The authors concluded that rosiglitazone appeared to be safe with respect to the risk of incidence of cancer, but that further studies were needed to confirm a possible protective effect. Given potential for bias in the review, issues with the statistical analysis and uncertain quality of the included studies, the authors' conclusion should be interpreted with caution.

To assess the effect of rosiglitazone on the incidence of cancer.

MEDLINE was searched for studies published in English up to February 2008. The website of GlaxoSmithKline (the manufacturers of the drug rosiglitazone) was also searched for relevant studies.

To be eligible for inclusion, studies had be randomised controlled trials (RCTs) that compared rosiglitazone with either placebo or other active hypoglycaemic drugs, with a minimum duration of 24 weeks. Eligible trials were required to report fatal and non-fatal serious adverse events (including incidences of cancer). Trials reporting no events were excluded.

Most included trials were of patients with type 2 diabetes; non-diabetic patients had metabolic syndrome, plaque psoriasis, insulin resistance, multiple sclerosis, rheumatoid arthritis, Alzheimer's disease, human immunodeficiency virus (HIV) infection, coronary artery disease, and polycystic ovary syndrome. Where reported, the mean age of participants ranged from 14 to 71 years (the review reported a weighted mean age of 55.3 years). Included trial duration ranged from 24 to 208 weeks. The type of cancer included gastrointestinal, pancreatic, pulmonary, mammary gland/female genital tract, male urogenital tract, other known origin, or not specified. Around half of the interventions were placebo-controlled; other comparators included metformin, glyburide, glimepiride, glipizide, pioglitazone, and sulphonylureas; some studies used no comparator.

The authors did not state how many reviewers assessed studies for inclusion.

The authors did not state that they assessed validity.

Data were extracted on incidence of cancer to calculate odds ratios (ORs), with 95% confidence intervals (CIs).

The authors did not state how many reviewers performed the data extraction.

Odds ratios, with 95% confidence intervals, for incidences of malignancies or cancer were pooled using a Mantel-Haenszel procedure. As all treatment and control arm durations were not of the same length, incidence densities (number of events per hundred patient-years) were pooled using random-effects models.

Where possible, subgroup analyses were undertaken for trials with different comparators, patients with type 2 diabetes, non-diabetic patients, trials with a duration of 52 weeks or more, and the most common types of cancer.

Eight RCTs were included in the review. The text reported 16,322 patients in treatment arms and 12,522 patients in control arms.

There was no statistically significant moderation of the risk of cancer associated with rosiglitazone (OR 0.91, 95% CI 0.71 to 1.16). The incidence density of cancer was 0.23 (95% CI 0.19 to 0.26) cases per hundred patient-years in the treatment group and 0.44 (95% CI 0.34 to 0.58) cases per hundred patient-years in placebo and control groups.

Sensitivity analyses did not significantly alter the cancer risk estimates.

The use of rosiglitazone appeared to be safe with respect to the risk of incidence of cancer, but further studies are needed to confirm a possible protective effect.

This review addressed a clear research question, with clear study selection criteria. The search was relatively restrictive but appeared adequate; search terms were not reported. Only English language studies were sought in the MEDLINE search, so the risk of language bias could not be ruled out. Review methods were not reported, so the risk of reviewer errors and/or bias affecting study selection, assessment of study quality and data extraction was unclear.

No study quality assessment was reported, so the reliability of the included studies was unclear. Few primary study details were reported, so the extent of clinical heterogeneity between studies was unclear. The study details that were reported indicated that some studies did not have a comparator arm, even though the study selection indicated they were necessary for inclusion. Data-extraction and synthesis approaches both appeared adequate, although the validity of excluding studies with no events was unclear. No tests for statistical heterogeneity were reported, although the review went some way to assessing heterogeneity using sensitivity analyses. Although the results were clearly reported, it was unclear how many studies were included in the meta-analysis and subgroup analyses. As the authors acknowledged, the RCTs included may typically be underpowered and too short in duration to detect risks or benefits associated with long-term use of the drug, and that the small number of incidences per trial could pose challenging problems in statistical analysis.

Given uncertainty regarding both the appropriateness of the data and the conduct of the review, the authors' conclusion should be interpreted with caution.

Practice: The authors did not state any implications for practice.

Research: The authors stated that the effect of the drug on incidences of malignancies can only be obtained through epidemiological studies, or meta-analyses of trials designed for other purposes, and that larger databases are needed to elucidate the risk profile for individual forms of cancer in rosiglitazone-treated patients.

Partially funded by Regione Toscana, TRESOR Project.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.