Risk of Acute Myocardial Infarction, Stroke, Heart Failure, and Death in Elderly Medicare Patients Treated with Rosiglitazone or Pioglitazone A Cohort Study Osler Journal Club September 8,2010 Faculty advisor: J Hunter Young, MD MHS Rupa Krishnaswamy, MD Background • Use of rosiglitazone may be associated with increased risk of serious CV events when compared with pioglitazone • 2007 Meta-analysis of 42 RCTs with rosiglitazone: 1.4 fold increase in AMI (compared with non-thiazolidinediones) • Meta-analysis of 19 RCTs with pioglitazone: reduction in composite outcome of non-fatal AMI, stroke, and all-cause mortality Study Outline • OBJECTIVE: If risk of serious CV harm is increased with rosiglitazone compared with pioglitazone • DESIGN: Observational, retrospective cohort study of 227 571 Medicare beneficiaries • ENDPOINTS: AMI, stroke, heart failure, death Cohort Study • Cohort assembled based on particular exposure • Followed over time for development of outcome Pros VS Cons • • • • PROS Time based study allows speculation of causality Observational nature allows study of toxic exposures Can study multiple outcomes of a specific exposure • • • • • CONS Time consuming Not randomized Not good for rare diseases or outcomes with a long natural hx Attrition Participants • Medicare beneficiaries • Age > 65 (mean age = 74.4yrs) • Began treatment with thiazolidinedione between July 2006 – June 2009 • Follow-up for up to 3yrs after initiation of treatment (median: 105 days, range 1-1093) • Similar background characteristics, including medical conditions and medications Design • New-user inception cohort design • At least 6 months continuous Part D enrollment and at least 12 months continuous Parts A&B enrollement prior to first thiazolidinedione prescription • Data collected on medical history and use of medications prior to initiating thiazolidinedione – CV, cerebrovascular, DM-related complications, chronic medical conditions – Medications for CV disease, DM, hyperlipidemia Baseline Characteristics of Cohort Members at Initiation of Thiazolidinedione Therapy Graham, D. J. et al. JAMA 2010;304:411-418. Core Cardiovascular Medical Conditions and Medications Used Among Cohort Members During the 12-Month (Medical Conditions) or 6-Month (Medications) Period Preceding Initiation of Thiazolidinedione Therapy Graham, D. J. et al. JAMA 2010;304:411-418. Additional Medical Conditions and Medications Used Among Cohort Members During the 12-Month (Medical Conditions) and 6-Month (Medications) Period Preceding Initiation of Thiazolidinedione Therapy Graham, D. J. et al. JAMA 2010;304:411-418. Study End Points • • • • • • Time to event evaluated for the following events: Acute myocardial infarction Stroke Heart failure All-cause mortality (death) *CV disease accounts for ~70% of deaths in patients with DM. All-cause mortality may be an indicator of CV mortality. Follow-up • Patient follow-up from time of cohort entry to: – First end-point occurrence – Gap in treatment > 7days – Switching to different thiazolidinedione – Hospitalization – End of study period Statistics • Baseline characteristics compared using standardized mean differences – NOT influenced by sample size – Useful for comparing cohorts in large observational studies – Value of 0.1 SD or less: negligible difference in means between compared groups Statistics • Kaplan-Meier cumulative incidence plots show time to event for each end point • Hazard ratios with 95% confidence intervals calculated with Cox proportional hazard models – HR: effect of an explanatory variable on the risk of an event • Number needed to harm estimated using attributable risk – Attributable risk: the difference in rate of a condition between an exposed population and an unexposed population Kaplan-Meier Cumulative Incidence of Time to Event for Acute Myocardial Infarction, Stroke, Heart Failure, and All-Cause Mortality in Elderly Medicare Patients Treated With Rosiglitazone or Pioglitazone Graham, D. J. et al. JAMA 2010;304:411-418. Kaplan-Meier Cumulative Incidence of Time to Event for the Composite of Acute Myocardial Infarction, Stroke, Heart Failure, and All-Cause Mortality in Elderly Medicare Patients Treated With Rosiglitazone or Pioglitazone Graham, D. J. et al. JAMA 2010;304:411-418. Results • • • • 1746 AMIs (21.7% fatal) 1052 Strokes (7.3% fatal) 3307 Heart failure hospitalizations (2.6% fatal) 2562 Deaths from all causes Results: Rosiglitazone vs Pioglitazone • For composite of all end-points, attributable risk is 1.68 (95% CI 1.27-2.08) per 100 personyears of rosiglitazone compared with pioglitazone • Number needed to harm is 60 (95% CI 48-79) persons treated for 1 year to generate 1 excess event Results: Rosiglitazone vs Pioglitazone • Kaplan-Meier plots show no differences in risk for AMI between rosiglitazone and pioglitazone • Increased risk of stroke, heart failure, and death with rosiglitazone compared with pioglitazone Results Hazard Ratios • Adjusted HRs for stroke, heart failure, and death increased with rosiglitazone • Adjusted HRs for composite of AMI, stroke, heart failure, or death increased with rosiglitazone • Adjusted HR for AMI NOT significantly increased Incidence Rates, Attributable Risks (Rate Differences), and Numbers Needed to Harm for AMI, Stroke, Heart Failure, All-Cause Mortality, and a Composite Individual End Point in Elderly Medicare Patients Treated With Rosiglitazone vs Pioglitazone Graham, D. J. et al. JAMA 2010;304:411-418. Post Hoc Analyses • Proportional hazards assumption met for AMI, stroke, heart failure – Proportional hazards assumption: that covariates multiply hazard • Assumption NOT met for death or composite of all end-points • Investigate importance of nonproportionality Post Hoc Analyses • Cohort 110 950 patients (prior to May 21, 2007) – Importance of date: Publication of rosiglitazone meta-analysis by Nissen and Wolski that showed increased risk of AMI with rosiglitazone • Proportional hazards assumption now met for death • Analysis of 15 009 patients who entered after May 2007 showed similar results Post Hoc Analyses • HRs for death and the composite increased with rosiglitazone compared with pioglitazone – Analyzed in time intervals: 0-2mon, 2-4mon, >4mon • Proportional hazards assumption met for death • HRs for rosiglitazone compared with pioglitazone were statistically significantly increased during the third interval – Death: HR 1.21 (95% CI, 1.05-1.39) – Composite: HR 1.23 (95% CI, 1.14-1.34) Post Hoc Analyses • Compare effects of the two thiazolidinediones on end points within subpopulations • Based on use or nonuse of: • Insulin, sulfonylureas, metformin, nitrates, and statins • HRs for each end point similar in patients with and without baseline use Conlusions • Rosiglitazone associated with increased risk of stroke, heart failure, death, and composite of AMI, stroke, heart failure, or death compared with pioglitazone • When compared with pioglitazo, rosiglitazone associated with: – 1.25-fold (95% CI, 1.16-1.34) increase in risk of heart failure – 1.27-fold (95% CI, 1.12-1.45) increase in risk of stroke – 1.14-fold (95% CI, 1.05-1.24) increase in risk of death Conclusions • Risk of AMI not different among the two thiazolidinediones in elderly Medicare patients • Studies that have revealed this risk conducted in younger patients • Assumption: Patients survive to hospitalization to be counted – Incidence of sudden cardiac death increases 6-fold between 6th and 8th decades of life Conclusions • Incidence rates of AMI, stroke, heart failure, and death for pioglitazone cohort similar to PROactive trial – CV trial comparing pioglitazone with other DM therapies – Mean age 61.1, but patients with established macrovascular disease Strengths and Weaknesses • Strengths – Long duration of follow-up – Temporal association of exposure with outcome – Increased generalizability • Wide array of variables among study cohort • 2 groups indistinguishable • Limitations – Not an RCT – Misclassification of end-points