Correspondence

Questions about a Placebo-Controlled Trial of Preventive Therapy for Tuberculosis in HIV-Infected Ugandans

N Engl J Med 1998; 338:841-843March 19, 1998

Article

To the Editor:

The first randomized trial demonstrating the efficacy of isoniazid (as compared with placebo) in preventing tuberculosis in Haitians with human immunodeficiency virus (HIV) infection and positive purified-protein-derivative (PPD) tests was reported by Pape and colleagues in 1993.1 The elegant trial, also placebo-controlled, reported by Whalen et al. (Sept. 18 issue)2 raises an important question: Is there a rationale for conducting additional trials after a positive result?

There seem to be two main methodologic issues that could justify additional placebo-controlled trials: validity, if there were suggestions of systematic biases in the study by Pape et al., and generalizability, if there were reasons to believe that the findings were specific to Haiti.

Potential threats to validity in the study by Pape et al. were in the direction of underestimating, rather than inflating, the true effect: a higher proportion of PPD-positive patients and longer follow-up in the isoniazid group, and the switch to treatment with isoniazid in some patients initially assigned to the placebo group. The fact that 40 percent of the cases were confirmed by culture in the study by Pape et al., as compared with 50 percent in the study by Whalen et al., would potentially invalidate the results only if differential misclassification had occurred. Other limitations3 — the relatively small sample (118 patients, because Pape et al. stopped enrollment early, on ethical grounds) and the small number of incident cases of tuberculosis (15) — have to do more with precision in the measurement of the estimate of an effect than with validity.

Was there reason to believe that the results would be confined to Haiti? Perhaps, but it is unlikely. Therefore, why replicate? It is often assumed that reliability is a substitute for validity. This is not the case: consistent results may be invalid. The best assurance of intrinsic validity, and of valid statistical inference, is in the randomization, not the replication. . . .

Moïse Desvarieux, M.D., Ph.D.
Columbia School of Public Health, New York, NY 10032

3 References

1
Pape JW, Jean SS, Ho JL, Hafner A, Johnson WD Jr. Effect of isoniazid prophylaxis on incidence of active tuberculosis and progression of HIV infection. Lancet 1993;342:268-272
CrossRef | Web of Science | Medline

2
Whalen CC, Johnson JL, Okwera A, et al. A trial of three regimens to prevent tuberculosis in Ugandan adults infected with the human immunodeficiency virus. N Engl J Med 1997;337:801-808
Full Text | Web of Science | Medline

3
De Cock KM, Grant A, Porter JD. Preventive therapy for tuberculosis in HIV-infected persons: international recommendations, research, and practice. Lancet 1995;345:833-836
CrossRef | Web of Science | Medline

To the Editor:

The study of HIV-seropositive and PPD-positive Ugandan adults included a placebo group of almost 500 persons.

The Journal apparently was influenced to publish this study despite reservations because it had been approved by the institutional review board of University Hospitals of Cleveland and Case Western Reserve University, and all the patients gave informed consent.1 It is hard to believe that a similar study of coinfected adults in Cleveland would ever have been approved by the same institutional review board. Just what is it that makes such a study possible in Kampala, Uganda?

The women in the study were of childbearing age, and close to 70 percent of the participants were women. There was a total of 31 cases of tuberculosis in the placebo group. Not only were adults unnecessarily included in a placebo group, but their children were also exposed and possibly infected as a necessary consequence of this ethically flawed study.

Marie T. Turner, M.D.
Lemuel Shattuck Hospital, Jamaica Plain, MA 02130

1 References

1
Angell M. The ethics of clinical research in the Third World. N Engl J Med 1997;337:847-849
Full Text | Web of Science | Medline

Author/Editor Response

The authors reply:

To the Editor: In her editorial,1 Angell raises questions about the ethics of the clinical trial of preventive therapy for tuberculosis in HIV-infected Ugandan adults, performed by the Ugandan–Case Western Reserve University Research Collaboration. She states that the ethical use of a placebo group in the trial depended “entirely on the strength of the preexisting evidence.” At the time the study began, in March 1993, the protective efficacy of preventive therapy for tuberculosis in patients with HIV infection had not been established. Observational cohort studies provided preliminary evidence that preventive therapy with isoniazid would be effective, but these studies were performed in selected populations at risk and were not properly controlled. Although preventive therapy with isoniazid is beneficial in HIV-seronegative persons, one could not assume an equivalent effect in HIV-infected persons because of impaired host immunity. Indeed, HIV infection confers the greatest known risk of the development of tuberculosis, with estimates of the annual risk as high as 8 to 12 percent.2 In areas where the transmission of Mycobacterium tuberculosis is high, such as sub-Saharan Africa, the beneficial effect of preventive therapy may be abrogated by reinfection, especially in persons with advanced HIV infection. Whether a brief course of preventive antituberculosis therapy could reduce the risk of tuberculosis in dually infected persons in this setting was genuinely unknown. Moreover, given the recognized susceptibility of HIV-infected persons to adverse drug reactions, the safety of preventive therapy had not been properly assessed. Thus, there was equipoise in regard to the efficacy and safety of preventive therapy at the time the study began. Dr. Turner's comments assume that the safety and efficacy of preventive therapy in HIV-infected persons were known at the start of the study.

The study in Uganda was designed and begun before the results of the Haiti study had been published.3 After reviewing these results, we concluded that the Haiti study was inconclusive and decided that the placebo group was justified. This decision was supported by the World Health Organization's Therapy of Mycobacterial Disease Steering Committee in April 1994. The Haiti study was inconclusive because the estimate of a protective effect against tuberculosis was confined to the PPD-positive subjects in the study and was based on only eight cases. Misclassification of only one case in either study group would render the result statistically insignificant.

We are dismayed that Angell alludes to the Tuskegee studies in discussing our clinical trial. The investigators in our research collaboration adhered to all guidelines for the ethical conduct of human research in both developed and developing countries. In fact, we believe that our work represents a model of international collaboration. Today, because a series of properly controlled, randomized trials have been performed, we have solid information on which to base recommendations for the prevention of tuberculosis in HIV-infected persons in both industrialized and resource-poor countries.

Christopher C. Whalen, M.D.
John L. Johnson, M.D.
Case Western Reserve University, Cleveland, OH 44106-4945

Roy D. Mugerwa, M.B., Ch.B.
Makerere University, Kampala, Uganda

Jerrold J. Ellner, M.D.
Case Western Reserve University, Cleveland, OH 44106-4945

3 References

1
Angell M. The ethics of clinical research in the Third World. N Engl J Med 1997;337:847-849
Full Text | Web of Science | Medline

2
O'Brien RJ, Perriens JH. Preventive therapy for tuberculosis in HIV infection: the promise and the reality. AIDS 1995;9:665-673
CrossRef | Web of Science | Medline

3
Pape JW, Jean SS, Ho JL, Hafner A, Johnson WD Jr. Effect of isoniazid prophylaxis on incidence of active tuberculosis and progression of HIV infection. Lancet 1993;342:268-272
CrossRef | Web of Science | Medline

Author/Editor Response

The editorialist, Dr. Msamanga, replies:

To the Editor: Desvarieux asks why Whalen et al.1 replicated the study reported by Pape et al. in 1993.2 Both studies used a randomized, placebo-controlled design. The aim of Pape et al. and Whalen et al. was to determine the efficacy of isoniazid (or other antituberculosis regimens in the case of Whalen et al.) in preventing tuberculosis in HIV-infected, PPD-positive patients in Haiti and Uganda, respectively. I agree with Desvarieux that the trial to examine the efficacy of antituberculosis prophylaxis should not have been replicated, because the question of efficacy was no longer an issue.

The greatest concern now, and this applies to Uganda and many other developing countries, is to implement this type of intervention. The operational issues include the need for voluntary counseling, screening, the type and competence of providers, questions of cost effectiveness, compliance, monitoring of side effects and drug resistance, and logistic problems. These issues are all site-specific, so each country needs to determine how to operationalize the intervention. Furthermore, the effect of antituberculosis prophylaxis on long-term survival has yet to be clarified. I also concur with the points raised by Turner. In our editorial on the report by Whalen et al., my coauthor and I stated “Given our current state of knowledge, however, future studies should not include a placebo group, since preventive therapy should be considered the standard of care.” 3 The issue of the affordability of drugs should be tackled by getting governments, pharmaceutical companies, donors, and other international agencies to cooperate in making drugs cheaper rather than by looking for other, probably inferior, regimens for people in less-developed countries.

Gernard Msamanga, M.D., Sc.D.
Muhimbili Medical Center, Dar es Salaam, Tanzania

3 References

1
Whalen CC, Johnson JL, Okwera A, et al. A trial of three regimens to prevent tuberculosis in Ugandan adults infected with the human immunodeficiency virus. N Engl J Med 1997;337:801-808
Full Text | Web of Science | Medline

2
Pape JW, Jean SS, Ho JL, Hafner A, Johnson WD Jr. Effect of isoniazid prophylaxis on incidence of active tuberculosis and progression of HIV infection. Lancet 1993;342:268-272
CrossRef | Web of Science | Medline

3
Msamanga GI, Fawzi WW. The double burden of HIV infection and tuberculosis in sub-Saharan Africa. N Engl J Med 1997;337:849-851
Full Text | Web of Science | Medline

Author/Editor Response

Dr. Angell replies:

Many readers, like Whalen et al., objected to my drawing an analogy between the ethics of the 1932–1972 Tuskegee study of the natural history of syphilis and some current placebo-controlled trials in the Third World (I was careful to say that the study by Whalen et al. was an arguable case).1 Nevertheless, there are parallels, which I described fully elsewhere2 in response to criticisms of my editorial. Two of these parallels are especially important.

First, like the current trials, the Tuskegee study had an important goal. In 1932, at its inception, the natural history of syphilis was poorly understood, and it was not even clear whether it should be treated at all after its initial stages, particularly given the ineffectiveness and toxicity of available therapy. The researchers continued their study long after penicillin became available in 1944, because they believed it would be the last chance to gather such important information,3 and the results are still cited in textbooks.4 In essence, it was argued then, as now, that the end justifies the means — an ethically untenable position.

Second, as in the Third World today, treatment was generally unavailable to the population from which the men in the Tuskegee study were drawn (poor, black men in rural Alabama). Untreated subjects were not being denied treatment they would otherwise have received. In today's exculpatory phrase, it was not the “local standard of care” to treat syphilis in this population. And it was true then, as now, that the study was sanctioned by responsible authorities in the country where it was performed — but people of very different status from that of the subjects.

One frequently mentioned difference is that the Tuskegee subjects did not give informed consent (not an established principle in those days, anyway). But a recent story in The New York Times raised questions about the quality of the informed consent in the current Third World studies.5 Some subjects claimed they had no idea they might receive a placebo instead of effective treatment, and some were told that entering the study was the only way to get general medical care. In any case, informed consent and official approval do not justify inherently unethical research, especially when there is great disparity between the authority of researchers and that of subjects.

It is easier to see the ethical flaws of the Tuskegee study from a safe distance of 25 years than the flaws of current research. Nevertheless, those offended by the analogy have yet to show how some of the current studies differ in their fundamental failure to protect the welfare of human subjects.

Marcia Angell, M.D.

5 References

1
Angell M. The ethics of clinical research in the Third World. N Engl J Med 1997;337:847-849
Full Text | Web of Science | Medline

2
Angell M. Tuskegee revisited. Wall Street Journal. October 28, 1997:A22.

3
The development of consent requirements in research ethics. In: Faden RR, Beauchamp TL. A history and theory of informed consent. New York: Oxford University Press, 1986:151-99.

4
Lukehart SA, Holmes KK. Syphilis. In: Isselbacher KJ, Martin JB, Braunwald E, Fauci AS, Wilson JD, Kasper DL, eds. Harrison's principles of internal medicine. 13th ed. New York: McGraw-Hill, 1994:728-9.