01-03-tl-opi-524-4th interim report (additional sensitivity analysis).pdf

Pioglitazone HCl (ACTOS)
KPNC 4th Interim Report, Cohort Study of Pioglitazone and Bladder Cancer
(Study No. 01-03-TL-OPI-524)
FDA Advice / Information request for sensitivity analysis and other analyses
1Center for Clinical Epidemiology and Biostatistics2Department of Biostatistics and Epidemiology3Department of Medicine4Department of Pharmacology CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 2 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
This report presents supplementary analysis to the 4th interim analysis of our ongoing study of the relative risk of bladder cancer with Actos (pioglitazone), as specified in the Request Advice/Information request made by the FDA dated 30 August 2011. The requested supplementary analyses presented in this report are: 1. A sensitivity analysis to assess change of cohort entry criteria to minimize left censoring of 2. Duration analysis for other antidiabetic medications.
3. To investigate effect modification of the pioglitazone association by age and sex.
4. To analyze the pioglitazone association by cancer stage.
Supplementary analyses on the potential confounding and/or detection bias related to the use of
screening for proteinuria and use of urinalyses are incomplete at the time of this report and will
be reported in the final report.
A sensitivity analysis to assess change of cohort entry criteria to minimize left censoring of
exposure
The sensitivity analysis requested by the FDA entailed altering the inclusion and exclusion
criteria. The original inclusion criteria were:
! all patients who are in the KPNC Diabetes Registry and who are age 40 or older as of
! all additional KPNC Diabetes Registry members who reach age 40 at any point before ! patients age 40 or older who enroll in KPNC between January 1, 1997 and December 31, 2002 and who are identified as having diabetes; ! all KPNC members age 40 or older who develop diabetes during this time period. Follow-up for bladder cancer began for each person when they become eligible (ie, on January 1, 1997, or a subsequent date when they are first identified as having diabetes and being 40 years of age).
The exclusion criteria were: age < 40 years, a diagnosis of bladder cancer recorded in the KPNC cancer registry prior to initiation of observation or within 6 months of entry into KPNC, patients without prescription benefits at the time of entry into the cohort, and those with a gap of more than four months in prescription or membership benefits where the gap started within the first four months of entering the cohort.
The inclusion criteria for this new sensitivity analysis are:! all patients who are in the KPNC Diabetes Registry and who are age 40 or older as of ! all additional KPNC Diabetes Registry members who reach age 40 at any point before December 31, 2009 and who have been enrolled in KPNC since January 1, 1997 without a gap in membership; CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 3 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
! all KPNC members age 40 or older who are newly diagnosed with diabetes between January 1, 1997 and December 31, 2009. The definition of newly diagnosed diabetes requires that the patient be a KPNC member for more than 2 years before being identified as having diabetes according the registry criteria. Follow-up and exclusion criteria for the sensitivity analysis are the same as for the original cohort. Using these new inclusion criteria, we repeated the primary analyses of the 4th interim analysis, examining the relative risk of pioglitazone exposure using the primary definition of exposure, the cumulative dose and duration analyses, the tests for interaction with sex and smoking, and the analysis of stage of disease at diagnosis. The results are summarized in Tables 1 through 9. The size of the cohort increased from 193,099 with 33,416 exposed to pioglitazone to 221,180 with 27,262 exposed to pioglitazone (Table 1). The reduction in the number exposed to pioglitazone despite the increase in the overall sample size is explained by the new requirement for continuous membership since 1997 (Figure 1 and Figure 2). In the original cohort, among the 33,416 patients treated with pioglitazone, there were 8,799 (26%) who entered the cohort after January 1, 1997 and had less than 2 years of follow-up in KPNC prior to cohort entry who were treated with pioglitazone (Table 2). By design, no patients met these criteria in the modified cohort. Patient characteristics were generally similar using the original and modified inclusion criteria (Table 1). There were fewer missing data at baseline in the modified cohort for race/ethnicity, renal function at baseline, hemoglobin A1c, duration of diabetes, and income. The prevalence of complications of diabetes among patients treated with pioglitazone was similar between the original and modified cohort; the prevalence of the complications among patients not treated with pioglitazone was lower in the modified cohort than in the original cohort.
The median time from cohort entry to the first exposure to pioglitazone was now 5.9 years. The median follow-up after the first exposure to pioglitazone was 4.4 years. The median total follow-up time for patients never exposed to pioglitazone was 5.2 years.
To further assess for the possibility of left censoring (i.e., missing pioglitazone exposure prior to the start of follow-up), we determined the number of patients who received a pioglitazone prescription within 4.5 months of cohort entry. A prescription soon after cohort entry suggests that a patient might also have been being using pioglitazone prior to cohort entry. There were 2,392 (7%) and 360 (1%) patients in the original and the modified cohorts, respectively with apioglitazone prescription soon after cohort entry. It is not known whether or not these patients used pioglitazone prior to cohort entry, although it is certainly possible. The proportions of patients who received a prescription within 4.5 months of cohort entry were substantially higher for use of other oral hypoglycemic drugs and insulin (Table 2). 76% of pioglitazone-treated patients had a prescription for another oral antidiabetic drug and 12% for insulin soon after cohort entry. In the modified cohort, these proportions were still high (68% and 11% among pioglitazone-treated, respectively). Thus, it seems reasonable to assume that there was unmeasured use of these medications prior to cohort entry. This is addressed in an analysis of patients with newly diagnosed diabetes that is described later in this report and in Table 10. CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 4 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
In general, there were no substantive differences between the results of the original analyses and
the analyses using the modified inclusion criteria. The fully adjusted hazard ratio (HR) for ever
exposure to pioglitazone was 1.07 (95% CI 0.87 – 1.30) in the original analysis and 1.01 (95%
CI 0.81 – 1.27) in the modified cohort (Table 5). Similarly, for the longest duration of therapy
category, more than 4 years, the fully adjusted HR was 1.30 (95% CI 0.91 – 1.86) in the original
analysis and 1.26 (95% CI 0.84 – 1.88) in the modified cohort. The fully adjusted models used
the same covariates as in the 3rd interim and 4th interim reports. Time updating variables, such as
diabetic complications, were not included in the models.
In the modified cohort, we were not able to include the category of papillary urethral neoplasm
of low malignant potential (PUNLMP), because we only utilized the categorizations in the
cancer registry (Table 9). PUNLMP was only identifiable from medical chart abstraction
completed as part of the nested case-control study, data which were not employed in this cohort
analysis. However, the proportion of bladder cancers among pioglitazone treated patients that
were regional stage or distant were similar: 5% vs. 6% in the original and modified cohorts,
respectively.
Duration analysis for other antidiabetic medications
The FDA also requested duration analyses for other diabetes medications. Because each of the
other medications were utilized prior to the start of the study window, and given that each of the
other medications is often used prior to TZDs, it likely there was substantial left censoring of use
of other antidiabetic drugs in the full cohort. We therefore limited this analysis to patients who
were diagnosed with diabetes after they registered with KPNC (incident sub-cohort), thereby
allowing for complete records of all diabetes medications dispensed to the patients. The
definition of newly diagnosed with diabetes required that the patient was a member of KPNC for
a minimum of 2 years before the first diabetes diagnosis and this diagnosis occurred after
January 1, 1997. Although initially proposed to use patients diagnosed with diabetes as early as
January 1, 1994, when the pharmacy records were online at all KPNC pharmacies, we elected to
limit this to those diagnosed after January 1, 1997 to be consistent with our primary analysis. We
have completed these analyses for duration of therapy with pioglitazone, metformin,
sulfonylureas, and insulin. Among the 59,070 patients who met the inclusion criteria, there were
8,710 patients treated with pioglitazone, 32,726 treated with metformin, 38,708 treated with
sulfonylureas, and 9,861 treated with insulin. In analyses adjusted for age, sex, race, smoking,
and calendar year of cohort entry, we observed no increase in the incidence of bladder cancer
with ever exposure to pioglitazone [0.68 (0.42-1.10)] or any of the other medication groups
(Table 10). Likewise, we did not observe an increase in the incidence of bladder cancer with
more than 4 years duration of therapy with pioglitazone [0.66 (0.21-2.09)], metformin
[0.94 (0.65-1.37)], or sulfonylureas [0.76 (0.55-1.05)] (Table 10). However, because of the small
sample size, particularly in the group with greater than 4 years of therapy, the confidence
intervals were very wide. For example, there were only 3 patients with pioglitazone treatment of
more than 4 years duration who were diagnosed with bladder cancer and the resultant confidence
intervals ranged from 0.21 to 2.09. In contrast to the oral hypoglycemic medications, the
incidence of bladder cancer appeared to increase with longer duration of insulin therapy
(HR 2.00, 95% CI 0.93 - 4.29) (Table 10).
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 5 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
To further address the potential impact of this finding on our previously observed association an increased risk of bladder cancer among patients with longer duration of therapy with pioglitazone, we conducted three additional analyses: 1. Adjusted for the variables in the initial analysis plus each of the other three diabetes therapies where the other therapy is treated as a time updating variable for never versus ever exposed.
2. Adjusted for model 1 variables plus each of the other three diabetes therapies where the other therapy is treated as a time updating variable using the duration categories of never exposed, less than 1.5 years, 1.5 to 4.0 years, and more than 4 years.
3. Adjusted for model 1 variables with follow-up censored at the time of initiation of insulin.
The results of the additional analysis were generally similar (Table 10). When we censored
follow-up at the time of initiation of insulin, the hazard ratio for more than 4 years of
pioglitazone was 0.95 (95% CI 0.30-2.99). Because of the wide confidence intervals, however,
we urge caution against over interpreting all of the results presented in Table 10.
Effect modification of the pioglitazone association by age and sex
The FDA requested that we perform sex-stratified analysis within 10-year age categories
(Table 11). It is noteworthy that there were several strata with small numbers, which prevented
computation of a hazard ratio and confidence intervals. As such, the results should be viewed
with caution. As requested by the FDA, we also conducted an additional analysis with a
dichotomous age definition. These results are presented in Table 12. Although the confidence
intervals are tighter in this analysis than in the analysis using 10-year age categories, caution is
again necessary when interpreting these results.
Pioglitazone association by cancer stage
The FDA requested analyses that excluded in situ cancer from the outcome and analyses that
excluded in situ and local stage cancer from the outcome. Using the original cohort, we
completed the analysis that excluded in situ stage cancer from the outcome. This analysis also
excluded those with an unknown stage from the outcome definition. Patients with in situ,
PUNLMP, or unknown disease stage were censored at the time of their bladder cancer diagnosis.
In the fully adjusted model, the hazard ratios for the longest duration and highest cumulative
dose exposure were greater than that in the primary analysis, although none were statistically
significant. The fully adjusted hazard ratio (95% CI) was 1.17 (0.88-1.55) for ever exposure to
pioglitazone, 1.48 (0.85-2.55) for initiation of pioglitazone 6.5 years prior or longer,
1.60 (0.99-2.59) for more than 4 years of pioglitazone exposure, and 1.49 (0.95-2.33) for more
than 35,000 mg of total dose (Table 13).
Because there were only 7 pioglitazone-exposed patients with regional or distant stage cancer,
we did not complete the analysis limiting the outcome to only regional or distant stage bladder
cancer.
Additional results requested by the FDA
Additional results requested by the FDA included the fully adjusted hazard ratios for all variables
included in the fully adjusted primary analysis from the original cohort. These data are provided
in Table 14. These data confirm expected positive associations with bladder cancer for increasing
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 6 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
age, male sex, Caucasian race, and smoking. No association was observed for diabetes duration
and risk of bladder cancer or for exposure to any of the other categories of diabetes medications.
Additional analyses planned and promised in the Fourth Interim Analysis Report from
24 May 2012
As promised, we tested for an interaction by sex and smoking in the test for trend analysis. None
of the tests for interaction were statistically significant at a threshold of p<0.1 (see revised
Table 6).
Results that are pending at the time of this report
We are actively exploring the issue of whether disproportionate use of testing for proteinuria
may have confounded the results or led to a detection bias. We have designed a series of analyses
to explore this issue and anticipate having these completed for inclusion in the final report.
Comments on the results of these supplemental analyses
In general, the sensitivity analysis using the alternative cohort inclusion and exclusion criteria
resulted in very similar results as the primary analysis (HR [95% CI]: 1.07 [0.87, 1.30] vs.
1.01 [0.81, 1.27]). This suggests that the inclusion of new members to the KPNC plan after
1997 for whom there could be missing data on pioglitazone use prior to cohort entry did not
appreciably bias the results. There was a slightly larger cohort but fewer pioglitazone exposed
patients and particularly a reduction in the number of patients with long-term exposure. The
result was slightly wider confidence intervals than in our primary analyses, reflecting lower
statistical power. This likely resulted from the smaller number of pioglitazone-treated patients
and fewer with long duration of exposure despite the larger overall sample size. We note that the
patients in the sensitivity analysis are likely to have missing data on exposure to other diabetes
medications prior to cohort entry.
The analysis limited to newly diagnosed patients (incident sub-cohort) did not observe an
association or exposure-response relationship between pioglitazone use and risk of bladder
cancer. For more than 4 years of pioglitazone use, the risk of bladder cancer was 0.66 [95% CI:
0.21, 2.09]. It is important to consider reasons why the results of this analysis differ from that of
the full cohort. This new analysis included only newly diagnosed patients while the full cohort
included incident and prevalent diabetes. The advantage of an incident cohort analysis is that it
assures that there is complete data on use of all anti-diabetic medications and that all patients
have comparable duration of diabetes. The major limitation is the reduced statistical power given
the much smaller sample size. The difference in the results could reflect reduced precision due to
decreased sample size in the incident sub-cohort, or alternatively inadequate adjustment for
confounding by insulin or other medication exposures in the full cohort, or to different duration
of diabetes in the full cohort and incident sub-cohort. If the latter were true, it would imply that
either there is an interaction between pioglitazone exposure and diabetes duration or that
duration of diabetes has been incompletely controlled for in the full cohort. Further analysis of
the incident sub-cohort at the final 10-yr follow-up should have improved statistical power and
may provide further insight into the differences between the full cohort and the incident
sub-cohort.
The possible association of long duration of insulin use and bladder cancer in the duration
analysis is worthy of consideration. Because pioglitazone is sometimes used as an insulin sparing
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 7 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
agent and at other times in combination with insulin, future studies directly comparing
pioglitazone to insulin while adjusting for diabetes duration would be informative. Likewise,
studies that excluded patients treated with insulin, as we have done in this subgroup analysis
(Table 10, Model 4) could be informative.
Summary
In response to requests by the FDA, we have provided a number of additional data items for
review. The hazard ratios for the covariates included in the primary analysis are consistent with
the known epidemiology of bladder cancer. The age- and sex-stratified analyses are more
difficult to interpret because of the small number of subjects per stratum. This is evident by the
markedly increased width of the confidence intervals surrounding the point estimates.
The analysis excluding in situ cancers and papillary neoplasms of uncertain malignant potential
resulted in slightly higher hazard ratios than the primary analysis, although these were not
statistically significant. Likewise, the test for trend was not significant in these analyses.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 8 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Flow Diagram of the Original Cohort Entry Criteria
Start with all patients who are in the KPNC diabetes registry and are members age 40 or older as of 01/01/97 (n=83,161) enrolled with existing diagnosis of diabetes between 01/01/97-12/31/02 (n=116,824) All KP members age 40 or older who are diagnosed with diabetes by 12/31/2002 (n=207,389) Exclude those had a diagnosis of bladder cancer CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 9 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Flow Diagram of the Modified Cohort Entry Criteria
Start with all patients who are in the KPNC diabetes registry members and who are age 40 or older as of 01/01/1997 (n=83,161) Add additional registry members who reach 40 All KP members age 40 or older who are diagnosed with diabetics by 12/31/2009 (n=238,238) diagnosed diabetes requires that the patient be a KPNC before being identified as having diabetes according to the registry criteria.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 10 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Demographics of the Study Cohort According to Pioglitazone Treatment at
any Time During Follow-up§
Original Cohort
Modified Cohort
Pioglitazone
No pioglitazone
Pioglitazone
No pioglitazone
treatment
treatment
(n=33,416)
(n=159,683)
(n=27,262)
(n=193,918)
the start of follow-up#Diabetes duration at baseline CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 11 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Demographics of the Study Cohort According to Pioglitazone Treatment at
any Time During Follow-up§ (continued)
Original Cohort
Modified Cohort
Pioglitazone
No pioglitazone
Pioglitazone
No pioglitazone
treatment
treatment
(n=33,416)
(n=159,683)
(n=27,262)
(n=193,918)
§ All variables are at any time during follow-up except for some baseline variables noted. All comparisons have p-
values <0.01 except female sex (original cohort p=0.64; modified cohort p=0.25).
* Creatinine >=1.4 mg/dL for women and >=1.5 mg/dL for men.
‡ Low income defined as median household income in census block below the cohort average (original cohort
$59,000; modified cohort $60,000).
# The variable used to define new diabetes differed in the primary and sensitivity analyses. In analysis of the original
cohort, new diabetes includes patients newly entering the diabetes registry, either because they were newly
diagnosed patients or who newly enrolled in Kaiser Permanente with an existing diagnosis of diabetes mellitus. In
the analyses of the modified cohort the definition of newly diagnosed diabetes requires that the patient be a KPNC
member for more than 2 years before being identified as having diabetes according to the registry criteria.
¥ Includes use of any other diabetes medications during follow-up.
† BPH medications to treat benign prostatic hypertrophy.
μ Number and percentage among males.
∑ Other bladder conditions include hematuria, retention, urgency, neurogenic bladder, catheter and other bladder/urethral symptoms.
β Diabetes complications include diabetic retinopathy, peripheral neuropathy, proteinuria, diabetic nephropathy or coronary artery disease.
∞ Includes diabetic neuropathy, foot ulcer, or amputation.
Ω Includes microalbuminuria or macroalbuminuria.
π Creatinine ≥=2.0 mg/dL for both men and women.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 12 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Comparison of When Patients Entered the KPNC Diabetes Registry and
Medication Use Within 4.5 Months (145 Days) of Cohort Entry
Original cohort entry criteria
Modified cohort entry criteria
Pioglitazone
No pioglitazone
Pioglitazone
No pioglitazone
treatment
treatment
1997 and with at least 2 years of KPNC membership at cohort entry 1997 and less than 2 years of KPNC membership at cohort entry * 1 or more prescription within 4.5 months of cohort entry.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 13 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Pioglitazone Exposures as of the End of Follow-up
Category
ORIGINAL COHORT
MODIFIED COHORT
Time since starting pioglitazone (median, range) CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 14 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Incidence Rate and Relative Hazard of Bladder Cancer with Pioglitazone Exposure in the Modified Cohort
Adjusted for age,
Bladder cancer
Adjusted for age,
sex, year of cohort
incidence rate
sex and year of
entry and
follow-up
(per 100,000
Unadjusted
cohort entry
Fully adjusted†
person-years)
(HR, 95% CI)
(HR, 95% CI)
(HR, 95% CI)
(HR, 95% CI)
Time since starting pioglitazone
Duration of therapy
Cumulative dose
†Fully adjusted refers to inclusion of all potential confounders in the statistical model from the last report plus year of cohort entry: age, sex, race/ethnicity, other diabetes medications, smoking, other bladder conditions, median household income, congestive heart failure, cancer other than bladder cancer, renal insufficiency, HbA1c and the interaction with new diagnosis of diabetes, duration of diabetes, and year of cohort entry. *Also adjusted for use of other diabetes medication.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 15 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Incidence Rate and Relative Hazard of Bladder Cancer with Pioglitazone Exposure
ORIGINAL COHORT
MODIFIED COHORT
Bladder cancer
Bladder cancer
incidence rate
adjusted†
incidence rate
adjusted†
Person-years of
(per 100,000
Person-years of
(per 100,000
follow-up time
person-years)
follow-up time
person-years)
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 16 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Incidence Rate and Relative Hazard of Bladder Cancer with Pioglitazone Exposure (continued)
ORIGINAL COHORT
MODIFIED COHORT
incidence
incidence
Person-years
adjusted†
Person-years
adjusted†
of follow-up
(per 100,000
of follow-up
(per 100,000
person-years)
person-years)
†Fully adjusted refers to inclusion of age, sex, race/ethnicity, other diabetes medications, smoking, other bladder conditions, median household income, congestive heart failure, cancer other than bladder cancer, renal insufficiency, HbA1c and the interaction with new diagnosis of diabetes, duration of diabetes, and year of cohort entry. *Also adjusted for use of other diabetes medication.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 17 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Revised for Original Cohort
Results of age, sex, and calendar year of cohort entry adjusted models stratified by sex and smoking status. Revised to include test for trend results interaction p values.
Non-smokers
HR, 95% CI
HR, 95% CI
HR, 95% CI
HR, 95% CI
* Reference group is unexposed to pioglitazone.
† Test for trend combines two strata. Interaction p values are for the interaction between the exposure and sex or smoking in the test for trend analysis.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 18 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Modified Cohort
Results of age, sex, and calendar year of cohort entry adjusted models stratified by sex and smoking status.
Men Women
Non-smokers
HR, 95% CI
HR, 95% CI
HR, 95% CI
HR, 95% CI
* Reference group is unexposed to pioglitazone† Test for trend combines two strata. Interaction p values are for the interaction between the exposure and sex or smoking in the test for trend analysis.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 19 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Results of Age, Sex, and Calendar Year of Cohort Entry Adjusted Models
Stratified by Sex and Smoking Status
Non-smokers
HR, 95% CI
HR, 95% CI
HR, 95% CI
HR, 95% CI
* Reference group is unexposed to pioglitazone.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 20 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Cancer Stage by Exposure Status
ORIGINAL COHORT
MODIFIED COHORT
No pioglitazone
No pioglitazone
Pioglitazone treated
treatment
Pioglitazone treated
treatment
Cancer stage
(n=137 cases)
(n=952 cases)
(n=105 cases)
(n=1,015 cases)
* Papillary urethral neoplasm of low malignant potential.
+Not available because this analysis includes patients for whom chart review has not been completed.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 21 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Analysis of Duration of Other Diabetes Therapies Among the 59,070 Patients
Who Were Newly Diagnosed§ with Diabetes During 1997-2002
Pioglitazone
Metformin
Sulfonylureas
HR (95% CI)
HR (95% CI)
HR (95% CI)
HR (95% CI)
N exposed cases/ N exposed by end of follow-up Duration of therapy Model 1
Ever/Never Exposed
Model 2
Duration of therapy
Model 3
Duration of therapy
Model 4
Duration of therapy
§ The definition of newly diagnosed with diabetes required that the patient was a member of KPNC for a minimum of 2 years before the first diabetes diagnosis.
Model 1 - Results of age, sex, race, smoking and calendar year of cohort entry adjusted models.
Model 2- Adjusted for model 1 variables plus each of the other three diabetes therapies where the other therapy is treated as a time updating variable for never versus ever exposed.
Model 3 – Adjusted for model 1 variables plus each of the other three diabetes therapies where the other therapy is treated as a time updating variable using the duration categories of never exposed, less than 1.5 years, 1.5 to 4.0 years, and more than 4 years.
Model 4 - Adjusted for model 1 variables with follow-up censored at the time of initiation of insulin.
N/A – Not applicable.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 22 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Sex-stratified Analysis Within 10-Year Age Categories, Original Cohort
Entry Criteria
Age 40-49
Age 50-59
Men Women
HR, 95% CI
HR, 95% CI
HR, 95% CI
HR, 95% CI
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 23 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Sex-stratified Analysis Within 10-Year Age Categories, Original Cohort
Entry Criteria (continued)
Age 60-69
HR, 95% CI
HR, 95% CI
HR, 95% CI
HR, 95% CI
* Reference group is unexposed to pioglitazone; models adjusted for year of cohort entry.
† Interaction p values are for the interaction between the exposure and sex or smoking in the test for trend analysis.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 24 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Sex-stratified and Age-stratified (<60 Or >=60) Analysis, Original Cohort
Entry Criteria
Age <60
Age >=60
Men Women
HR, 95% CI
HR, 95% CI
HR, 95% CI
HR, 95% CI
* Reference group is unexposed to pioglitazone; models adjusted for year of cohort entry.
† Interaction p values are for the interaction between the exposure and sex or smoking in the test for trend analysis.
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 25 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Analysis of the Original Cohort But Limiting the Outcome to Local,
Regional, or Distant Stage Tumors
Bladder cancer
incidence rate (per
follow-up
100,000 person-
Fully adjusted†
(HR, 95% CI)
†Fully adjusted refers to inclusion of all potential confounders in the statistical model from the last report plus year of cohort entry: age, sex, race/ethnicity, other diabetes medications, smoking, other bladder conditions, median household income, congestive heart failure, cancer other than bladder cancer, renal insufficiency, HbA1c and the interaction with new diagnosis of diabetes, duration of diabetes, and year of cohort entry. CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 26 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Hazard ratios and Confidence Intervals for All Variables Included in the
Analysis of Ever Exposure to Pioglitazone in the Original Cohort Entry
Criteria
Fully Adjusted
All covariates
HR (95% CI)
Diabetes Medications
Never treated with any diabetes medications Received at least one prescription for a diabetes medication but never met the Cohort Entry
Smoking Before Censor Date
Any Bladder Conditions on or Prior to Baseline
Median Household Income Above Average ($59K) Congestive Heart Failure Prior to or on Baseline
Had Cancer Other than Bladder Cancer Prior to Baseline
CONFIDENTIAL
Pioglitazone HCl (ACTOS)
Study No. 01-03-TL-OPI-524
Page 27 of 27
Supplemental Analyses to 4th Interim Report
10 December 2012
Hazard ratios and Confidence Intervals for All Variables Included in the
Analysis of Ever Exposure to Pioglitazone in the Original Cohort Entry
Criteria t (continued)
Fully Adjusted
All covariates
HR (95% CI)
Serum Creatinine
Baseline HbA1c
Newly Diagnosed Diabetic
Interaction Term
HbA1c >=10 and Newly Diagnosed Diabetic HbA1C Missing and Newly Diagnosed Diabetic Diabetes Duration
CONFIDENTIAL

Source: http://general.takedapharm.com/Trial-Disclosure/01-03-TL-OPI-524-4th_Interim_Report_(Additional_Sensitivity_Analysis).pdf