We agree with Drs. H. Yazici and Y. Yazici that studying the adverse effects of a drug is difficult and complicated. As they point out, meta-analyses of randomized controlled trials that focused on the relationship between TNF antagonists and cancer have yielded conflicting results; observational studies can play an important complementary role by supplying information on typical populations observed long term. The point made in their letter regarding the importance of using the same source population when selecting exposed patients and comparator patients, and ensuring that the exposed and comparator groups have similar indications for treatment and similar risks of experiencing adverse effects, cannot be overemphasized.
As the authors suggest, the risk of adverse effects from many rheumatic disease treatments has yet to be adequately defined; this may be especially true for pediatric populations, patients receiving newer biologic agents, and patients in specific high-risk subgroups (e.g., patients with a prior cancer). Table 1 highlights several key considerations that should be taken into account when assessing observational studies of adverse effects of drugs. When it comes to important issues, such as the risk of cancer related to therapies, clinicians want the truth, and patients generally desire straightforward answers. In these difficult areas, however, one is reminded of a quote from Oscar Wilde: “The plain and simple truth is rarely plain and never simple.”
|1. Study design||Strong consideration should be given to cohort studies. These allow for clear definition of the start of the followup period (start of exposure and/or comparator drug), covariate assessment period (prior to the start of drug exposure), end point assessment after an appropriate latency period following the start of followup, estimation of incidence rates, and calculation of relative risks.|
|2. Source population||Exposed and comparator subjects should be chosen from the same source population and should have similar indications for treatment and similar risks of experiencing adverse effects.|
|3. Exposure assessment||The start and end of the exposure periods must be clearly defined for exposed and comparator groups, considering how concurrent exposures are handled.|
|4. Reference exposure||Nonusers are often a poor choice for comparator subjects because of confounding by indication, unclear index date (in cohort studies), and the risk that nonusers differ across exposures of interest. Strong consideration should be given to using subjects receiving alternative drugs as comparator subjects.|
|5. New users||When hazards are likely to change over time, new users of the drug of interest and the comparator drug may be the optimal subjects.|