Prophylaxis for tuberculosis in children with HIV: randomised controlled trial

Could prophylactic isoniazid prevent tuberculosis in HIV positive children? Kirsten Patrick considers the latest research

The effect of HIV and AIDS is greatest in sub-Saharan Africa. Researchers in the Western Cape region of South Africa wanted to see what effect prophylaxis with isoniazid would have on death rates (mortality) of HIV positive children living where rates of tuberculosis infection are high. They found that isoniazid prophylaxis improved survival in HIV-infected children. "Prophylaxis," of course, means using a drug to prevent an infection rather than to treat an infection.

In sub-Saharan Africa, HIV and tuberculosis are dual pandemics. The immune suppression associated with HIV makes tuberculosis and HIV co-infection common. Tuberculosis infection speeds up the progression of HIV related illness and also increases the likelihood of other opportunistic infections. In children infected with HIV who are admitted for pneumonia in areas with high prevalence of HIV and tuberculosis, 8% have positive Mycobacterium tuberculosis sputum cultures. In Zambia, a study showed that 18% of HIV positive children who died from respiratory infection had tuberculosis infection and that tuberculous pneumonia was the second leading cause of death in children older than 1 year.

Studies looking at HIV positive adults in Africa have shown that isoniazid prophylaxis effectively reduces the number of new cases of tuberculosis and lowers mortality overall. But how isoniazid prophylaxis affects tuberculosis infection or mortality in children with HIV is not known. In many developing countries only a small minority of children with HIV receive antiretroviral treatment. Finding another means to improve survival in children with HIV would clearly be a good thing. So these researchers sought to find out whether isoniazid prophylaxis would produce the same effects in children as it does in adults. They also hoped to evaluate how prophylaxis would affect disease outcomes in children taking highly active antiretroviral treatment (HAART), and how well isoniazid would be tolerated in children.

The authors conducted a double blind randomised controlled trial (RCT) comparing prophylaxis with placebo. An RCT is considered to be the ideal study for evaluating a drug's effect. When implementing a new treatment, it is not clear whether the effects that are observed are caused by the drug or other factors associated with taking the drug. Statisticians call these other factor "confounders." An RCT, by virtue of its design, reduces confounders by randomising, controlling, and blinding.

"Randomised" means that the allocation of a child to receiving either isoniazid or placebo was done by chance and not according to the severity of their disease, who their doctor was, or when they were enrolled in the study. For proper randomisation, it should not be done by the clinician enrolling the patient in the study because something might influence (or bias) them to choose one group over another for a particular patient. Instead, a trial coordinator, who knows nothing of the patient's backgrounds or clinical details, should allocate patients to groups. This is usually done at a remote site and before enrolment of participants begins, as was done in this study.

"Controlled" means that the group that receives the intervention, in this case prophylaxis, is compared with another group that receives a different treatment or a placebo. This allows researchers to say that an intervention works better than, as well as, or less well than something else. Without a control group, any observed effect of an intervention is essentially meaningless because we cannot draw useful conclusions from it. If randomisation is done properly the control group should have a similar mix of patients, both in terms of disease presentation and demographics, as the intervention group. This allows us to assume, pretty confidently, that any different outcomes between groups are as a result of the intervention and not as a result of other factors.

"Double blind" means that neither the patients nor the doctors administering the treatment knew which drug was being given. This is not always possible but is the ideal because it prevents doctors and patients from modifying their behaviour according to their knowledge and thus influencing outcomes.

The trial was approved by a local ethics committee.

Children infected with HIV aged 8 weeks and older who attended one of two children's hospitals in a single region of South Africa were eligible to take part in the study. They had to weigh 2.5 kg or more (for safe isoniazid dosing), to have access to transport, and to have a parent or guardian give written consent to their inclusion in the trial. Children couldn't take part if they had chronic diarrhoea, severe anaemia, neutropenia, thrombocytopenia, or non-reversible renal failure because this would mean that the drug might be unsafe. If the children were already taking isoniazid or couldn't take it because of a previous reaction they were also excluded because had they been included, their group allocation would not have been random. Children who were already being treated for tuberculosis-that is, already receiving the intervention drug-were allowed to finish their treatment before being enrolled in the study.

Children who were taking HAART could take part in the study if they had been stable taking their treatment for at least two months. This was a decision that may have influenced outcomes because HAART is known to slow the progression of HIV disease and thus could influence mortality. In this study, however, the researchers hoped to analyse outcomes in subgroups, and one subgroup was patients taking HAART. In other words, subgroup analyses can tell us whether the effect of isoniazid was caused, in part, by HAART drugs or not.

The dose of prophylactic isoniazid was 10 mg/kg/day. Because tablets were 100 mg and could only be broken into quarters, a variability of 8-12 mg/kg/day was allowed. Placebo tablets were made by the same manufacturer, looked identical to the isoniazid, and were prescribed in the same way for rigorous double blinding.

In this study, both the prophylaxis and the placebo groups also received 5 mg/kg co-trimoxazole dose. Co-trimoxazole prophylaxis is standard in children who have category B or C HIV disease in South Africa, so the researchers used it in all children so that it would not act as a confounding variable. Some children took this drug three times a week and some took it daily because the researchers hoped to look at that "subgroup" too, but I shall not discuss that here. Multivitamin supplementation and immunisations were given as normal.

Children who developed confirmed or probable tuberculosis during the trial were "unblinded" and given tuberculosis treatment modified according to the antimicrobial susceptibility of their cultured bacilli.

The primary outcome measure was mortality. If a child died, the hospital records were examined where possible, otherwise the researchers obtained verbal information if they could.

After a year, 148 children were enrolled in the study, of which 21 had died. Sixteen deaths occurred in the placebo group and 5 in the prophylactic isoniazid group. Even though the numbers were small, statistical analysis showed that this was highly significant. A board had been set up to monitor the safety of the study. It reviewed these preliminary findings and recommended that the placebo arm of the study be stopped on ethical grounds. They communicated this recommendation after the trial had been going for a total of 16 months. At that time 277 children had been enrolled in the study, of which 263 were included in the final analysis. The overall mortality was 12.2%-32 deaths in the whole group. Mortality in the prophylaxis group was almost half that in the placebo group (8.3% v 16%). The benefit of isoniazid prophylaxis applied regardless of age or the severity of HIV disease. The drug was well tolerated, with none of the children in the trial having a severe reaction to it.

The incidence of culture positive or probable tuberculosis was lower in the prophylaxis children than in the placebo group (5/132 cases, 3.8% v 13/131, 9.9%). The time to becoming infected with tuberculosis was also longer in the prophylaxis group compared with the placebo group. Statistical analysis found both these findings significant.

The researchers did the best study they could to answer their research question. Because the preliminary findings caused the trial to be stopped early, the study numbers were relatively small, and long term follow-up was not possible. The main outcome findings were robust, however. Subgroups within the study population could not be adequately evaluated as the authors had intended. For example, the usefulness of isoniazid prophylaxis in patients receiving HAART could not be assessed because few patients were taking it.

Because this study found that prophylaxis significantly reduces mortality in HIV infected children, it is unlikely that another identical study would get ethical approval. This study was approved because, at the time it began, children did not routinely receive isoniazid prophylaxis if they were diagnosed as having HIV. So giving placebo was not unethical-it was standard management. Further trials like this one would increase the evidence base and allow us to be surer of the study findings, but allowing another study such as this would involve a difficult ethical dilemma.

Further research would need to look at the cost efficacy of isoniazid prophylaxis and at its usefulness in patients taking HAART.

This paper confirms that using isoniazid as prophylaxis in children infected with HIV who don't yet have tuberculosis lengthens lives and staves off tuberculosis infection, in the short term at least.

Isoniazid prophylaxis may be an important intervention to improve the survival of children with HIV in areas where antiretroviral treatment is not routinely available-the case in many parts of the developing world.