The purpose of this letter is to draw attention to the misuse of the results of some clinical studies. I refer to a study published in Annals in June 1994 [1]. The aim of this study was to determine whether corticosteroids affect the findings in temporal artery biopsy in patients with temporal arteritis. The authors retrospectively reviewed the charts of 535 patients who underwent temporal biopsy. Each specimen was reexamined by a pathologist. However, every decision about treatment was based on clinical grounds, and there was no gold standard for the diagnosis other than the biopsy result. Biopsy results were positive in 31% of all patients. Therefore, an undetermined number of patients had diseases other than temporal arteritis, and most of these patients were probably included in the nontreatment group because they were statistically less likely to have headache, jaw claudication, visual symptoms, and polymyalgia rheumatica. Even the adjustment for clinical variables would not suffice if different populations are being considered. Thus, the authors' conclusion that treated and untreated patients had similar biopsy findings [1] must be viewed with extreme caution. In the original report the authors did so, and they recognized that "these results do not prove that histologic features are unaffected by corticosteroids."

The following year, however, a second study based on the same population [2] and a review on histopathologic specificity of vasculitis [3] (both appearing in highly cited journals) were published by the authors involved in the first study. In these papers, and without any evidence other than the previous study, the authors stated that "we have demonstrated (elsewhere) that biopsy positivity rate was unrelated to prior corticosteroid treatment" [2] and that "prior corticosteroid treatment has no significant masking effect on the temporal arteritis in biopsy specimens" [3]. In 1 year, the original inconclusive result was turned into an unquestionable one.

Authors should make every effort to guarantee that the bibliography-derived contents of papers are based, as much as possible, on evidence. Readers should also make sure that such evidence exists before assuming that information is definitive.