Community engagement in a randomised trial
Peter Diggle, Lancaster Medical School, Lancaster University, UK
11th April 2021
In my career as a medical statistician I have been involved in the design and analysis of a number of randomised trials in the UK, but in none of these was I invited to meet a trial participant, and a data-set was never more than a spreadsheet of numbers. Indeed, it’s all too easy to think that the impersonal nature of a statistician’s involvement in a randomized trial is essential to maintain their objectivity. Nowadays, most of my work is oriented towards population health research in low and middle income countries, where my experience has been somewhat different, and all the better for it. I have been privileged to accompany field researchers as they engage with remote communities to explain why they want to measure health outcomes, how the communities might benefit and, more importantly, what the proposed study cannot do. I have observed in these communities an understanding of how small risks to individuals can be outweighed by community-wide benefits that contrasts favourably with the way such issues are often presented in the UK news media, for example in relation to Covid19 vaccine safety. The story I want to share here is about a trial conducted in the Chikwawa District, southern Malawi.
The perimeter of the Majete Wildlife Reserve in Chikwawa district is home to around 25,000 people distributed over 65 villages. Malaria is endemic and most houses offer only rudimentary protection against night-biting mosquitos. A few years ago I was invited to act as the statistician for a trial of two community-led interventions intended to reduce the risk of malaria transmission.
My first involvement in the trial was to join a two-day field-trip to visit several of the villages and explain our reasons for wanting to run the trial. In one, I met a nurse who, with his one assistant, acted as pharmacist, general health advisor and midwife for five neighbouring villages. In the room where he dispensed medicine by day and slept on his camp bed by night, we showed him computer-generated maps of the seasonal ebb and flow of malaria incidence throughout the area (see photo above), and he was immediately able to relate the hot-spots on the map to local environments that provided good breeding sites for anopheles mosquitos. In the other room of his workplace and home, a mother-to-be was in labour. My overall impression was of serenity in the face of constant struggle to meet the health care needs of the local community with inadequate resources.
Having gained a first-hand appreciation of the importance of the trial, my next task was to design it. Put simply, which villages were to receive either or both of the interventions – supplying villages with house-improvement materials or the means to spray mosquito breeding sites – and which were to be used as controls? Because mosquitos fly, a constraint on the design to avoid contamination of treatment effects was that different treatments could only be applied to villages, or sets of villages, separated by at least 500 metres. This made it impossible for us to include every village in the trial, which raised an obvious ethical issue – how to decide which villages to exclude whilst maintain the trust and engagement of the whole of the local community. To address this, we came up with six designs that were equally efficient from a statistical perspective and collectively included every village at least once. The study team then visited all 65 villages to explain the rationale for the study design and invite villagers to attend a community event at which one of their number chose the design from one of six sealed envelopes.
Did the trial work? Well, in one sense it exceeded our expectations. Over the duration of the trial the incidence of malaria fell substantially, but by amounts that did not differ greatly across the different arms of the trial. Given that most trials are set up to find differences between treatment arms, you could say that this constitutes a failure. An alternative interpretation is that a general increase in the community-wide awareness of the importance of limiting exposure to malaria-transmitting mosquitos helped to reduce disease risk everywhere.
As a student, I was taught that statistics is about extracting information from quantitative data, and that the randomised trial is the gold-standard method for doing so. Now, when I teach statistics I emphasise that its purpose is to contribute to a better understanding of complex social, biological and environmental processes using a combination of data and context, and that in the health arena the context is the community. From that perspective, the randomised trial is but one of a number of methods at our disposal, and not always the one best-suited to answering the important questions.
McCann, R.S., van den Berg, H., Diggle, P.J., van Vugt, M., Terlouw, D.J., Phiri, K.S., Di Pasquale, A., Maire, N., Gowelo, S., Mburu, M., Kabaghe, A.N., Mzilahowa, T., Chipeta, M.G. and Takken, W. (2017). Assessment of the effect of larval source management and house improvement on malaria transmission when added to standard malaria control strategies in southern Malawi: study protocol for a cluster-randomised controlled trial. BMC Infectious Diseases, 17,
DOI: 10.1186/s12879-017-2749-2