Health Economists in Clinical Trials: Why Early Engagement Matters

Health economists in clinical trials are most useful when they join early, explain their methods in plain language and plan for the data they will need. That was the central message of a webinar brought together by University College London and the Global Health Network at the University of Oxford. A professor of health policy and systems management in Uganda moderated three panelists with trial experience across Africa and Asia.

Why trials need economic questions

The moderator framed the session as a move from efficacy to decisions. A trial can show whether an intervention works. Decision makers then ask whether it is affordable and what it displaces. One panelist put it simply: economics is about choices. Every resource has a cost, even when the patient does not pay for it. A drug supplied free by government still has a cost, and the payer is simply someone else.

Three panelists, three settings

The health economist from the University of York described four trials. In Nepal, a team modelled a new ulcer dressing for neuropathic ulcers caused by leprosy, using a Markov model built in Excel to estimate lifetime health gain and cost. In Pakistan, the team compared a technology-assisted, peer-delivered perinatal mental health intervention with delivery by community health workers. A trial of behavioural activation for people with depression and diabetes in Bangladesh and Pakistan analysed costs from the health-payer, individual and societal perspectives, then extended them with lifetime modelling. The speaker now co-leads a hypertension trial in community pharmacies in low-income urban areas of both countries, where cost-effectiveness modelling started while the trial was still being designed.

Ad

The professor from the Centre for Global Health Economics at UCL said the work spans about 15 randomised trials over roughly 14 years, mostly in low- and middle-income countries in Southeast Asia and East Africa. Topics include HIV, tuberculosis, maternal and child health, early childhood development and noncommunicable diseases. Many of the trials were community-based or multisectoral. The professor cited a participatory women’s group trial in rural India on child stunting, and a trial in Bangladesh that combined community groups with mobile health messages to prevent type 2 diabetes. That trial was highly effective and highly cost effective. The professor also warned that economic evaluations are sometimes requested only after data collection, which makes retrospective analysis hard.

The public health specialist from Amref Africa described two Tanzanian studies. The first is a quasi-experimental evaluation of integrated primary care clinics that combine hypertension, diabetes and HIV services. Each integrated facility is matched against a vertical programme comparator on size, type, location and patient volume. The second converts HIV clinics into comprehensive chronic care clinics across 20 primary health facilities over about three years. Facilities switch in waves of three or five, with the order set by a statistician who is not involved in running the intervention, stratified by district and HIV patient numbers.

Practical advice from the panel

• Join early and stay close to the team. The York speaker advised getting to know clinicians and statisticians from the start and explaining methods in a session the whole team attends. The Amref speaker met the field team weekly rather than waiting for quarterly steering committees, which exposed a confusing patient cost diary before messy data arrived.
• Design the case report form with the analysis in mind. Resource-use questions often sit at the end of long questionnaires, when participants are tired. One team had so much detail on drug names and prices that the data were hard to use.
• Pilot tools in another area at the same facility level. The Amref speaker estimated that around 30 percent of tools were changed after piloting.
• Model before the trial ends. The York team built a lifetime model at the start of a five-year project. It suggested a ceiling of $10 per patient for the intervention to remain cost effective. Value of information analysis then showed which parameters to measure more precisely during the trial.
• Report by audience. For community interventions, the UCL professor costs community time and presents provider, health system and societal perspectives, since funders and governments need different views.
• Ask about what people track. Amref’s cost diary asked patients for transport fares in shillings, but patients rarely tracked them. Asking by mode of travel, such as motorbike, dala dala or walking, improved the data.
• Show horizons beyond the trial. Markov and microsimulation models can extend trial results over a lifetime. Intermediate horizons, such as five or ten years, can suit some decision makers better.