Case-control or cohort study?

Ashutosh Tamhane, Researcher,

Other Contributors:

February 07, 2011

To The Editor,

In a recently published article by Chandra et al. the authors have stated that they conducted a "case-controlled" study to examine the anticoagulant effectiveness of warfarin in vitreoretinal surgery.(1) However, this study was conducted on a cohort of patients receiving pars plana vitrectomy (PPV); 60 patients who received warfarin (exposure) on the day of PPV were selected. These 'exposed' patients were matched to other patients receiving PPV but not taking warfarin (unexposed) on age, sex, and presenting complaint. These patients were then followed for perioperative and long term complications. Thus, this is not a case- control study rather it is a matched cohort study. As a result, the statistical analysis raises concerns. Matching in a cohort study does not necessarily eliminate the need for control of matching factors in the analysis; censoring (competing risks, death, loss to follow-up) can produce imbalances thereby limiting the extent of matching only to those among the original counts of persons for whom matching was applied.(2) In other words, an association of exposure and the matching factors could be observed for the remaining persons and the observed person-time.2 Furthermore, even if no censoring occurs and one examines risk, control of matching factors would still be necessary to obtain valid standard deviation estimates.(3,4) In contrast, case-control studies require control of matching factors associated with exposure rather than risk. Moreover, with variable follow-up duration and censoring a matched survival (time-to-event) analysis producing hazard ratios would have been a better strategy than reporting the p-values alone; censoring has not been mentioned in the article.

REFERENCES

1. Chandra A, Jazayeri F, Williamson T. Warfarin in vitreoretinal surgery: a case controlled series. Br J Ophthal 2010. http://bjo.bmj.com/content/early/2010/11/11/bjo.2010.187526.full.html. 2010.

2. Rothman K, Greenland S. Modern Epidemiology. 2nd ed. Philadelphia: Lippincott Williams & Wilkins; 1998.

3. Greenland S, Robins J. Estimation of a common effect parameter from sparse follow-up data. Biometrics. 1985;41:55-68.

4. Weinberg C. On pooling across strata when frequency matching has been followed in a cohort study. Biometrics. 1985;41:103-116.

Conflict of Interest:

None declared

Conflict of Interest

None declared