Abstract
Voluntary licences are increasingly being used to expand access to patented essential medicines in low-income and middle-income countries (LMICs). Since 2014, non-exclusive voluntary licences have been issued by Gilead and Bristol-Myers Squibb for key drugs for hepatitis C virus (HCV) infection. We aimed to evaluate the effect of these licences on access to HCV treatment. We conducted a difference-in-differences analysis, exploiting the staggered and selective introduction of voluntary licensing in different countries, to identify the effect of voluntary licensing agreements on treatment uptake. We extracted Polaris Observatory data on the total number of people infected with HCV, diagnosed with HCV, and treated for HCV, and constructed a longitudinal panel of LMICs over a 13-year period (2004-16). Countries were included if they were classified as LMICs by the World Bank in 2014, and had available data on HCV outcomes. The exposure was defined as inclusion in any voluntary licence agreement for HCV drugs. Treatment uptake was calculated as the number of people treated for HCV in a given year per 1000 living people ever diagnosed with HCV. We fit difference-in-differences linear regression models controlling for different confounders that could influence treatment access and uptake, including country and year fixed effects and a range of country-level factors. We additionally assessed the dynamics of the effect and the robustness of our findings. 35 countries were included in the panel: 19 in the intervention group and 16 in the control group. In the simplest model, adjusting only for country and year fixed effects, voluntary licences were associated with an increase in the annual number of people accessing HCV treatment of 69·3 per 1000 diagnosed (95% CI 46·7-91·9; p=0·0060). After adjusting for country-level covariates, this increase was 53·6 per 1000 diagnosed (25·8-81·5; p=0·0354). The effect of licensing increased over time, and was largest in the second year after implementation. Results were robust to alternative specifications. Voluntary licensing initiatives appear to substantially improve HCV treatment uptake in eligible countries. This evidence supports the expansion of licensing strategies to include more countries and more treatments. Unitaid and Médecins Sans Frontières.
Highlights
Introduction of Gilead licenceIntroduction of Bristol-Myers Squibb (BMS) licence and expansion of Gilead licenceMean number of people treated for hepatitis C virus (HCV)Year www.thelancet.com/lancetgh Vol 7 September 2019 e1191In all specifications the difference-in-differences panel estimator could be subject to serial and within-cluster correlation, potentially leading to standard errors that are not robust.[31]
Our final panel consisted of 35 low-income and middle-income countries (LMICs) with treatment data within the period 2004–16
The intervention group consisted of 19 countries included in either voluntary licensing agreement, and the remaining 16 countries with no such agreement formed the control group. 14 countries had both the Gilead and BMS licensing agreements, one had the Gilead licensing agreement only, and four had the BMS licensing agreement only. 14 (74%) countries in the intervention group and 12 (75%) in the control group had medium-quality or high-quality HCV data inputs
Summary
Mean number of people treated for HCV (per 1000 diagnosed). In all specifications the difference-in-differences panel estimator could be subject to serial and within-cluster correlation, potentially leading to standard errors that are not robust.[31] With a small number of clusters and few observations per cluster, the asymptotic justification of clustered standard errors can fail, resulting in biased estimates. We used rand omisation inference, a method that estimates inference using the distri bution of treated groups and addresses both the serial correlation issues and the potential over-rejection of the null hypothesis with other methods.[31,32,33] We did randomisation inference using the user-written Stata package ritest.[33]. We ran sensitivity analyses and robustness checks to probe the main assumptions that support the validity of our methodology.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
