Author Affiliations: Department of Medicine, The Johns Hopkins University School of Medicine, Baltimore, Maryland.
The thiazolidinediones, rosiglitazone and pioglitazone, are peroxisome proliferator–activated receptor γ (PPAR-γ) agonists. They effectively reduce glycated hemoglobin among patients with type 2 diabetes mellitus by approximately 1 to 1.5 percentage points compared with placebo and are used as second-line treatment agents. However, rosiglitazone and pioglitazone have been associated with peripheral edema, congestive heart failure, and bone fractures.1,2 An increased risk of myocardial ischemia has been attributed to rosiglitazone.2 An increased risk of bladder cancer associated with pioglitazone is noted in the current label. Spontaneous reports of macular edema with the use of thiazolidinediones has resulted in regulatory warnings on this potential association. However, the causality remains unclear.
The prevalence of diabetic macular edema (DME) in one or both eyes among patients with diabetes is estimated to be 7.1%.3 Approximately one-third of these affected eyes lose vision. Since PPAR-γ receptors are present in the retinal vasculature, PPAR-γ–mediated fluid retention has been postulated to contribute to DME. In this issue of the Archives, Idris et al4 report the results of a retrospective, population-based study of adult patients with type 2 diabetes in the United Kingdom. The investigators used data from The Health Improvement Network (THIN), which allowed for long observation of the exposed patients and adequate adjustment for important confounders, including body weight, glycated hemoglobin levels, and concomitant medications. After propensity score adjustment, the authors reported a significantly increased odds of DME in users of thiazolidinediones compared with nonusers (odds ratio [OR], 2.3 [95% CI, 1.5-3.6]) at 1 year. Both pioglitazone and rosiglitazone were associated with a similar increased risk. This increased risk persisted at 10 years of follow-up.
Importantly, the authors appropriately used multiple imputation methods to handle missing data in this cohort and used propensity score methods to address confounding by indication, which is a prominent risk with observational data; however, several limitations preclude definitive conclusions. First, the authors did not have information on the duration of thiazolidinedione exposure or duration of diabetes in the THIN data; that is, they compared prevalent users of thiazolidinediones—patients taking thiazolidinediones for some unspecified time—to nonusers of the drugs.5 This may have led to underascertainment of early-onset macular edema, which would have underestimated the OR.
Second, the positive predictive value of the READ Codes (a thesaurus of coded medical terms maintained and distributed by the United Kingdom Terminology Center) for DME in the THIN Database is unknown. This may have caused misclassification of individuals, which would have underestimated the OR.
Third, since thiazolidinediones are typically used as second-line agents, the reported differences may reflect the underlying diabetes severity rather than effect of the drug. The users of thiazolidinediones had higher glycated hemoglobin levels and were more likely to be on insulin regimens, while nonusers had higher use of cardioprotective agents such as aspirin and angiotensin-converting enzyme inhibitors. Although the authors used propensity score methods to address these differences, these methods cannot adjust for either unknown or unmeasured confounding factors. The direction that this biases the results is unknown.
Finally, the authors did not control for risk factors altered by the thiazolidinediones, such as glycemic control, which is a prominent concern when the exposed individuals are observed for a long time.
These findings should be seen in the context of other studies. A prospective study of 170 000 patients with type 2 diabetes in the Kaiser Permanente database in Southern California reported that thiazolidinedione users were more likely to develop macular edema (OR, 2.6 [95% CI, 2.4-3.0]).6 The authors of that study used administrative codes of unknown validity for the diagnosis of DME, and they could not account for duration of diabetes, duration of thiazolidinedione use, and other confounders. A disproportionality analysis of data from the US Food and Drug Administration Adverse Event Reporting System (FDA-AERS) database reported an increased reporting OR (ROR) for macular edema associated with the thiazolidinediones (ROR, 3.88 [95% CI, 2.79-5.39]).7 However, the increased risk was limited to rosiglitazone (ROR, 5.55 [95% CI, 3.94-7.79]) and was not found for pioglitazone (ROR, 0.99 [95% CI, 0.51-1.88]). The limitations of this study include the lack of information on confounders.
In contrast, other observational studies have reported no elevated risk. A large cross-sectional analysis from the baseline data of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) Eye Study8 reported no association between thiazolidinedione use and macular edema (OR, 0.97 [95% CI, 0.67-1.40]). The clinical examination for DME was standardized and conducted by masked evaluators who graded all photographs for the presence of DME. However, the assessment of thiazolidinedione exposure was by self-report, and the authors were unable to establish the timing of thiazolidinedione use with this cross-sectional design. Another small prospective study with a mean follow-up of 2.8 years also reported no significant difference in macular edema comparing rosiglitazone users (n = 124) with matched controls (n = 158), but the study was significantly limited by its very small sample size.9
The large, long-term clinical trials of rosiglitazone and pioglitazone have not reported an increased risk of serious adverse events of macular edema associated with the thiazolidinediones (eg, Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes [RECORD trial]10 [NCT00379769; n = 2220 in the rosiglitazone group and n = 2227 active controls; mean follow-up, 5.5 years] and the Prospective Pioglitazone Clinical Trial in Macrovascular Events11 [NCT00174993; n = 2605 in the pioglitazone group and n = 2633 in the placebo group; average follow-up, 34.5 months]). However, 7 patients being treated with rosiglitazone reported macular edema as an adverse event compared with 3 active controls in the RECORD trial.10 A meta-analysis of clinical trials is likely to be low yield given the rarity of this outcome and its poor ascertainment.
In conclusion, we can neither be certain that thiazolidinediones cause macular edema nor be reassured that such a risk does not exist. Future studies using new-user incipient cohort designs with validated exposure and outcome definitions and appropriate adjustment for diabetes severity may provide additional information on this potential association.
There are important lessons for evaluation of benefits and risks of prescription drugs. Current approaches are excessively focused on determining the probability of rare but potentially serious risks with newer treatments. The population impact of commonly occurring events such as hypoglycemia associated with older drugs such as sulfonylureas and insulin receives little attention. Very little is known about the relative importance that patients assign to both benefits and risks and whether patients are willing to trade off certain serious risks for potential benefits.
The conduct of timely postmarketing safety studies is particularly important when drugs are approved based on improvements in surrogate end points of efficacy such as glycated hemoglobin. Appropriately interpreting the benefits and risks of a drug requires quantification of the uncertainty of the benefits and harms and an understanding of the relative importance of the outcomes to patients. What should clinicians do when faced with imperfect evidence? Despite the uncertainty regarding the risk of macular edema and thiazolidinediones, the occurrence of characteristic visual symptoms among patients taking thiazolidinediones or any other diabetic medication should prompt evaluation and ophthalmologic referral for DME evaluation, as noted in the current drug labels. The optimal choice for therapy for an individual patient should be determined after eliciting patient preferences for various benefits and risks in a shared decision-making context.
Correspondence: Dr Singh, Department of Medicine, The Johns Hopkins University School of Medicine, 624 N Broadway, Ste 680 B, Baltimore, MD 21205 (email@example.com).
Published Online: June 11, 2012. doi:10.1001/archinternmed.2012.2461
Financial Disclosure: None reported.
Singh S, Segal JB. Thiazolidinediones and Macular EdemaComment on “Association Between Thiazolidinedione Treatment and Risk of Macular Edema Among Patients With Type 2 Diabetes”. Arch Intern Med. 2012;172(13):1011-1013. doi:10.1001/archinternmed.2012.2461