Correspondence

Zidovudine in Asymptomatic HIV Infection

N Engl J Med 1993; 329:1895-1896December 16, 1993

Article

To the Editor:

Cooper et al. (July 29 issue)1 conclude that zidovudine confers benefits that last “more than 2 1/2 years” in asymptomatic patients with HIV infection and more than 400 CD4+ cells per cubic millimeter.

The authors use an “on-treatment” approach to the analysis, as evidenced by the fact that “data were censored three months after the termination of blinded treatment with the study medication.” For a true intention-to-treat analysis, this approach should be supplemented with analyses in which follow-up information is not censored after the termination of blinded treatment2. These analyses should consider the original primary end point for the trial -- AIDS or severe AIDS-related complex -- as well as the end points adopted later. We are not even told whether AIDS developed by the end of the trial in approximately half the patients (308 patients who withdrew plus most of the 190 with a non-AIDS-related end point). One hundred seventy-three patients in the zidovudine group withdrew, as compared with 135 in the placebo group -- a difference that makes the suggested additional analyses especially important. If there was a tendency for patients about to reach an end point to withdraw from the trial, then an appreciable bias in favor of zidovudine could have been created.

Given that it is widely believed that the effect of zidovudine is of limited duration,3-5 a suggestion that the benefit lasts more than two years should be supported by the demonstration of a statistically significant difference in risk between zidovudine and placebo when one considers only the number of person-years at risk after two years of treatment. The small number of patients with end points after more than two years of therapy makes it doubtful that such a significant difference was present. Therefore, the assertion that zidovudine has a beneficial effect that lasts for more than 2 1/2 years in these patients is not justified on the basis of the results presented.

Andrew N. Phillips, Ph.D.
Caroline A. Sabin, M.Sc.
Royal Free Hospital School of Medicine, London NW3 2PF, United Kingdom

5 References

1
Cooper DA, Gatell JM, Kroon S, et al. Zidovudine in persons with asymptomatic HIV infection and CD4+ cell counts greater than 400 per cubic millimeter. N Engl J Med 1993;329:297-303
Full Text | Web of Science | Medline

2
Pocock SJ. Clinical trials: a practical approach. New York: John Wiley, 1983.

3
Friedland GH. Early treatment for HIV -- the time has come. N Engl J Med 1990;322:1000-1002
Full Text | Web of Science | Medline

4
Swart AM, Weller I, Darbyshire JH. Early HIV infection: to treat or not to treat? BMJ 1990;301:825-826
CrossRef | Web of Science | Medline

5
Gazzard BG. When should asymptomatic patients with HIV infection be treated with zidovudine? BMJ 1992;304:456-457
CrossRef | Web of Science | Medline

To the Editor:

The report of the European-Australian Collaborative Group study on early intervention with zidovudine is premature in its conclusion that zidovudine was beneficial. In terms of the data on progression to AIDS or severe AIDS-related complex, the group given zidovudine and the group given placebo differ by only four patients, in whom Pneumocystis carinii pneumonia developed. No individual case data are presented on prophylaxis against P. carinii pneumonia. . . .

Part of the difference in the probability curves is due to differences in the incidence of oral hairy leukoplakia and oral candidiasis (six more cases of oral hairy leukoplakia and two more of oral candidiasis in the placebo group than in the zidovudine group, according to the criteria of the Centers for Disease Control and Prevention [CDC] for group IV disease). But the authors use these conditions in different ways. The definition of severe AIDS-related complex includes “unexplained oral candidiasis” and “oral hairy leukoplakia.” “Modified criteria” for assignment to CDC group IV-C2 include strict clinical qualifications for the diagnosis of oral hairy leukoplakia and oral candidiasis, including a requirement for treatment. The authors' Table 4 does not indicate which definition of oral hairy leukoplakia and oral candidiasis is used. The different definitions could influence the authors' conclusions about quality of life, prognostic importance, and survival.

The probability curves for disease progression are further confounded by the inclusion of CD4+ cell counts below 350 per cubic millimeter as a study end point. The disease-progression data included 113 such CD4+ cell counts as end points in the placebo group (total end points, 129) and 70 as end points in the zidovudine group (total, 76). The vast majority of beneficial effects on overall disease progression came from these end points. The Concorde trial casts serious doubt on the value of CD4+ cell counts as a surrogate marker for the clinical efficacy of zidovudine monotherapy in asymptomatic HIV infection1,2.

Philip B. Berger, M.D.
597 Parliament St. Suite 208, Toronto, ON M4X 1W3, Canada

2 References

1
Aboulker J-P, Swart AM. Preliminary analysis of the Concorde trial. Lancet 1993;341:889-890
CrossRef | Web of Science | Medline

2
Seligman M. Report of the Co-ordinating Committee of the Concorde Trial. Presented at the Ninth International Conference on AIDS, Berlin, Germany, June 6-11, 1993.

To the Editor:

The benefit of zidovudine therapy for patients with more than 500 CD4+ cells per cubic millimeter has yet to be demonstrated. Unfortunately, the size of the sample and the end points were not revised in the study by Cooper et al. to allow for a separate analysis of the group of patients with more than 500 CD4+ cells per cubic millimeter, and data of three groups of patients (those with 400 to 499, those with 500 to 749, and those with ≥ 750 CD4+ cells per cubic millimeter) were pooled when statistical significance was analyzed. This approach is misleading, because the differences between zidovudine-treated patients and placebo recipients with lower CD4+ counts are obviously greater, and pooling these results with those of the other patients might lead to a significant overall result, even when differences in the group with more than 500 CD4+ cells might not be significant.

R. Martinez
Hospital Virgen de las Nieves, 18012 Granada, Spain

Author/Editor Response

The authors reply:

To the Editor: Phillips and Sabin are concerned that there may have been a tendency for patients to withdraw from treatment just before reaching an end point. As we stated in our paper, the study was analyzed on an intention-to-treat basis, so that all data collected up to three months after the termination of blinded treatment were included. However, patients continued to be followed. Survival data available at the termination of the study indicated that in the group of patients who withdrew early, there were six deaths each among the zidovudine-treated patients and the placebo recipients (relative risk, 0.77; 95 percent confidence interval, 0.26 to 2.30). In addition, all available survival data up to mid-1992 showed that after a median follow-up of 127 weeks, there had been 16 deaths in the placebo group and 11 deaths in the zidovudine group (relative risk, 0.69; 95 percent confidence interval, 0.32 to 1.49). Longer-term data will be reported in due course.

The overall relative risk of disease progression was 0.56 (95 percent confidence interval, 0.43 to 0.75), with no evidence to support an increasing or decreasing trend in this risk during the study. Hence, our conclusion, allowing for follow-up, that the benefit lasts more than 2 1/2 years is justifiable.

A single patient in the placebo group received prophylaxis for P. carinii pneumonia during the study. The definition of the end points used for each group of clinical diseases is clearly given in the Methods section. Dr. Berger suggests that the results are confounded by the inclusion of CD4+ cell counts as a study end point. However, we presented the results both separately and combined and demonstrated the efficacy of zidovudine in terms of clinical as well as immunologic end points.

Dr. Martinez questions the benefit of zidovudine for patients with more then 500 CD4+ cells per cubic millimeter. Since the primary objective of our study was to consider disease progression in all patients with more than 400 CD4+ cells, it is most appropriate to present our results for the overall group. Nevertheless, the risk ratios and 95 percent confidence intervals confirm a benefit for patients with 500 to 749 CD4+ cells per cubic millimeter (relative risk, 0.63; 95 percent confidence interval, 0.41 to 0.95) as well as those with 400 to 499 CD4+ cells per cubic millimeter (relative risk, 0.40; 95 percent confidence interval, 0.25 to 0.65).

David A. Cooper, M.D.
University of New South Wales, Sydney NSW 2010, Australia

Jose M. Gatell, M.D., Ph.D.
Hospital Clinic I, 08016 Barcelona, Spain

Citing Articles (1)