Cukier and colleagues' [1] analysis of the putative association between bovine collagen implants and subsequent dermatomyositis and polymyositis raises substantial issues about establishing causality, especially in the absence of biologic plausibility [2]. Our previous case–control study [3] of newly diagnosed cases of dermatomyositis and polymyositis failed to identify any cases associated with bovine collagen implants.

Questionable assumptions exist in the epidemiologic methods they used; all increased the likelihood of apparent association. First, the number of cases observed must be a subset of the number of persons at risk. The authors included in their numerator patients who only received collagen skin tests as well as those who were treated with collagen implants, but the number of patients at risk (denominator) excluded patients who were skin tested only; this figure includes two cases that should have been excluded and overestimates the true rate by a factor of perhaps 30%. Second, rather than counting all exposures, the authors limited the period of observation to that in which the "cases" occurred. For example, they did not include the period between July 1988 and June 1991, when no new cases of dermatomyositis or polymyositis were reported, despite the fact that an estimated 320 000 additional patients were treated with collagen injections. Third, the incidence rates reported by Oddis and colleagues [4] are known to be underestimates because of omission of patients with myositis who were diagnosed and treated outside of the hospital and because of probable underascertainment. Fourth, the most recent 5-year period studied by Oddis and colleagues had much higher incidence rates than did the reference period used by the authors. These methodologic errors systematically decrease the total number of estimated cases by severalfold.

Their discussion failed to note that a Food and Drug Administration (FDA) scientific panel examined all clinical, immunologic, and epidemiologic studies presented (including theirs) in October 1991 and concluded that "there is insufficient statistical or biological evidence to support a conclusion that collagen injections cause autoimmune or connective tissue diseases in persons without a history of these diseases" [5].