Veerle Vanlerberghe, Marleen Boelaert, François Chappuis, Shyam Sundar, Epco Hasker, Bart Ostyn, Shri Prakash Singh, Joris Menten, Albert Picado, Murari Lal Das, Basudha Khanal, Kamlesh Gidwani, Jean-Claude Dujardin, Suman Rijal, and Marc Coosemans
In a recent paper, Nagpal et al. [1] voiced concerns about the limited or biased use of scientific evidence to support public health interventions to control neglected tropical diseases (NTDs). Visceral leishmaniasis (VL), also known as kala-azar, is one of the major NTDs and does not escape this problem. Transmission is vector-borne and the Indian subcontinent is the region reporting most of the VL cases worldwide. In this region, the main causative species is Leishmania donovani and Phlebotomus argentipes is the vector. Transmission is considered anthroponotic and peridomestic—occurring at night when female sand flies bite people sleeping inside their house. The World Health Organization and the governments of India, Nepal, and Bangladesh set out in 2005 to eliminate VL from the region by 2015 through a combination of early treatment of cases and vector control. However, while recent advances in diagnostic tools and drugs have significantly improved case management strategies, the available vector control tools against P. argentipes remain limited. The elimination initiative promotes the use of indoor residual spraying (IRS) of households and cattle sheds to reduce vector density, but the evidence underpinning the effectiveness of IRS in this region is scanty. Historical observations show that L. donovani transmission declined concomitantly with dichlorodiphenyltrichloroethane (DDT) spraying during the 1950s–60s to eradicate malaria. In the aftermath of this malaria eradication campaign, very few VL cases were observed in endemic regions until the mid-seventies, when there was resurgence of a VL epidemic in India [2]. To date, there are no randomized trials showing the effect of IRS on the incidence of clinical VL [3,4], though some studies showed a reduction in vector density. When the VL elimination initiative was launched in 2005, there were no clear alternatives for IRS as a vector control strategy. Insecticide treated nets (ITNs) were proposed as an alternative or complement to IRS on the basis of analogy arguments regarding their given efficacy against malaria [5] or on data from observational studies suggesting ITNs reduce the risk of VL [2]; but as for IRS, there were no randomized trials evaluating the effect of ITNs on L. donovani transmission. In this context, a number of field studies were conducted in the Indian subcontinent in the past decade to evaluate the effectiveness and impact of ITNs and other vector control tools on VL. Most of these studies have been reviewed in detail in two recent papers [3,4]. The only two studies evaluating the impact of vector control interventions on clinical outcomes found conflicting results. First, the KALANET project, a cluster randomised controlled trial (CRT) in India and Nepal, showed that mass-distribution of ITNs did not reduce the risk of L. donovani infection or clinical VL [6]. Then, an intervention trial in Bangladesh suggested that widespread bed net impregnation with slow-release insecticide may reduce the frequency of VL [7]. Technical (e.g., type of nets and insecticides, lack of replicas and randomisation in Bangladesh) and biological factors (e.g., insecticide susceptibility and sand fly behaviour) may explain the different results observed. This apparent contradiction raises the question about the role that ITN may play in controlling VL in the Indian subcontinent but has also triggered a lot of discussion on methodology and evidence levels required when evaluating vector control tools for VL. In this paper, we would like to summarise the lessons learned from the KALANET CRT in terms of methodology to inform the generation of future evidence and discuss interpretation of findings against this background. The KALANET trial was designed to evaluate the distribution of ITNs as a public health intervention to prevent VL in the Indian subcontinent. The objective was to answer the question: “in the current context in India and Nepal, would free mass distribution of ITN significantly reduce the incidence rate of VL in endemic regions?” This was a question asked not at the individual level, whether sleeping under an ITNs protects an individual against VL compared to an individual not sleeping under an ITN, but a question asked at program level: does such prevention measure reduce VL incidence in communities. Furthermore, the question was about effectiveness in real life conditions and not a question about efficacy of ITNs in “laboratory conditions,” a difference that is clear for public health experts but not necessarily for all readers. The answer to this question may, for instance, be entirely different in a country such as Bangladesh, where no vector control program was operating for many years, no spraying was implemented at the time, and fewer households were using untreated nets, compared to India and Nepal. The first lesson we learned was about the importance of clarifying the research question itself. To answer the above question, we adopted a study design used previously in a successful intervention trial on zoonotic VL transmission [8]. We designed a CRT to demonstrate a 50% reduction on the risk of L. donovani infection associated to the village-wide distribution of ITN [6]. Long-lasting insecticidal nets (LNs) were chosen as intervention as they remain effective for three to four years in the field [2]. Incident L. donovani infection, measured as seroconversion in the Direct Agglutination Test (DAT), was used as the main outcome. Measuring the impact of LN on the risk of clinical VL would have been the preferred primary outcome, but its low incidence and the long incubation period precluded this. Incidence of VL cases was nonetheless measured as secondary outcome. The trial was conducted in 26 high-incidence clusters (16 in India and 10 in Nepal) with over 20,000 inhabitants followed over 24 months. After randomisation, LNs were distributed in all households in the 13 intervention clusters, with the number of LNs proportional to household size, to make sure that all household members could sleep under the nets. Participants in the control clusters were allowed to continue using their untreated nets. The effect of LNs on the incidence rate of seroconversion and VL was compared after 24 months between intervention and control clusters. No LNs were used in the control clusters. The results of the trial, analysed as suggested by Hayes and Moulton [9], showed that the large scale distribution of LN did not reduce the risk of L. donovani infection [6]. These results were consistent across several endpoints measured, as no difference was observed in (1) incidence rate of clinical VL [6], (2) seroconversion in rK39 ELISA (a second serological marker) [10], and (3) mean P. argentipes exposure measured at cluster level by a sand fly saliva antibody detection ELISA [11]. Similarly, the reduction of P. argentipes density indoor in the study clusters was limited (24.9%) [12]. The main conclusion of the trial was that “there is no evidence that using LNs as a public health intervention provides additional protection against VL at community level compared with existing control practices in India and Nepal (e.g., irregular use of untreated nets and IRS). This does not mean that the use of LNs in those VL endemic regions should be dismissed, as they may provide some degree of personal protection against sand flies [13] and have been shown to reduce the risk of malaria [6]. However, the VL elimination initiative in India and Nepal cannot rely on the stand-alone use of LNs to effectively control transmission.” The above message is complex and was disappointing for many, in the first place for the researchers themselves, as the hopes for a user-friendly, household-controlled tool to control VL in the Indian subcontinent were given a serious blow. Moreover, results from a negative trial are hard to communicate. Criticism of peers focused on four main areas: 1. the biological rationale for the intervention, 2. the choice and number of units of analysis, 3. the choice of endpoint and 4. the adherence to the intervention. Stockdale and Newton also identified these methodological issues as key factors to evaluate studies testing preventative methods against human leishmaniasis infection [3].