Monitoring the elimination of human African trypanosomiasis: Update to 2014

Background The World Health Organization (WHO) has targeted the elimination of Human African trypanosomiasis (HAT) ‘as a public health problem’ by 2020. The selected indicators of elimination should be monitored every two years, and we provide here a comprehensive update to 2014. The monitoring system is underpinned by the Atlas of HAT. Results With 3,797 reported cases in 2014, the corresponding milestone (5,000 cases) was surpassed, and the 2020 global target of ‘fewer than 2,000 reported cases per year’ seems within reach. The areas where HAT is still a public health problem (i.e. > 1 HAT reported case per 10,000 people per year) have halved in less than a decade, and in 2014 they corresponded to 350 thousand km2. The number and potential coverage of fixed health facilities offering diagnosis and treatment for HAT has expanded, and approximately 1,000 are now operating in 23 endemic countries. The observed trends are supported by sustained surveillance and improved reporting. Discussion HAT elimination appears to be on track. For gambiense HAT, still accounting for the vast majority of reported cases, progress continues unabated in a context of sustained intensity of screening activities. For rhodesiense HAT, a slow-down was observed in the last few years. Looking beyond the 2020 target, innovative tools and approaches will be increasingly needed. Coordination, through the WHO network for HAT elimination, will remain crucial to overcome the foreseeable and unforeseeable challenges that an elimination process will inevitably pose.


Introduction
In the last decade of the 20 th century, the number of cases of human African trypanosomiasis (HAT), also known as sleeping sickness, reached alarming levels [1,2,3,4]. In reaction to this epidemiological situation of a lethal disease, a number of stakeholders came together to support the affected countries. In the early years of the 21 st century, the World Health Organization (WHO) launched a public-private partnership that, together with important efforts from bilateral cooperation and non-governmental organizations (NGOs), enabled to reverse the epidemiological trend [5]. In this process, the key role was played by the National Sleeping Sickness Control Programmes (NSSCPs) of endemic countries and their committed health workers.
The steady reduction in the number of HAT cases reported during the first decade of the current century prompted first the HAT focal points of endemic countries [6], then the WHO Strategic and Technical Advisory Group on Neglected Tropical Diseases (NTDs) [7] and finally the WHO Expert Committee on control and surveillance of HAT [8] to set the elimination of HAT as a goal. The technical viability of HAT elimination rests on the existence of vulnerable points in the transmission cycle, the present as well as prospective availability of control tools, and the evidence of having reached elimination in several HAT transmission areas [9,10,11].
As a consequence, HAT was included in the WHO NTD roadmap as one of the diseases targeted for elimination as a public health problem by 2020 [7]. The indicators to measure the progress towards elimination were defined, and a reporting calendar was established [8]. The selected indicators for HAT elimination should be monitored every two years, and be presented in HAT stakeholders meetings [12].
A first progress report on HAT elimination looked at the gambiense form only [13], and it provided an update to 2012 for the main indicators of elimination. HAT elimination was shown to be on track, even though the exclusion of rhodesiense HAT data rendered that progress report incomplete. The present paper provides the first comprehensive biennial update, including both the gambiense and the rhodesiense form, and it covers the progress made in HAT elimination from 2000 to 2014. Reported data on HAT occurrence are compared to the milestones set by the WHO roadmap on NTDs [7], in particular as it concerns the target of fewer than 2,000 reported cases by 2020, which is the first global indicator of HAT elimination as a public health problem. Regarding the second global indicator (i.e. 'number of foci reporting less than 1 case per 10,000 inhabitants'), we present here a revised metric based on the concept of 'areas at risk of HAT' [14], which enables a more robust and objective quantification. This revised metric was recently endorsed by the WHO HAT elimination Technical Advisory Group.

Ethics statement
The research does not directly involve human participants. No individual data is used in the paper. All the data used are provided routinely by National Control Programmes as epidemiological information and are fully anonymized.

Number of HAT cases reported annually
Detection of HAT cases is currently undertaken by NSSCPs, NGOs and Research Institutions. HAT morbidity data in disease-endemic countries are collected by NSSCPs or dedicated departments in the Ministries of Health, and subsequently reported to WHO on an annual basis. Field activities including active and passive case finding are regularly reported. Transboundary cases (i.e. individuals who contracted the infection in one country but who were detected by the health facilities in a neighbouring country) are also reported and allocated to the country of infection; national authorities are informed accordingly for appropriate action. Sporadic cases are also detected in non-endemic countries, amongst travellers and migrants. They are reported to WHO thanks to the centralized distribution of anti-trypanosome drugs. Information on the likely area of infection is used to allocate these 'exported' cases to the country of infection [15]. All data are entered in the database of the Atlas of HAT [16].
In the present paper, the number of HAT cases reported from 2000 to 2014 is provided for all endemic countries. These figures include a few minor revisions as compared to previously published counts for the period 2000-2013 [8,13,16,17]. The revisions stem from in-depth verifications carried out for the continuous improvement of the Atlas of HAT (e.g. a more accurate allocation of transboundary cases).

Geographic distribution of HAT
The geographic distribution of HAT reported cases is mapped at the village level following already described methodologies [16,18]. The database includes, from the year 2000 onwards, not only the cases detected actively and passively but also the people examined per village during active screening activities carried out by mobile teams. All records in the database are linked to the source files from which the information was derived, and all source files are safely stored in a digital data repository.
In this paper, emphasis is given to the distribution of HAT cases for the five-year period 2010-2014. Because of the inherent epidemiological features of HAT, and in the context of the elimination strategy, a five-year window is considered as the most useful to analyse and present the updated picture of the extent of the disease [8,13]. In particular, the 5 year window is believed to strike a good balance between temporal resolution (which would call for a shorter window) and robustness (which would call for a longer window, so as to smooth the year-toyear variations in screening intensity).
The risk of HAT infection is estimated from the number of reported cases (numerator-Atlas of HAT) and the exposed population (denominator-Landscan [19]). Previously published methods enable point level data from the Atlas of HAT (village-level mapping) to be converted into continuous, smoothed surfaces of disease intensity and risk [14,20]. Smoothing is based on a 30-km radius kernel density, and although HAT risk was initially estimated over ten-year periods [14,20], more recently five-year periods have been considered more informative to monitor elimination [13]. In the present paper, the progress over time was investigated through a five-year sliding window (i.e. from 2000-2004 to 2010-2014). The 30-km smoothing for the estimation of HAT risk is meant to account for a variety of complex and not easily quantifiable epidemiological features such as the mobility of people and of the vector, whilst at the same time reducing the effect of mapping inaccuracies in the input data. This methodology and its rationale are described in detail elsewhere [14].
On the basis of the number of HAT cases per annum (p.a.) as compared to the exposed population, HAT risk is ranked into five categories: very high (! 1 HAT case per 10 2 people), high (! 1 HAT case per 10 3 people and < 1 per 10 2 people), moderate (! 1 per 10 4 people and < 1 per 10 3 people), low (! 1 per 10 5 people and < 1 per 10 4 people), and very low (! 1 per 10 6 people and < 1 per 10 5 people) [13]. Risk is considered 'marginal' below the threshold of 1 HAT case p.a. per 10 6 people. It is noteworthy that below the category of 'moderate', the risk level fits the WHO general definition of elimination of HAT as a public health problem.
For the present risk estimates, the Atlas of HAT provided village-level mapping for 92.9% of HAT reported cases (period 2000-2014). For the 7.1% of the cases which were not mapped at the village-level, information on the area of occurrence was used (i.e. unmapped cases were proportionally allocated to the endemic villages of the same area [20]).
Population at risk potentially covered by fixed health facilities with capacities for HAT diagnosis and treatment Fixed health facilities play a crucial role in the control and surveillance of HAT. With a view to estimating their physical accessibility and potential coverage of at-risk populations, time-distance analysis was used [21]. In the present paper the coverage of the population at risk of gambiense HAT is updated, and that of rhodesiense HAT is presented for the first time.
Data on the fixed health facilities that are active in HAT control and surveillance were provided by NSSCPs through standardized forms. For each health structure, information was collected on the name, location and capacities for HAT diagnosis and treatment. Data were harmonized, mapped and assembled in a geo-spatial database [21]. The survey was conducted between September 2015 and April 2016.
A time-distance function was used to map the physical accessibility to HAT diagnosis and treatment [21]. A 'landscape friction' geospatial layer for Africa provided the travel time through each 1-km/30 arcseconds pixel [22]. Landscape friction takes into account terrain and transportation network. The cumulative travel time was calculated from any location to the nearest health facility ('shortest weighted distance' or 'least cumulated time'). The economic cost of travel (i.e. affordability) is not considered in our analysis; only travel time is computed.
For presentation purposes, results were summarized for three thresholds of travel time (i.e. 'one hour', 'three hours' and 'five hours') and stratified by risk categories. With a view to exploring trends, results for gambiense HAT are compared to those of a previous survey (completed in August 2013 [21]). For the previous study [21], stratification relied on a 10-year risk layer (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009). To ensure consistency with the present estimates, which are stratified on a 5-year risk layer (2010-2014), previous estimates were recalculated on the basis of the corresponding 5-year risk layer (i.e. 2007-2011).

Number of HAT cases reported annually
A total of 3,797 new HAT cases (including both gambiense and rhodesiense HAT) were reported in 2014 (Fig 1). For this indicator, the continental target set for 'HAT elimination as a public health problem' is fewer than 2,000 cases, a level that was planned to be reached by the year 2020 [7]. The intermediate milestone of 5,000 cases in 2014 was not only reached but surpassed by 1,203 cases.
The number of gambiense HAT cases reported by year and by country is shown in Table 1. In 2014 a total of 3,679 cases were reported, corresponding to an 86% reduction compared to 2000. It is worth noting that the Democratic Republic of the Congo (DRC) continues to account for the vast majority of gambiense HAT cases. In 2014, the DRC accounted for 87% of the total number of cases (3,205 out of 3,679). Concerning rhodesiense HAT, results are shown in Table 2. With 118 cases reported in 2014, the rhodesiense form of the disease continues to represent a small part of the total HAT reported cases (3%). With 709 cases reported in the year 2000, a reduction of 83% in 14 years was observed. Over the last 4 years, the number of rhodesiense HAT cases has stabilized at around 100 per year. For the whole period 2000-2014 (2000 being the start year of the Atlas of HAT), a total number of 206,570 cases has been included in the database. Of these, 92.9% cases have been mapped at the village level, for a total of 30,278 mapped villages. The average accuracy for mapped HAT cases is presently estimated at 1.3 km, and it is being continuously improved.

Geographic distribution of HAT
Gambiense HAT. In West Africa, disease transmission continues in Guinea. The decrease in reported cases in 2014 can be ascribed to the abrupt reduction in case-finding activities during the Ebola outbreak, during which only weak passive case detection was in place. In Côte d'Ivoire, the number of recorded cases keeps decreasing even following a reinforcement of active and passive case detection. In Nigeria, cases are only sporadically reported, against a backdrop of insufficient case detection activities. One autochthonous case was detected by     Kenya  15  10  11  0  0  0  1  0  0  1  0  0  2  0  0  40   Malawi  35  38  43  70  48  41  58  50  49  39  29  23  18  35  32  608   Mozambique  --1  -1  ----------2   Uganda  300  426  327  338  335  473  261  119  138  129  112  84  71  43  70 3,226 United Republic of Tanzania  350  277  229  113  159  185  127  126  59  14  5  1  4  1  1 1,651   Zambia  9  4  5  15  9  7  6  10  13  4  8  3  6  6  12  117   Zimbabwe  -----3  --0  3  2  4  9  1  3  25   Total  709  755  616  536  552  709  453  305  259  190  156  115  110  86  118 5,669 Other In Central Africa, active screening activities are routinely carried out in Cameroon, Chad, Congo, Equatorial Guinea and Gabon, and the system of passive case detection has been reinforced. In these countries, the decreasing trend in the number of cases continues and it is believed to reflect a real abatement in disease transmission (5,720  The reported trend is likely to reflect a real reduction of infections, as the intensity of active and passive surveillance remained remarkably stable over these years, with approximately 2 million people screened per year. Interestingly, vector control activities were recently initiated in a few foci of gambiense HAT, with a view to complementing medical activities. Vector control is therefore believed to have contributed to the observed positive trends in such countries as Guinea [24] and Chad. Rhodesiense HAT. The number of rhodesiense HAT reported cases has been decreasing steadily in the majority of affected countries ( Table 2). Uganda reported 65% of all rhodesiense HAT cases during the period 2010-2014, even while experiencing a reduction of 66% as compared with the previous 5 years (from 1,120 cases in 2005-2009 to 380 in 2010-2014). In Uganda, although a few cases are reported from protected areas probably due to the wildlife reservoir, the main reservoir involved in transmission is cattle [25]. Multisectoral disease control activities including the veterinary dimension were reinforced over the years in a One Health framework, thus enabling disease transmission to be brought under control [26]. A very substantial reduction was observed also in Tanzania [27]. Despite the increase in reported cases, it is difficult to establish whether disease transmission is on the increase in Zimbabwe, as in the period 2000-2004 the reporting system was weaker than in following years. For gambiense HAT, 1.08 million square kilometres are estimated to be at risk of infection (period 2010-2014), with almost sixty thousand square kilometres at very high and high risk, and almost two hundred and eighty thousand at moderate risk. Over 62% of the total area at risk is in the DRC. Compared to the previous 5 years (2005-2009), the total area at risk decreased by 12% (Table 3), with more marked declines for the categories at higher risk (i.e. 60% and 36% for 'high and very high' and 'moderate' respectively). Areas at risk shrank in virtually all endemic countries, with the exception of Chad (+55%), and Gabon (+17%). In absolute terms, the most sizable reductions were observed in Angola (-62,000 km2) and DRC (-44,000 km2).

Areas and population at risk of HAT
Fig 5a focuses on the cumulated categories of moderate, high and very high risk. These categories correspond to the areas where HAT is considered a public health problem (i.e. > 1 HAT case p.a. per 10,000 people). Also, at the global level, they correspond to the second primary indicator to monitor the elimination process, i.e. 'number of foci reporting less than 1 case per 10,000 inhabitants'. We observe that the areas at moderate, high and very high risk halved from 2004 to 2014 (from approximately 680 thousand km 2 to less than 340 thousand km 2 ). The decrease was sharper in the higher risk categories, with the areas at high or very high risk shrinking by 78% (from 276 to 59 thousand km 2 ). Looking at individual countries, the largest strides in reducing the areas where gambiense HAT is a public health problem were made in the DRC (with a reduction of approximately 125,000 km 2 over the last 11 years), Angola (107,000 km 2 ), South Sudan (45,000 km 2 ) and Congo (40,000 km 2 ).
For rhodesiense HAT, approximately 100,000 km 2 are estimated to be at risk of infection ( Table 4). Most of these areas are in the low and very low risk categories (92,000 km 2 ), where HAT is not considered a public health problem. Only 10,000 km 2 are at moderate risk, while no area affected by rhodesiense HAT was estimated to be at high or very high risk in the period 2010-2014. The trends over the past 11 years are presented in Fig 5b. We note that for rhodesiense HAT, in the first monitored period 2000-2004, the extent of the transmission areas where the disease can be considered a public health problem were already much smaller than for gambiense HAT (i.e. 27,600 km 2 for rhodesiense HAT versus 681,000 km 2 for gambiense HAT). Even if already starting from these relatively small areas, by 2010-2014 the reduction was even sharper in relative terms (i.e. 62% reduction for rhodesiense HAT as compared to 50% for gambiense HAT). Focusing on individual countries, the most sizable elimination of rhodesiense HAT as a public health problem was achieved in Tanzania, with a reduction of over 13,000 km 2 , followed by Uganda (almost 4,000 km 2 ), although the reduction in Tanzania has to be interpreted in a context of considerably reduced surveillance activities.
Population at risk of HAT. For gambiense HAT, 55 million people are estimated to be at risk of infection (period 2010-2014), with 1.2 million at very high and high risk, and 9.2 million at moderate risk (Table 5). Therefore, over 10 million people live in areas where gambiense HAT is still considered a public health problem. Most of these people live in DRC (8.5 million). As compared to the previous 5 years (i.e. 2005-2009), the total population at risk increased by over 5 million, but the increase only concerns the low and very low risk categories (+11 million). General population growth and people changing status from higher to lower risk categories account for the growth in the population at low and very low risk, which in  In relative terms, the most dramatic reduction was in the very high risk category (99%, from 265 thousand to 2 thousand people). For population at high and moderate risk the reductions were respectively 78% (from 5.4 to 1.2 million) and 14% (from 10.8 to 9.2 million).
For rhodesiense HAT, over 6 million people are estimated to be at risk (Table 6). Most are in the low and very low risk categories (5.6 million), while only four hundred thousand people are in the moderate risk category. In Fig 6b, we note that the reduction in the three combined risk categories for rhodesiense HAT was 73% (from 1.47 to 0.4 million people). The population at high risk decreased from 65 thousands to zero (2004 to 2014), while the population at very high risk was always zero for rhodesiense HAT throughout the monitored period. Population at risk potentially covered by fixed health facilities with capacities for HAT diagnosis and treatment Survey and mapping of fixed health facilities. For gambiense HAT, the survey completed in February 2016 revealed the existence of 882 fixed health facilities with capacity for diagnosis (+28% as compared to the survey in 2013 [21]), out of which 516 also have capacity  for treatment. Sixty-three percent of the 882 facilities are found in the DRC. The complete results of the surveys, and the comparison with the 2013 survey [21], are provided in Table 7.
While clinical suspects can be identified in all of the 882 facilities, serological testing is available in 72% of them (639, i.e. +46%). The more complex and advanced types of diagnosis, i.e. parasitological diagnosis and disease staging, are available in only 326 (+7%) and 262 (+8%) of the facilities respectively.
A more limited number of facilities offer treatment for gambiense HAT (516, +4%), all of which can administer pentamidine for first-stage infections. First line treatment for secondstage infections (NECT) is provided by 226 facilities (+20%), a relatively low number as this treatment requires highly skilled personnel for administration.
For rhodesiense HAT, 111 facilities offer diagnosis in six endemic countries, i.e. Kenya, Malawi, Uganda, United Republic of Tanzania, Zambia and Zimbabwe ( Table 8). All of these perform clinical diagnosis, while parasitological diagnosis and disease staging are offered by 44% and 31% of the facilities respectively. Thirty-two health facilities are involved in rhodesiense HAT treatment, and all provide both suramin for first-stage infections and melarsoprol for second-stage.
The geographic distribution of the health structures involved in HAT diagnosis and treatment are shown in Fig 7. S1 Text provides maps of the centres able to offer the different types of diagnosis and treatment.
Population at risk potentially covered by fixed health facilities. Table 9 summarizes the potential coverage of the population at risk of gambiense HAT by fixed health facilities. For diagnosis, 28, 43, and 48 million people at risk are respectively within one, three and five hours travel of a competent facility (corresponding to 52%, 77%, and 88% of the at-risk population). For treatment, the corresponding figures are 23 (42%), 40 (73%), and 47 million (86%). These figures show an improved situation as compared to the previous survey, when accessibility for the same categories was 43%, 74% and 81% for diagnosis and 39%, 72% and 85% for treatment.
When looking at the coverage by risk categories, clear improvements were observed in the very low, low and moderate risk, whilst comparatively lower gains were measured in the high and very high risk categories. It is noteworthy that an improvement in accessibility was observed for the more complex levels of diagnosis (serological, parasitological and disease staging) as well as treatment (second-stage infections) (S2 Text).
Among the population at risk of rhodesiense HAT (Table 10), 2.5, 4.6, and 5.2 million people are respectively within one, three and five hours travel of a diagnostic-competent facility (corresponding to 41%, 75%, and 86% of the at-risk population). For treatment, the corresponding estimates are 2.0, 4.4, and 5.1 million (i.e. 34%, 72%, and 84% of the at-risk population).
Importantly, health facilities providing HAT diagnosis and treatment also cover a substantial number of people who are at marginal risk, including areas that were at risk of the disease in the past.

Epidemiological trends
The data presented in this paper, covering the 15 years between 2000 and 2014, indicate clear progress towards HAT elimination as a public health problem, which is on track to be achieved by 2020. For the first time, in 2014 fewer cases than the set milestone were reported (i.e. 3,797 against the 5,000 milestone). The level reported in 2014 had been planned to be reached by 2016-2017. This decrease in reported cases was observed in a context of fairly constant intensity of active screening activities and reinforced passive surveillance in several countries, so the trend is very likely to reflect a real abatement in disease transmission. Preliminary data for 2015 (not presented here) show a further reduction in reported cases, thus corroborating the observed trend. Areas at risk of HAT, which are estimated from reported cases and exposed population, are also shrinking. In particular, the areas where HAT is still a public health problem (i.e. where ! 1 HAT case per 10 4 people p.a. is reported), have been decreasing steadily. Between 2004 and 2014 we estimate a reduction of approximately 360,000 km 2 , i.e. -51% as compared to the 2004 level.
The number of fixed facilities providing gambiense HAT diagnosis and treatment increased, and thus their potential coverage of the at-risk population. Improvements were observed in the basic levels of diagnosis (clinical, serological) and treatment (first-stage), as well as in the more advanced levels (i.e. parasitological diagnosis, disease staging, and first-line treatment for second-stage infection with NECT).
Looking at rhodesiense HAT, we note that this form of the disease continues to represent a relatively small proportion of the total number of HAT reported cases (i.e. 2.7% average for 2000-2014). Against this backdrop, a sizeable decrease was observed between 2000 and 2011, with a reduction of 594 cases p.a. (i.e. 84%). However, since 2011 progress has stagnated and the number of reported cases has stabilized at around 100 p.a. Several factors may have contributed to the stagnation. One is the expanded use of rapid diagnostic tests (RDT) for diagnosing malaria instead of microscopic examination. In fact, microscopic examination enabled the accidental diagnosis of HAT while looking for malaria parasites. Another is the fact that, in the first decade of 2000, the maximum reduction of cases was observed in areas where livestock is the reservoir of T. b. rhodesiense, and where strengthened veterinary public health brought about the decrease [26]. The wildlife reservoir is much more difficult to manage and it represents the main source of the scattered but constant rhodesiense HAT cases that are reported from protected areas [16]. It is also important to note that for rhodesiense HAT the rate of under-detection is likely to be higher than for gambiense HAT. The causes of this are manifold, and they include a faster disease progression, the poor effectiveness of active screening that makes rhodesiense HAT only detectable by passive screening, a lower incidence and therefore a lower awareness and preparedness of health staff, and the occurrence of the disease in sparsely populated areas. As a result of this likely higher rate of under-detection, the reliability of all indicators based on reported cases is bound to be lower for rhodesiense than for gambiense HAT. Applying to rhodesiense HAT the same risk threshold as for gambiense HAT, we observe that the areas where the disease is still a public health problem are very few and very small. In the period 2010-2014 they were mostly limited to a restricted area in central Uganda (Kaberamaido and Dokolo Districts [28]) and the area surrounding the Vwaza Marsh Wildlife Reserve in Malawi [29]. A few additional scattered areas at moderate risk are also found in sparsely populated zones of Zambia (mostly corresponding to the Luangwa National Parks [30]).
Looking at the potential coverage of the at-risk populations by passive surveillance, we provide here the first continental survey of fixed facilities having capacity for diagnosis and treatment of rhodesiense HAT. At this stage, we only note that the coverage of population at risk of rhodesiense and gambiense HAT is comparable. One hundred and eleven facilities for rhodesiense HAT were identified, compared with 882 for gambiense HAT, which at first glance seems proportionate to the respective reported burdens. However, when looking at these figures, the substantial differences between the two diseases have to be kept in mind, and especially the generally lower transmission intensity and the likely higher under-detection for rhodesiense HAT. In this context, we argue that there is a need to expand and improve the network of fixed health facilities for rhodesiense HAT (e.g. in Zimbabwe [27]). In this analysis, financial barriers are not considered for both gambiense and rhodesiense HAT, but it is important to underline that in many countries, it can be an important barrier that prevent a wider use of the existing facilities.

Methodological challenges in monitoring HAT elimination
In this paper, efforts were made to present a comprehensive update of the indicators of elimination, as set by WHO [8]. We show that, in the framework of the Atlas of HAT, data were assembled and methodologies were developed that enable us to follow the progress of HAT elimination with a high level of geographic detail (village-level mapping), and completeness of data in time and space (comprehensive data from all reporting countries are systematically collated, harmonized and analyzed from the year 2000 onwards). However, challenges in monitoring the process of elimination still remain.
Regarding the primary global indicators of elimination, while the cumulative number of HAT reported cases has been effectively followed by WHO for a long time, monitoring the second primary indicator, i.e. 'the number of foci where HAT is no longer a public health problem' [8], is proving more challenging. The main reason for this is that the available definition of focus as 'a zone of transmission to which a geographical name is given (locality, region, or river)' [31] is useful for operational purposes, but it is vague and not particularly helpful for measuring. In particular, it is difficult to define the geographical boundaries of HAT foci in an objective and standardized way. The challenge is compounded by our incomplete understanding of the focal nature of HAT [16], and by the fact that different countries use different criteria to define foci. On the other hand, the data collected in the HAT Atlas allow to quantify the area at risk [13,14], which is, in essence, a 'zone of transmission' that can be measured in a robust and objective way. As such, it represents a much more suitable metric for the second primary indicator of HAT elimination, originally proposed as 'number of foci reporting less than 1 case per 10,000 inhabitants'. This metric also easily lends itself to monitoring over time, and at various spatial scales (from global to subnational and local levels). A recently created WHO HAT elimination Technical Advisory Group (HAT-e-TAG) recognized the impossibility of enumerating and delineating HAT foci objectively, and endorsed the revised global metric to assess elimination as a public health problem (i.e. the 'total area at risk reporting ! 1 case /10,000 people/year', which corresponds to the risk categories of 'moderate', 'high' and 'very high'). HAT-e-TAG also proposed the 2020 target for this indicator, i.e. a reduction of 90% by the year 2020 as compared to the baseline calculated for the period 2000-2004.
Regarding the secondary indicators of elimination, they currently include population at risk, coverage of active and passive screening activities, and the geographic distribution of the disease [8].
For population at risk, the available data and methodologies enable an effective monitoring. Provided that attention is paid to the various levels of risk, the population at risk provides useful complementary information to the primary indicators.
Regarding the coverage of passive surveillance, at this stage we are in a position to estimate only a potential coverage (i.e. physical accessibility), and there is still a two-year lag between the survey of health facilities and the risk map used for stratification. In the future, efforts will be made to shorten the lag, and, much more importantly, to estimate the actual coverage of passive surveillance from the number of individuals passively screened by the health facilities. As to active screening activities, data are already systematically included in the Atlas of HAT that will enable the actual coverage to be estimated and mapped. To this end, a methodology is presently being developed. It is worth pointing out that the present methodologies to estimate coverage fail to capture issues of quality of coverage, such as what age, sex, or occupational groups are covered, quality and performance of the services provided and the varying efficacy of detection methods used.
As regards geographic distribution, the last secondary indicator, it is still considered as a very useful aspect of the epidemiology of HAT to be monitored. However, measuring this indicator quantitatively is not deemed particularly relevant at this stage, especially because the indicator 'area at risk' already captures the main quantitative aspects of the geographic distribution of HAT.
One cross-cutting aspect that affects virtually all indicators is their reliance on reported HAT cases, with all the uncertainties that an unknown level of underdiagnosis and underreporting brings. Efforts are being made to estimate and map these uncertainties through geospatial and environmental modelling [32,33].

Conclusions
The advances in the process of HAT elimination [13] are confirmed in this new comprehensive report for gambiense and rhodesiense HAT. In particular, the milestone for the number of HAT cases reported in 2014 was achieved and even surpassed. Case-finding efforts were sustained in most of the affected countries, which gives confidence in a real progress in disease elimination. These results were accomplished mainly through sustained efforts in disease surveillance and control by NSSCPs.
The strength of the epidemiological knowledge continues to improve. The database of the Atlas of HAT is regularly improved in terms of completeness and accuracy, thus resulting in a robust estimation of the indicators. At the same time, models trying to predict the level of underdetection and the presence or absence of the disease in grey areas are being developed.
In a few affected areas, access to diagnosis and treatment is still constrained by insecurity (e.g. in Central Africa Republic and South Sudan) and remoteness (e.g. in some area in the DRC). Also, the progressive loss of expertise and motivation of health staff dealing with HAT is one of the inevitable effects of the reduction of cases. New innovative approaches are required to sustain the quality of interventions. Looking to the future, another inevitable consequence of reduced number of cases will be a progressive shift from active screening to a combination of passive surveillance and reactive screening. This shift, and the related integration of gambiense HAT surveillance into the health system, will be one of the main challenges to elimination.
Rhodesiense HAT represents a relatively small part of the global HAT problem. Because of its zoonotic dimension, the approach to tackle rhodesiense HAT must consider the epidemiological role of the domestic and wild animal reservoir in a One Health framework. As a result, disease elimination will require a multisectoral approach that should involve the veterinary services and include a vector control component [34,35]. As to the interruption of transmission, it is likely to remain elusive for some time to come, unless a breakthrough in control tools enables to tackle the animal reservoir (especially its wildlife compartment).
Despite the recent advances, it is crucially important to sustain the commitment of all stakeholders. Appropriate funding must be ensured if the 2020 and the 2030 goals are to be achieved. It is likely that both targets can be met, although the latter (i.e. interruption of transmission) is expected to pose a more severe challenge, and it is only applicable to gambiense HAT [8]. In the process of elimination, increased ownership of the fight against HAT by endemic countries must be ensured. The challenges to integration of HAT activities in weak national health systems raise concerns. All efforts and policies aiming to strengthen health systems, especially in rural areas, will contribute to the sustainability of HAT elimination.
Looking at gambiense HAT, in this new context of strongly reduced prevalence, human asymptomatic carriers [36] and the possible animal reservoir [37,38] need to be studied in more detail, as they could play a role in disease maintenance, resurgence and reintroduction. Development of new control tools, including diagnosis, treatment and vector control, could change the current control and surveillance scenario by enabling innovative, adapted and more cost-effective strategies to be implemented.
While the process of HAT elimination is progressing as planned, many challenges still lie ahead. At this juncture, the WHO network for HAT elimination set up in 2014 [12] ensures crucial coordination of stakeholders and maximum effectiveness in the support to endemic countries. Only by maintaining the synergy and coordination of interventions will sustainable elimination of HAT be achieved. Disclaimers: The boundaries and names shown and the designations used on the maps presented in this paper do not imply the expression of any opinion whatsoever on the part of WHO and FAO concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. The views expressed in this paper are those of the authors and do not necessarily reflect the views of WHO and FAO.