Graham DJ, Drinkard CR, Shatin D, Tsong Y, Burgess MJ. Liver Enzyme Monitoring in Patients Treated With Troglitazone. JAMA. 2001;286(7):831-833. doi:10.1001/jama.286.7.831
Author Affiliations: Office of Postmarketing Drug Risk Assessment (Dr Graham) and Office of Biostatistics (Dr Tsong), Center for Drug Evaluation and Research, US Food and Drug Administration, Rockville, Md; and Center for Health Care Policy and Evaluation, UnitedHealth Group, Minneapolis, Minn (Drs Drinkard and Shatin and Ms Burgess).
Context Soon after initial marketing in March 1997,
troglitazone, the first thiazolidinedione antidiabetic agent, was found
to cause life-threatening acute liver failure. The drug was removed
from the market in March 2000.
Objective To evaluate the effect of US Food and Drug
Administration (FDA) risk management efforts, including repeated
labeling changes and "Dear Healthcare Professional" letters, on
periodic liver enzyme monitoring of patients taking troglitazone.
Design, Setting, and Participants Claims data from a large,
multistate managed care organization were used to establish 4 cohorts
of patients (N = 7603) with at least 90 days of health
plan enrollment before first troglitazone prescription during 4
consecutive periods spanning April 1997 to September 1999 and
representing 4 progressively stringent liver monitoring
Main Outcome Measures Percentage of eligible troglitazone users in
each cohort with baseline, monthly (for up to 6 months of continuous
use), and complete (baseline and monthly) enzyme monitoring, based on
computerized records of laboratory claims.
Results Baseline testing increased from 15% before any FDA
monitoring recommendations (cohort 1) to 44.6% following 4 separate
FDA interventions (cohort 4; P<.001). In cohort 4,
33.4% of users had follow-up testing after 1 month of therapy, falling
to 13% after 5 months of continuous use. In all cohorts, less than 5%
received all recommended liver enzyme tests by the third month of
Conclusions The FDA risk management efforts did not achieve
meaningful or sustained improvement in liver enzyme testing. Evaluation
of the impact of regulatory actions is needed before such actions can
be regarded as effective or sufficient.
Troglitazone is a thiazolidinedione antidiabetic agent that decreases
insulin resistance. It was approved by the US Food and Drug Administration
(FDA) in January 1997 for treatment of type 2 diabetes mellitus. During premarketing
clinical trials, serum liver transaminase elevations exceeding 3 times the
upper limit of normal were observed in 1.9% of troglitazone-treated patients
vs 0.6% in comparators.1 Of 2510 patients
treated with troglitazone in these trials, 2 were hospitalized with drug-associated
hepatitis and 2 developed jaundice.
From March 1997, when marketing began, through the end of October 1997,
there were no recommendations in product labeling that liver enzymes should
be monitored. Soon thereafter, cases of acute liver failure (ALF) began to
be reported among patients taking troglitazone. In response to a steadily
increasing number of reported cases, 4 separate "Dear Healthcare Professional"
letters were sent by the manufacturer to US physicians nationwide, each successively
recommending an increase in liver enzyme monitoring.2- 5
Despite these warnings, as well as a widely publicized FDA advisory meeting
in March 1999, cases of ALF continued to be reported.6- 13
In mid-1999, 2 other thiazolidinediones were approved. By March 2000, when
troglitazone was removed from the market, 94 cases of liver failure had been
The issuance of successive warning letters and associated changes in
labeling for liver enzyme monitoring provided an opportunity to study their
impact on physician and patient behavior and to assess their effectiveness
as a risk management tool.
UnitedHealth Group is a national health care company that includes 43
health plans across the United States. A historical cohort study of troglitazone
recipients was performed using longitudinal claims data from 12 UnitedHealth
Group–affiliated health plans covering about 3 million persons from
10 different states. Data consisted of separate files (pharmacy, facility,
physician, and enrollment) linked by unique encrypted identifiers, protecting
Members with at least 1 troglitazone prescription between April 1, 1997,
and September 24, 1999, and with at least 90 days of continuous enrollment
before their first (index) prescription were selected for study (n = 9136).
The prior enrollment requirement ensured that the index prescription was the
patient's first prescription. Enrolled patients were assigned to time-specific
cohorts based on the date of their index prescription and were followed for
up to 6 months while taking the drug. Four study cohorts were established,
correlating with specific FDA regulatory actions and liver enzyme monitoring
recommendations (Table 1). Patients
whose index dates fell outside the inclusion dates for cohorts 1 through 4
were excluded (n = 1533).
Longitudinal claims were screened for testing of alanine aminotransferase
or aspartate aminotrasferase, either specifically or as part of a multitest
panel. Baseline liver enzyme monitoring was defined as a laboratory claim
within 30 days before to 7 days after the index troglitazone prescription.
Monthly follow-up liver enzyme monitoring was defined as a claim from 14 days
before to 14 days after the 30-day calendar "anniversary" date following the
index prescription for up to 6 months of use. During the study period, none
of the 12 health plans limited the number of liver enzyme tests that could
Liver enzyme monitoring completeness at baseline and after each month
of troglitazone use was calculated as the percentage of eligible patients
with a laboratory claim at each relevant time point. Statistical analysis
of differences in degree of enzyme testing was performed using the χ2 test with adjustment for multiple comparisons according to Hochberg.14,15 Statistical analyses were
performed using SAS version 6.12 (SAS Institute Inc, Cary, NC).
There were 7603 first-time troglitazone recipients enrolled in the 4
study cohorts. The proportion of patients undergoing baseline liver enzyme
testing rose from 15% in cohort 1 to 44.6% in cohort 4 (χ23 = 429; P<.001). Monthly follow-up enzyme
monitoring at 1 month of therapy increased from 3.8% for cohort 1 to 33.4%
for cohort 4 (Figure 1; χ23 = 325; P<.001). This level
of monitoring was not maintained in cohort 4, falling to 13% by 5 months.The
largest increase in monthly testing occurred between cohorts 1 and 2, with
successively smaller increases between subsequent cohorts. For the first 5
months of treatment, monthly testing was generally 3- to 4-fold higher in
cohort 2 than cohort 1 (P<.001 at each month).
There was no difference between these cohorts in the sixth month. Monthly
testing in cohort 3 was 1.7-fold higher than cohort 2 in the first month (P<.001). With each successive month, the differences
in monitoring completeness decreased and were gone by month 5. There was no
statistical difference in monthly monitoring between cohorts 3 and 4 in any
month. Although one third or fewer of patients underwent follow-up testing
in any month, the same patients were not typically tested each month. For
example, among patients in cohort 4 treated with troglitazone for 3 months,
only 10% had all 3 follow-up tests performed, while 60% had at least 1.
Complete adherence to monitoring requirements occurred at low levels
(Table 2). In cohort 4, only 18.4%
of patients had a liver enzyme test at both baseline and month 1. By 3 months
of use, less than 5% of patients in any cohort received the full complement
of recommended testing.
Liver transaminase monitoring was infrequently and irregularly performed
despite repeated FDA regulatory interventions, including labeling changes,
"Dear Healthcare Professional" letters to physicians, and heightened national
publicity and awareness of liver failure risk associated with troglitazone.2- 13
Although baseline testing increased from 15% to 45%, more than half of the
patients who started taking troglitazone during the last period did so without
a baseline test. Monthly follow-up testing also increased somewhat between
the baseline and last period studied, but was modest and not sustained. Differences
in testing rates between cohorts quickly waned and were indistinguishable
by months 5 or 6 of continued use.
Liver enzyme monitoring was chosen by the FDA as the primary means of
reducing ALF risk with troglitazone. Testing was not performed consistently
or frequently enough to provide meaningful protection to patients. It is unknown
whether monthly monitoring would have prevented the development of ALF had
it been performed as recommended. A review of 43 ALF cases reported to the
FDA found 12 in which monthly monitoring was performed.13
Of these, 9 (75%) went from normal enzyme levels to irreversible liver injury
within a 1-month interval. In the other 3, troglitazone use continued for
a month beyond the first detected abnormality, so that it is unknown whether
ALF would have been avoided had the drug use been stopped a month earlier.
Also, 1 case of ALF occurred in each of 3 separate postmarketing studies,
despite monthly monitoring. These data call into question the utility of enzyme
monitoring for the prevention of ALF with troglitazone. In addition, the risk
of ALF did not abate with continued use of the drug, suggesting that if monitoring
were capable of preventing liver failure, it would need to be performed at
high levels of completeness for as long as patients continued treatment.13 If monitoring were shown to be protective against
ALF, linkage of drug dispensing with normal laboratory test results might
be considered. This approach worked well in reducing the incidence of agranulocytosis
The influence of labeling recommendations and warning letters on physician
and patient behavior is poorly understood. Walker et al18
found that the cumulative incidence of liver function testing during the first
8 weeks of diclofenac use was below 20%, despite recommendations in the "warnings"
section of its label. Recently, Smalley et al19
found that an FDA "black box" warning for cisapride, with an accompanying
"Dear Healthcare Professional" letter, did not meaningfully reduce the high
level of contraindicated use of this product.
There are several potential study limitations. Data were obtained from
computerized claims, and it is possible that not all laboratory claims were
submitted to UnitedHealth Group. However, because facility reimbursement depends
on filing of claims, this probably did not occur frequently. Claims lag is
another means by which laboratory tests might be undercounted. To minimize
this, data collection was performed 6 months after the end of the final cohort
period. More than 95% of facility claims typically are filed within 6 months
of the date of service.
This study suggests that labeling changes, including "black-box warnings"
and instructions to monitor patients closely, as well as repeated "Dear Healthcare
Professional" letters to physicians, cannot be assumed to be effective means
of risk management. More effective strategies are needed if drug risk management
programs are to benefit patient safety. Such methods should be pilot-tested
and evaluated before they are presumed successful.