these end points (14%) were (187, or 88%) were surrogate end that more patients were enrolled than
designated, all in cases where there was points. Examples of surrogate end analyzed; for these primary end points,
no explicitly stated primary end point points include target lesion revascular- a median of 50 patients was enrolled but
for a study. ization for a coronary stent, primary not analyzed (range, 1-604 patients).
Of the 213 primary end points, 111 patency for an endoprosthesis, and lead These 113 primary end point discrep-
(52%) were compared with controls and implant success for an electrophysiol- ancies totaled 10 351 patient exclu-
of these, 34 (31%) were retrospective ogy device. sions, 27% of the total enrolled. In 9 of
controls. Studies without controls were In the SSEDs, there were 157 pri- 122 primary end points (7%), there was
compared with objective performance mary end points for which both the a greater number analyzed than en-
criteria, which specified safety and/or ef- number enrolled and analyzed were rolled. All of these were due to retro-
ficacy targets for the device. One hun- stated. Of these, 122 (78%) had a dis- spective controls, meaning that pa-
dred nineteen primary end points (56%) crepancy between the number en- tients from a previous study were
were composites (TABLE 2). rolled and those analyzed. One hun- included in the PMA analysis, leading
to more patients being analyzed than
enrolled. For these, the median dis-
Table 1. Characteristics of Studies crepancy was 238 excess analyzed pa-
Studies tients (range, 50-848 patients).
for Which
Data
Available, Interpretation of Primary
No. (%)
Value a Range (N = 123) End Point Results
Demographic Data Of 213 primary end points in the SSEDs,
Age, mean, y 62.7 0.1-99 87 (71)
the results of 32 (15%) were noninter-
Age, SD, y b 11.4 55 (45)
pretable (Table 2). The most common
Male sex, % 66.9 28-92 89 (72)
Race, mean, %
reason was that no target goal for de-
White 87 77-94 11 (9) vice performance was stated in 25 end
African American or black 6 2-11 8 (7) points (78%), and in one instance the re-
Hispanic or Latino 5 2-6 8 (7) sults were not stated. Forty primary end
Other c 3 0-6 9 (7) points (19%) had training, lead-in, or
Studies with total enrolled patients stated, No. (%) 98 (80) roll-in patients excluded from analyses
Enrolled patients, mean (SD) 308 (284) 23-1548 and 21 (10%) had a post hoc analysis of
Enrolled patients/site, mean (SD) 40 (172) 2-1492 80 (65)
the primary end point (Table 2).
Enrolled patients/site, interquartile range 8-21
In some instances, end points were
Study Strength
Randomized studies 33/123 (27) interpreted to meet their targets when
Blinded studies 17/123 (14) they may have met only a part of them.
Double-blind 11/123 (9) In 1 PMA, for example, 107 of 226 pa-
Single-blind 6/123 (5) tients (47.3%) had chronic success in
Studies stating multicenter or single center 114/123 (93) the effectiveness analysis cohort (de-
Multicenter 110/114 (96) fined as no recurrence of clinically rel-
Multicenter but did not state No. sites 20/110 (18)
evant monomorphic ventricular tachy-
No. of sites, mean (SD) 23/123 (16.9) 1-80 94 (76)
cardia that were targeted at ablation17),
Studies with location stated 83/123 (67)
All US sites 43/83 (52)
with a 95% lower confidence bound of
Some US sites 18/83 (22)
41.7%. The target end point on that
No US sites 22/83 (27) same table in the SSED,17 however, is
Median follow-up time for primary end points by device type, d shown to be 50% chronic success with
Cardiac stents 180 1-284 a 95% lower confidence bound of 40%.
Noncardiac stents 270 1-720 The SSED explains this discrepancy as
Bridge to transplant 180 30-180 follows: The results demonstrate that
Intracardiac devices 365 180-2920
the percentage of subjects achieving
Hemostasis devices 1 1-30
chronic success (47.3%, 95% lower con-
Electrophysiology devices 90 0.02-609
Endovascular grafts 365 30-365
fidence bound of 41.7%) met the pro-
Miscellaneous/other 180 0.01-365 tocol end point for chronic success. This
a Values are number/total number (percentage) unless otherwise stated. is due to the fact that although the point
b Standard deviation is presented separately for age because it was not available for all studies for which age was available.
c Asian, American Indian or Alaskan Native, Pacific Islander or Hawaiian Native, nonwhite, other, and unknown. estimate for chronic success was lower
than the protocol end point, the 95%
2682 JAMA, December 23/30, 2009Vol 302, No. 24 (Reprinted) 2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015


EVIDENCE IN PREMARKET APPROVAL OF CARDIOVASCULAR DEVICES

lower confidence bound of the esti-


Table 2. Characteristics of Primary End Points and Data Analyses
mate was higher than the protocol end
End Points,
point.17 Characteristic Subcharacteristic No. (%) a
PEPs with control comparisons 111/213 (52)
PMA Analyses Of control groups Retrospective 34/111 (31)
Of 78 PMAs, 24 (31%) had at least 1 Composite PEPs 119/213 (56)
randomized study and 10 (13%) at least Composite PEPs reporting adverse 50/119 (42)
events/complications
1 blinded study (TABLE 3). Four PMAs
Composite PEPs not reporting adverse 69/119 (58)
(5%) were supported by at least 2 events/complications
blinded randomized studies. More than Of composite PEPs as adverse Results for all components 50/50 (100)
one-third, 28 PMAs, did not report a events/complications
study with at least 1 US center (Table 3). Of composite PEPs not due to adverse Results for all components 36/69 (52)
events/complications
Seven PMAs had no primary end point
Results for some components 11/69 (16)
for any study. Of the 71 PMAs with at No results for any 22/69 (32)
least 1 primary end point, 16 (23%) components
used at least 1 retrospective control Surrogate PEPs 187/213 (88)
group and 16 (23%) based device ap- PEPs excluding training/lead-in/roll-in patients 40/213 (19)
proval at least in part on a post hoc from analysis
PEPs with post hoc analysis 21/213 (10)
analysis.
PEPs with discrepancy between No. enrolled 122/157 (78)
For example, 1 device was ap- and No. analyzed
proved wholly on a post hoc analysis No. enrolled No. analyzed 113/122 (93)
for a single subgroup studied in the No. analyzed No. enrolled 9/122 (7)
single preapproval study discussed in PEPs not interpretable 32/213 (15)
the SSED. However, because this sub- No target goal 25/32 (78)
group did not meet the prespecified tar- No statistical analysis 4 (13)
get performance standard, the SSED Insufficient data 2 (6)
stated the following: CDRH [Center for No results 1 (3)
Devices and Radiological Health] de- Abbreviation: PEP, primary end point.
a Total primary end points/outcomes/objectives=213 unless otherwise indicated.
termined that the OPCs [objective per-
formance criteria] designated for the
overall study population should not be vice. The vast majority of end points are
Table 3. Characteristics of Premarket
used to evaluate the safety and effec- surrogates, which may not be reliable Approvals
tiveness of the device in the AVNRT predictors of actual patient benefit.19 Al- PMAs,
[Atrioventricular Nodal Reentry Tachy- though surrogate outcomes are attrac- No./Total
cardia] subgroup. As a result, the de- tive because they decrease the time and Characteristic No. (%)
1 Randomized study 24/78 (31)
vice system for the proposed indica- costs required to do a study, they must
1 Blinded study 10/78 (13)
tion was evaluated on its merits, as be linked to a clinically meaningful end
No study reported with 1 US center 28/78 (36)
CDRH qualitatively considered the de- point to be valid. 1 Primary end point 71/78 (91)
vice risk-benefit profile.18 Composite outcomes are also com- 1 Post hoc analysis 16/71 (23)
mon, and in cardiovascular trials they 1 End point using retrospective 16/71 (23)
COMMENT have been shown to comprise indi- controls
The evidence presented in the SSEDs vidual end points that often vary in Abbreviation: PMAs, premarket approvals.
for FDA-approved cardiovascular de- clinical significance and do not con-
vice PMAs from 2000 through 2007 tribute equally to the composite mea-
showed that the majority of studies are sure.20 The frequent discrepancies be- because this preferentially excludes pa-
not blinded or randomized. Blinding for tween the number of enrolled patients tients in whom the device may not be
some devices such as left ventricular as- and the number analyzed for primary associated with a favorable outcome. In
sist devices is not possible,1 but most end points, despite the short fol- some instances where bias may not ini-
cardiovascular device PMAs do not have low-up times, may introduce bias be- tially be present, devices are approved
even 1 blinded or 1 randomized study. cause patients with less favorable out- using a post hoc analysis of data, which
Controls are used in a little more than comes may be lost to follow-up and could introduce bias favoring the de-
half of studies, and the common use of safety concerns may underlie this miss- vice. The PMA is the most rigorous de-
retrospectively selected controls can in- ing data. The common practice of ex- vice approval process, and strict stan-
troduce bias by allowing for the selec- cluding data from the training/roll-in/ dards for cardiovascular devices are
tion of control groups that favor the de- lead-in period also introduces bias expected given their far-reaching ef-
2009 American Medical Association. All rights reserved. (Reprinted) JAMA, December 23/30, 2009Vol 302, No. 24 2683

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015


EVIDENCE IN PREMARKET APPROVAL OF CARDIOVASCULAR DEVICES

fects, permanent nature, and use in benefit is sufficient before FDA ap- tionalists often are reluctant to ran-
critically ill patients. proval and marketing. However, the bar domize patients to a medical control
Study populations should be repre- for evidence of benefit should be higher group.32
sentative of the patient populations in for devices because they are im- The FDA approval process is an im-
which these devices will be used. The planted and cannot simply be discon- portant determinant of health care spend-
FDA guidelines stating that data from tinued, as drugs can. In addition, al- ing: when the FDA approves devices
outside the United States be appli- though devices can be lifesaving, they more quickly, spending on new devices
cable to the US population and US also have great potential for risk and ad- increases.6 Given that health care spend-
medical practice21 are increasingly im- verse events. For example, after 268 000 ing in the United States was 16.2% of the
portant as more trials are conducted in- implantations in 3 years after ap- gross domestic product in 200733 and is
ternationally. In more than one-third proval, the Medtronic Sprint Fidelis im- projected to increase to 31% by 2031, and
of PMAs, however, we were not able to plantable cardioverter-defibrillator lead that rapid increase in health care spend-
ascertain that even 1 study had been was found to have an increased risk of ing is attributed principally to new tech-
conducted in the United States. This re- fracture.24 Further, despite the risks of nologies,34 rigorous outcome evalua-
sults in uncertain generalizability of ap- using this device, lead removal is quite tion prior to approval is critical to
proved medical devices to the US popu- dangerous for patients.25 increasing value for US health care ex-
lation.22 The importance of FDA device ap- penditures. Fewer than half of medical
There are several possible reasons the proval is magnified as it preempts con- decisions are supported by firm evi-
criteria on which FDA device ap- sumer lawsuits on device safety. Drug dence of effectiveness, and many incen-
proval is based appear to be less rigor- approval by the FDA no longer guar- tives in the US health care system en-
ous than those for drug approvals. First, antees preemption26; current legisla- courage use of expensive treatments and
device approvals are a more recent ac- tion in Congress seeks to overturn this procedures unrelated to evidence of pa-
tivity for the FDA, having begun in 1976 inconsistency and allow consumers tient benefit.35 There is a new focus on
with the FDA Device Amendment,6 so lawsuits to become a part of the regu- comparative effectiveness research,
the agency has less experience with de- latory framework for devices.27 Post- which has been allocated $1.1 billion by
vices than it does with drugs. Further, market surveillance is also weak pro- the Obama administration.36 The suc-
the last decade has brought a signifi- tection because, although postmarket cess of comparative effectiveness de-
cant increase in the number and com- studies are sometimes required, manu- pends on the use of its principles in FDA
plexity of devices. In addition, on the facturers are not actively required to reviews.37 Cost containment would likely
FDA approval continuum, devices, seek out device malfunctions, so device- occur if rigorous clinical effectiveness re-
which are almost always implanted, are related adverse events are substan- views are used for new drugs and tech-
between drugs, which have relatively tially underreported.24,28,29 In addi- nologies and spending concentrated on
strict criteria for approval, and new sur- tion, although FDA approval may devices shown to benefit patients.38 Car-
gical operations, which do not require address only a specific, narrow popu- diovascular and peripheral vascular dis-
FDA approval. lation and indication, physicians may ease rank second among the primary re-
The importance of the seal of FDA use devices for unapproved indica- search areas designated by the Institute
approval cannot be overstated. Many tions.6 For example, Medicare data of Medicine.39 This study suggests that
manufacturers immediately encourage show that 69% of current drug- the FDA device approval process would
widespread use of their devices based on eluting stent use is off-label.30 benefit from such rigorous research,
FDA approval through direct-to- All of these factors make it critical to using meaningful clinical outcomes and
consumer advertising,11,12 detailing to public health that FDA device ap- valid, active (not historical) controls in
physicians, and continuing medical edu- proval require sufficient high-quality randomized, blinded studies con-
cation venues. An oft-repeated asser- evidence to support device safety and ducted in populations that reflect the US
tion by the sponsor is that FDA ap- effectiveness, as determined in the PMA population in which they are intended
proval is sufficient grounds for insurance process. Yet, 65% of PMAs were based for use.
coverage and rapid dissemination of new on a single study, which suggests that A limitation of this study may be that
devices. This rapid diffusion encour- there may not be adequate evidence the data source is primarily publicly
ages use beyond evidence or overutiliza- prior to FDA approval. Another op- available SSEDs. However, it is pos-
tion of the health care system.23 The find- tion is to rely more on independent, sys- sible that at least some of the data miss-
ings in this study raise questions about tematic evidence-based assessments, al- ing from the SSEDs were also omitted
the quality of data on which some car- though these will be hindered by the from the proprietary reports to the FDA.
diovascular device approvals are based. lack of rigorous clinical trial data and The SSEDs should contain all data pre-
There is a balance between getting disincentives after approval for such sented to the FDA. If sufficient data are
new drugs and devices to market studies.31 For example, after FDA ap- not presented or are inconsistent, the
quickly and ensuring the evidence of proval of a medical device, interven- SSED should be checked for thorough-
2684 JAMA, December 23/30, 2009Vol 302, No. 24 (Reprinted) 2009 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015


EVIDENCE IN PREMARKET APPROVAL OF CARDIOVASCULAR DEVICES

ness prior to making a decision and User Fee Act has been rapid approval Drafting of the manuscript: Dhruva, Redberg.
Critical revision of the manuscript for important in-
posting on the FDA Web site. Given of new drugs. This study suggests that tellectual content: Dhruva, Bero, Redberg.
that the specific stated purpose of SSEDs the emphasis for the FDA in 2009 and Study supervision: Redberg.
is to present the basis of the FDAs de- beyond must be approvals based on re- Financial Disclosures: Dr Redberg reported being a
member of the FDA Circulatory System Devices
cision,15 the SSEDs should be a thor- search that meets rigorous scientific Panel and a member of the California Technology
ough and accurate compilation of the standards for evidence of benefit and Assessment Forum. No other disclosures were
reported.
FDAs critique of evidence. Further, lack of harm to patients. To uphold the Additional Information: The eTable is available at http:
SSEDs are the only FDA-reviewed evi- FDAs mission of ensuring safe and ef- //www.jama.com.
dence available for clinicians, and they fective medical devices, it is essential Additional Contributions: Mark Pletcher, MD,
Department of Epidemiology and Biostatistics,
form the sole basis of data that can be that high-quality studies and data are UCSF, provided statistical assistance, which was
used for systematic reviews and guide- available. supported by NIH/NCRR grant UL1 RR024131 to
the UCSF Clinical and Translational Science Insti-
line development. This study rein- Author Contributions: Dr Redberg had full access to tute. Jeffrey Tice, MD, and Steven Schroeder, MD,
forces the need for improved access to all of the data in the study and takes responsibility for Department of Medicine, UCSF, provided help-
the integrity of the data and the accuracy of the data ful comments, and Deborah Airo, BA, Department
complete FDA reviews40 for both phar- analysis. of Epidemiology and Biostatistics, UCSF, pro-
maceutical and device data. Study concept and design: Dhruva, Bero, Redberg. vided editorial assistance. The UCSF Pathway to
Acquisition of data: Dhruva, Redberg. Discovery in Health and Society also provided assis-
The emphasis at the FDA in the last Analysis and interpretation of data: Dhruva, Bero, tance. None received compensation for their contri-
17 years since the Prescription Drug Redberg. bution.

REFERENCES
1. Muni NI, Zuckerman BD. The process of regula- 13. Gostin LO. The deregulatory effects of preempt- Administration. http://www.fda.gov/ForConsumers
tory review for new cardiovascular devices. In: Ant- ing tort litigation: FDA regulation of medical devices. /ConsumerUpdates/ucm103022.htm. Accessed
man EM, ed. Cardiovascular Therapeutics: A Com- JAMA. 2008;299(19):2313-2316. November 22, 2009.
panion to Braunwalds Heart Disease, 3rd Edition. 14. Assessing risk of bias in included studies. Higgins 26. Wyeth v Levine, 555 US __ (2009).
Philadelphia, PA: Elsevier; 2007. JPT, Altman DG, eds. In: Higgins JPT, Green S, eds. 27. Curfman GD, Morrissey S, Drazen JM. The Medi-
2. Zhan C, Baine WB, Sedrakyan A, Steiner C. Car- Cochrane Handbook for Systematic Reviews of In- cal Device Safety Act of 2009. N Engl J Med. 2009;
diac device implantation in the United States from 1997 terventions. http://www.cochrane-handbook.org. Ac- 360(15):1550-1551.
through 2004: a population-based analysis. J Gen In- cessed July 22, 2009. 28. Meier B. Maker of heart device kept flaw from
tern Med. 2007;23(suppl 1):13-19. 15. PMA application contents: summary of safety and doctors. The New York Times. May 24, 2005:1.
3. Rising K, Bacchetti P, Bero L. Reporting bias in drug effectiveness data. US Food and Drug Administration. 29. Maisel WH. Safety issues involving medical de-
trials submitted to the Food and Drug Administra- http://www.fda.gov/MedicalDevices/Device vices: implications of recent implantable cardioverter-
tion: review of publication and presentation. PLoS Med. RegulationandGuidance/HowtoMarketYourDevice defibrillator malfunctions. JAMA. 2005;294(8):
2008;5(11):e217. /PremarketSubmissions/PremarketApproval 955-958.
4. Feigal DW, Gardner SN, McClellan M. Ensuring safe PMA/ucm050289.htm#ssed. Accessed July 22, 30. Douglas PS, Brennan JM, Anstrom KJ, et al. Clini-
and effective medical devices. N Engl J Med. 2003; 2009. cal effectiveness of coronary stents in elderly per-
348(3):191-192. 16. Premarket approval: Center for Devices and Ra- sons: results from 262,700 Medicare patients in the
5. Medical Device User Fee and Modernization Act diological Health SuperSearch. US Food and Drug American College of CardiologyNational Cardiovas-
of 2002 frequently asked questions. US Food and Drug Administration. http://www.accessdata.fda.gov/scripts cular Data Registry. J Am Coll Cardiol. 2009;53
Administration. http://www.fda.gov/MedicalDevices /cdrh/cfdocs/cfPMA/pma.cfm. Accessed December (18):1629-1641.
/DeviceRegulationandGuidance/Overview 1, 2009. 31. Feldman MD, Petersen AJ, Karliner LS, Tice JA.
/MedicalDeviceUserFeeandModernizationAct 17. Summary of safety and effectiveness data: PMA Who is responsible for evaluating the safety and ef-
MDUFMA/ucm109208.htm. Accessed July 22, No. 040036, NaviStar ThermoCool deflectable diag- fectiveness of medical devices? the role of indepen-
2009. nostic/ablation catheter [August 2006]. http://www dent technology assessment. J Gen Intern Med. 2008;
6. Maisel WH. Medical device regulation: an intro- .accessdata.fda.gov/cdrh_docs/pdf4/P040036b 23(suppl 1):57-63.
duction for the practicing physician. Ann Intern Med. .pdf. Accessed July 22, 2009. 32. Furlan AJ, Fisher M. Devices, drugs, and the Food
2004;140(4):296-302. 18. Summary of safety and effectiveness data: PMA and Drug Administration: increasing implications for
7. Device classes: general and special controls. US Food No. 020045, cardiac cryoablation catheter and con- ischemic stroke. Stroke. 2005;36(2):398-399.
and Drug Administration. http://www.fda.gov sole system. http://www.accessdata.fda.gov 33. National health expenditures: 2007 highlights.
/MedicalDevices/DeviceRegulationandGuidance /cdrh_docs/pdf2/P020045b.pdf. Accessed Novem- http://www.cms.hhs.gov/NationalHealthExpendData
/Overview/GeneralandSpecialControls/default.htm. ber 8, 2009. /downloads/highlights.pdf. Accessed July 22,
Accessed July 22, 2009. 19. Fleming TR, DeMets DL. Surrogate end points in 2009.
8. Premarket Approval device advice: overview. US clinical trials: are we being misled? Ann Intern Med. 34. Technological change and the growth of health
Food and Drug Administration. http://www.fda.gov 1996;125(7):605-613. care spending: January 2008. Congressional Budget
/MedicalDevices/DeviceRegulationandGuidance 20. Lim E, Brown A, Helmy A, Mussa S, Altman DG. Office. http://www.cbo.gov/doc.cfm?index=8947.
/HowtoMarketYourDevice/PremarketSubmissions Composite outcomes in cardiovascular research: a sur- Accessed December 2, 2009.
/PremarketApprovalPMA/default.htm. Accessed July vey of randomized trials. Ann Intern Med. 2008; 35. Orszag PR, Ellis P. Addressing rising health care
22, 2009. 149(9):612-617. costs: a view from the Congressional Budget Office.
9. Testimony on FDAs medical device program by 21. Food and drugs: premarket approval of medical N Engl J Med. 2007;357(19):1885-1887.
Michael Friedman, MD, lead deputy commissioner, US devices: research conducted outside the United States. 36. Comparative effectiveness research in the USA.
Food and Drug Administration, before the House Com- 21 CFR 814.15. Lancet. 2009;373(9665):694.
mittee on Commerce, Subcommittee on Health and the 22. Glickman SW, McHutchinson JG, Peterson ED, 37. Alexander GC, Stafford RS. Does comparative ef-
Environment [April 30, 1997]. http://www.dhhs.gov et al. Ethical and scientific implications of the global- fectiveness have a comparative edge? JAMA. 2009;
/asl/testify/t970430a.html. Accessed July 22, 2009. ization of clinical research. N Engl J Med. 2009; 301(23):2488-2490.
10. Yock CA, Yock PG. The drug-eluting stent infor- 360(8):816-823. 38. Mongan JJ, Ferris TG, Lee TH. Options for slow-
mation gap. Am Heart Hosp J. 2004;2(1):21-25. 23. Emanuel EJ, Fuchs VR. The perfect storm of ing the growth of health care costs. N Engl J Med. 2008;
11. Boden WE, Diamond GA. DTCA for PTCA: cross- overutilization. JAMA. 2008;299(23):2789-2791. 358(14):1509-1514.
ing the line in consumer health education? N Engl 24. Maisel WH. Semper fidelis: consumer protection 39. Iglehart JK. Prioritizing comparative-effectiveness
J Med. 2008;358(21):2197-2200. for patients with implanted medical devices. N Engl research: IOM recommendations. N Engl J Med. 2009;
12. Mitka M. Direct-to-consumer advertising of medi- J Med. 2008;358(10):985-987. 361(4):325-328.
cal devices under scrutiny. JAMA. 2008;300(17): 25. Medtronic recalls Sprint Fidelis cardiac leads: ques- 40. OConnor AB. The need for improved access to
1985-1986. tions and answers for consumers. US Food and Drug FDA reviews. JAMA. 2009;302(2):191-193.

2009 American Medical Association. All rights reserved. (Reprinted) JAMA, December 23/30, 2009Vol 302, No. 24 2685

Downloaded From: http://jama.jamanetwork.com/ by a California State University - Fresno User on 07/23/2015

You might also like