Competing global statistics on prevalence of injecting drug use: why does it matter and what can be done?

Variation in global review estimates of the prevalence of injecting or opioid use is explained partly by the use of alternative information sources and value given to unreferenced country estimates, but such variation also highlights a prevailing problem with the reporting and robustness of prevalence estimates. Unfortunately, there is no quick solution: we need investment both in ongoing information or surveillance of drug-related harms and in the implementation of more refined statistical methods to estimate prevalence. Addiction presents the latest global statistics on alcohol, tobacco and illicit drug use 1 and highlighted discrepancies in estimates of the prevalence of injecting drug use between the 2017 World Drug Report (WDR - https://www.unodc.org/wdr2017/) led by the United Nations Office on Drugs and Crime (UNODC) and the most recent 2017 systematic review led by the global consortium 2 (Table 1). Global evidence on the prevalence of injecting drug use has increased and improved over the past decade—with the most recent review identifying information on injecting from 179 countries covering 99% of the global population, and synthesizable estimates from multiple sources for 83 countries covering 83% of the global population 2-4. The estimates of global injecting prevalence overlap substantially between the 2017 World Drug Report (mid 11 810 000 or 0.25%, ranging from 11 810 000 to 17 420 000) and the 2017 systematic review (mid 15 648 000 or 0.33%, ranging from 10 219 000 to 23 737 500). However, there are more marked differences by country and region (see Table 1), including Latin America (estimates range from 1 392 000 to 2 380 000 versus 380 000 to 1 340 000), Central Asia (189 500 to 416 500 versus 400 000 to 510 000) and South Asia (783 500 to 1 263 000 versus 290 000 to 300 000). There are several potential reasons for this variation. First, the criteria for inclusion of studies differs between the two reports. The World Drug Report uses sources without detail on the origin of the study data, such as estimates reported via its Annual Reports Questionnaire received from member states, which had a range of estimates, the source of which could not be located or obtained. The systematic review excluded estimates that did not have information on the origin of the source study included. Secondly, the two reports used different imputation methods in the absence of local data. The systematic review pooled estimates from studies of similar quality to maximize geographic coverage within a country (where possible), whereas the World Drug Report tended to use data from single studies. Ongoing concerns faced when trying to provide global estimates is that many country estimates are based on methods that are likely to be biased and often the original data used to derive the prevalence estimate are not reported transparently. It is acknowledged widely that population surveys will underestimate prevalence of comparatively rare, hidden and stigmatized behaviours, such as opioid, cocaine, methamphetamine and injecting drug use 5, 6. Alternative indirect estimates that utilize information on drug-related harms, and partial data sets of people in treatment or criminal justice, have been developed, but there has been a lack of investment in critical evaluation—in testing, applying, developing and then implementing the best methods. Estimates of injecting prevalence from the United States are a notable example. These estimates were based on several forms of multipliers across different years, including population survey and other data for specific areas in the United States, and then extrapolated up for the United States 7-10. Further, the exact method of generating multipliers and the data on which it is based were spread confusingly across multiple publications 7-10. In Europe, the surveillance of ‘problem drug use’, including injecting drug use, is moribund—few countries report up-to-date estimates, with no consistency in methods between countries (e.g. see http://www.emcdda.europa.eu/data/stats2017/pdu_en_), and several country estimates are overly reliant upon inherently biased methods 11. This sorry state of affairs has led the consortium at IHME leading the Global Burden of Disease (GBD) study—which has done so much to raise the profile of addictive behaviours and mental health as key drivers of ill health throughout the globe 12, 13— to consider excluding most of the evidence from so-called ‘indirect’ estimates and, instead, re-calibrate estimates of the prevalence of opioid and injecting drug use from ‘direct’ population surveys (based on extrapolation from a few countries that have both a ‘reliable’ indirect estimate and a general population survey). If carried through, this action would be both deleterious and retrograde. It is likely that the prevalence estimates for opioid disorders and injecting drug use would be lower than current estimates, down-weighting their public health importance, as the modelled contribution to disease burden would also fall 14. Indubitably, the estimates would be wrong, based on the heroically unjustifiable assumptions: that only a few countries have reliable indirect estimates and that the relationship between these and population surveys in these countries can apply to the rest of the world. Unfortunately, the right approach to fixing the problem is going to take more time, although there are several pressing reasons to improve prevalence estimates. Addressing the opioid overdose crisis in North America, where opioid related deaths have increased more than threefold during the last 15 years 15, and HCV elimination targets in many countries, where the majority of new infections are in people who inject drugs (PWID) 16, would have benefited from better information on the dynamics and size of the population of opioid users and PWID. Data used for prevalence estimation need to be reported more transparently, so that the methods can be replicated, and the information be made more readily available for quality assessment and synthesis. Such action could be taken now by academic journals, global agencies and those who fund work to deliver interventions to address injecting drug use and illicit drug harms to prevent unreferenced, opaque estimates being circulated. However, this is only part of the solution. More importantly, there has been a lack of investment in strengthening the epidemiological data in countries and cities affected by problematic drug use, which is the foundation of any prevalence estimate. As an epidemic of injecting or opioid use matures, the connections between data sources (and potential biases) also become more complex. In such contexts, investment in record linkage between administrative data sources will support better estimates of drug-related harm, and more refined statistical approaches to prevalence estimation. It has been demonstrated that it is feasible to measure drugs consumed by a population in their sewage/wastewater. Disappointingly, there have been few wide-scale implementations of wastewater monitoring or integration of wastewater data with other epidemiological data to size the drug market or estimate prevalence of drug use in the population 17-19. Therefore, let us ask IHME, the United Nations system and other agencies with an interest in monitoring and preventing drug-related harm to support countries to invest in information systems that document key harms more effectively, and to test and implement the most reliable statistical approaches to estimating disease burden and prevalence.