To the Editor.
—In their well-done study of the duration of survival benefit of zidovudine in patients with AIDS, Dr Lundgren and colleagues1 comment that visual inspection of product-limit plots for our study2 suggest a survival benefit of about 1 year's duration. Our observational study reported survival estimates on 352 patients who received zidovudine and 362 patients who did not receive zidovudine who were first diagnosed with AIDS from April 1987 through June 1989 and had longitudinal follow-up through June 1990. A total of 470 patients (66%) died during follow-up. When survival from diagnosis of AIDS was analyzed by a time-dependent Cox regression in a manner similar to Lundgren et al, we derived the results listed in the Table. A second time-dependent Cox regression analysis was also done using the following model: relative hazard (t) = exp(B1+[B2×time]), where B1 provides an estimate of the relative hazard of use vs