Case study 3: A gender analysis of a double blind placebo controlled trial of pneumococcal vaccine in HIV- positive adults in Entebbe, Uganda

Dr. Charles Gilks
Senior Lecturer in Tropical Medicine
Liverpool School of Tropical Medicine


The aim of the trial was to test whether polysaccharide pneumococcal vaccine will reduce the incidence of bacterial (pneumococcal) pneumonia and associated diseases in adults with HIV infection in the Ugandan context.


This vaccine has been in use for many years. The most recent 23-valent preparation was licensed in the US in 1983 (replacing the 14-valent product) and licensed in the UK in 1989. It has been widely used without problem in many people with a specific vulnerability to disease caused by the pneumococcus bacterium, Streptococcus pneumoniae. It has been recommended for individuals with HIV/AIDS in the USA since 1989 and the UK since 1992 However it has never been established whether it protects HIV-immunosuppressed people from the different types of invasive pneumococcal disease such as pneumonia, meningitis and septicaemia.

The study was carried out in the Medical Research Council, Department for International Development, Uganda Virus Research Institute and AIDS Programme in Uganda with the full collaboration and support of The AIDS Support Organisation (TASO) Entebbe, a local NGO which has pioneered HIV/AIDS counselling, support and community-based care in Uganda (and across Africa): the TASO philosophy is to live positively with AIDS.

The study was carried out in Uganda because the attack rates for pneumococcal disease in people weakened by HIV/AIDS are far higher in sub-Saharan Africa than in Europe as a result of poverty and overcrowding. Furthermore, with limited access to underfunded and often inadequate health care, mortality with established disease is far higher in Uganda than in the UK. Finding out whether the vaccine works under these circumstances would be the first step in establishing the appropriate use and public health role for this relatively cheap and potentially highly cost-effective vaccine in the AIDS era.


The study was double-blind and placebo controlled - a standard approach in which neither participant, nor researcher knew whether placebo or vaccine had been given. Participants were randomly allocated to one arm of the trial in order to minimise potential bias. Ethical review boards in Liverpool and Uganda approved the study design, but gender was not considered as a specific issue during the design process.

The study recruited 1400 TASO clients (who already know their HIV status and have been counselled properly), enrolled them with their full consent and then followed them up at six-monthly intervals until the end of the study in July 1998.


Unfortunately the study seems to show that the polysaccharide vaccine is ineffective and therefore will have no role in the prevention of pneumococcal disease in HIV-infected adults in Uganda. In fact in the preliminary results there is an increased incidence in the group which received the vaccine of some of the diseases against which the vaccine was supposed to protect. Further analysis is underway to clarify this issue.

1. Gender Analysis:

The following discussion indicates the types of gender issues that might be raised in a standard study design of this type.

Research Questions:

The research questions aimed to benefit HIV/AIDS patients in Uganda, including TASO clients who were involved in the study. Both men and women are infected with HIV in Uganda and enrolled with TASO. The research team did not explicitly question which groups would benefit most either from the process of the trial itself or from a positive outcome (the success of the vaccine in reducing infection). The team were aware that around 60% or more of the TASO clients were generally women, but were not sure what the implications of this might be for study design. The following section discusses some of the implications of gender relations for study design, focusing on the study group.

Study design:

To identify potential biases in the study groups, Step 1, Patterns of ill health, was applied. Two potential areas for bias were identified: differantial risks of infection and differential responses to the vaccine.

Differential risks of infection
If one sex is at higher risk (both biological and social) of contracting the diseases under investigation, the sex which is at higher risk will need to be adequately represented in the trial to answer the research. This has implications both for the calculation of sample size, and for recruitment methods. In this case evidence from the USA shows that pneumococcal disease has a higher attack rate for males across the age spectrum.21 These findings have been corroborated by other data internationally, although no data on gender differences in pneumococcal infections exists for Africa. If infection rates are higher in men, men need to be adequately represented in the sample in order to answer the research question.

In this particular trial, potential gender differences in the risk and impact of contracting HIV must also be considered. In some contexts women may be more vulnerable to HIV infection, both biologically and socially (See 2.3 What is the difference between Women’s Health and Gender Analysis?). In this case, the particular social and economic vulnerability faced by poor women who are HIV positive led to the enrolment of more women than men in the trial (see Implementation, below).

With a predominantly female trial population and with females at lower risk of developing pneumococcal disease, it might have been that too few pneumococcal disease episodes occurred in the trial to allow it to detect a difference between pneumococcal disease rates in the vaccine and placebo arms. In other words the sample size would have been too small. Fortunately in this study, generous allowances were made for drop out rates after enrolment, which were lower than expected, so the margin for error remained wide enough to compensate for this oversight. However, if drop out rates had been as high as expected, the sample size might have been too small to have given the trial enough power to answer the research question.

Gender differences in the risk of contracting pneumococcal pneumonia are attributed to ‘life-style’ factors such as higher rates of smoking and drinking among men. The trial matched characteristics including behavioural factors for each arm of the trial (placebo and non-placebo) in conducting the analysis. In matching smoking it could be argued that this behaviour is not an independent variable as it is influenced by gender. Logistical regression analysis will be used to minimise the influence of inter-dependent variables, but it may be useful to note this as a potential source of bias in interpreting results.

Differential responses to the vaccine:
The research team did not consider whether men and women would respond similarly to the vaccine. It was assumed that there was no likelihood of a different reaction, so that a study of equivalence was not required. However, in retrospect the research team concede that other expectations were challenged by the study results, so that their assumptions about sex and gender similarities might be open to question. The results are currently being analysed and responses to the vaccine will be analysed by sex. This is possible because men were adequately represented in the trial, but if the representation of each sex is not considered the proportion of men or women participating in a trial might be too small to allow for statistically significant differences to be observed. If differences by sex are noted, this might suggest the need to conduct an equivalence study of future interventions targeted at similar diseases. The presentation of the findings of the study will also need to make clear their degree of applicability to men and women.


Responses to ill-health (See: Gender Analysis, Part 3)

How does gender affect men's and women's responses to illness?


How does access to and control over RESOURCES influence how men and women respond to ill-health?

When men die as a result of AIDS, their wives become female heads of households and have fewer resources with which to seek care for HIV /AIDS related illnesses

How do GENDER NORMS affect responses to illness?

When men’s wives die as a result of AIDS, female relatives often care for them, whilst widows may receive no care from family members

Costs and benefits of enrolment
More women than men were recruited to the study. There are a number of likely reasons for this (see matrix, above). Women often discover their HIV status when their spouses die of AIDS related infections. The numbers of women contracting HIV is currently rising. There is some evidence to suggest that men are looked after by women in their households when they become sick, but following the death of their spouses, women may have no source of care available in the household or community. This encourages poor women to seek care from organisations such as TASO. The services offered by TASO in Entebbe are particularly extensive as a result of extra funding provided by the research project. TASO intermittently provides material supplies such as soap and food. It also offers a free, on demand clinical service and in-patient care to those who need it with subsidised medication. Three doctors are employed full-time, as well as two additional counsellors, a nurse sister and two field workers. These services are offered to all TASO clients and are not conditional on participation in the research. Follow up visits for participants are additional to normal TASO attendance but bus fares are paid.

The free services, bus fares and supplies offered are likely to act as stronger incentives for women to attend TASO than for men because women are particularly economically stretched, especially when they have lost a spouse or need free care. However, there appear to be no additional incentives to enrol in the study over and above those to join TASO. Extra costs of study participation, such as extra travel to the clinic, which could act as disincentives, are covered. Potential bias in the sample is therefore probably determined by the practices of the collaborating organisation. In this case, the practices of the organisation aim to support the most disadvantaged groups by offering free services on a self-selection basis and covering the costs of travel.

In other circumstances it is possible that certain groups of women or men will be less able to meet the transport costs of attending services where they might be recruited for a study and that formal or seasonal employment opportunities might prevent them from attending services at particular times of the day or year. These factors might also affect their willingness to be involved in a study. Most of the clients enrolled in this study did not have formal employment. If the study had collected detailed information on the livelihood strategies and household situations of clients, it might have been possible to identify particular groups which were represented, such as married or widowed women. For ethical reasons it would not have been possible to collect such information from TASO clients who declined to be involved in the study. Financial and opportunity costs might also affect the proportion of men and women who drop out of or complete participation in a study, although the continuation rates in this study appear to be the same for men and women.

It seems that in this study the benefits of participation outweighed the costs for the most disadvantaged groups due to the aims and practices of TASO and due to the particular way in which gender shapes the impact of HIV/AIDS in Uganda. However it is possible that in other situations the costs of participating in the study might be too high for certain groups, such as poor, married women with child care responsibilities or men who rely on seasonal labour opportunities. It is therefore important that researchers and the implementing organisations with which they work consider potential barriers to participation for those groups which might benefit most from an intervention.

How does gender affect men's and women's responses to illness?


How do the relative BARGAINING POSITIONS of men and women influence responses to illness?

Women may need to ask the head of household or senior members’ consent before joining the study

Consent procedures:
Consent procedures are particularly important because of the ways gender may influence decision making norms. As the matrix illustrates, women may be less able than men to decide to join a study without consulting the male head of household. The powerful position of researchers, who are often male, may lead women to feel that they are obliged to join a study because it is suggested to them by medical professionals who are seen as ‘knowing best’.

Enrolment in this study was carried out at the normal TASO clinic and took a full morning. All clients at TASO were approached and asked if they wished to take part in the study, but it was made clear that there was no obligation to join. The consent agreement was translated into the local language and then back again into English to check for accuracy and appropriateness of translation. If a client was unable to read the form it was read to them. There are no recorded cases of women stating that they needed to consult other family members, although reasons for declining to participate were not recorded for ethical reasons.

Gender differences in access to resources and work patterns will influence measurable disease outcomes. (See: Gender Analysis, Part 2)

The reliability of methods of identifying end points will also be influenced by gender relations.

How does gender affect men's and women's responses to illness?


How do GENDER NORMS affect responses to illness?

There may be differences in community norms of health and illness between men and women, which may lead to different recognition of symptoms and willingness to accept the sick role between different groups of men and women

The main end point of this study was developing invasive pneumococcal disease. In order to diagnose this effectively a microbiology lab was set up and individuals were investigated fully for each intercurrent morbidity event, in a more systematic way than would normally be practised. Other end points were death, quality of life, and other illness events such as TB or salmonellosis. The study used a variety of methods such as self-reporting by participants and diagnosis on presentation at the clinic, either due to an illness episode or at a regular check up. Gender differences in the self identification of symptoms or in the ability to attend the clinic in response to an illness episode would affect the accuracy of recorded episodes for each sex. Using a variety of methods and providing transport costs, as in this study, are ways of minimising this bias.

Presentation of the findings of the study will include any gender differences which are observed in trial outcomes. If there are no gender differences observed it will be helpful if this is made clear. If the trial had proved the efficacy of the vaccine it would have been important to clarify how applicable the findings were to each gender and to report some of the gender issues which might arise in the distribution of the vaccine such as inequities in access to services which provide it, or in the ability to pay for the vaccine.


Although this trial did not explicitly consider gender issues in its design and implementation, it is clear that working with TASO, an organisation which is aware of the problems faced by disadvantaged groups, helped to foster working practices which addressed some of their needs and prevented these issues from affecting trial outcomes. In some respects, such as the calculation of the sample size, the use of a wide margin of error prevented some of the potential negative consequences of the failure to consider gender differences in infection. If the trial had been successful, further issues about who would benefit most from its use as a medical intervention would have been raised.