Assessment of tuberculosis spatial hotspot areas in Antananarivo, Madagascar, by combining spatial analysis and genotyping

The study was carried out in Antananarivo, the capital of Madagascar (surface area = 90 km²; 1.1 million inhabitants). Antananarivo is divided into six districts and sub-divided into 192 “Fokontany” that are administrative neighborhoods. Population density per Fokontany and population size is highly heterogeneous, as in many large cities (Fig. 1a and b); the poorest neighborhoods were classified by average household income; the 25% of all neighborhoods with the lowest average income were assigned as “poorest neighborhoods”; the most populous and poorest neighborhoods are localized in the periphery (Fig. 1a and b) [22]. Some public markets are distributed in the densest zones (Fig. 1a). There is circulation of poorest and most vulnerable populations in these public markets. The first district is among the poorest and most populous in Antananarivo. Cases were diagnosed and treated in TB diagnosis and treatment centers (DTCs) indicated in Fig. 1c.

Fokontany were identified by conventional alphanumeric codes and were geolocalized by their centroid. Data of the population census of the 192 Fokontany were provided by the Development Office of Antananarivo (DOA) and poorest neighborhoods were described previously [22].

For each recruited patient, the objectives and nature of the study were explained by a DTC agent. Information written in Malagasy language was given to any patient or parents for children under 18 years. He/she was then asked if he/she agrees to participate in the study. Study subjects were allowed to refuse to participate in the study or withdraw from the study at any time without prior justification. When the subject agreed to participate in the study, he/she signed the informed consent form. An interview was conducted and the information was collected on a form. The residence address and Fokontany were recorded for every patient. The residence address provided by the patients in the questionnaire was confirmed by the DTC registers.

Fresh sputa were collected, stored, and cultured on Lowenstein Jensen (LJ) solid medium. Raw DNA was obtained from the clinical isolates by heating and killing a suspension of a culture colony in a hot dry bath at 80 °C for 30 min, After centrifugation at 13,000 * G for 5 min, supernatant was transferred in a new tube and conserved at −20 °C. Raw DNA was shipped at −20 °C to Orsay, France where the spoligotyping was performed directly using a Luminex 200 platform as described by Zhang [23]. The spoligotype profiles were identified by their Shared-type (SIT – Spoligotyping International Types) and lineage designation by using the SpolDB4/SITVITWEB rules and classification [4]. Moreover, to fit with current genome-based lineage designations, L1-L7 lineage labels were added to spoligotype families [24, 25]. Given the modest benefit of spoligotyping value when used alone in molecular epidemiological studies, we did not infer any recent transmission rate based on spoligotyping-based clustering, but used the genotyping information to correlate spatial and genetic clustering as a clue to define “Hot Spot TB area”.

TB patients were localized according to their Fokontany of residence. All data on the population denominators per Fokontany and metadata used for mapping were provided by the DOA. All TB patients and patients with genotypic clustered isolates (patients associated with isolates with repeated spoligotype or PRS) were scanned separately using the Kulldorff spatial scanning method. The populations per Fokontany with SaTScan™ were considered for the spatial analyses (http://​www.​satscan.​org) [26]. For Satscan® parameters, a maximum radius of 1 km was defined for the spatial scanning. A p-value <0.05 determined by conducting Monte Carlo replications was considered to be statistically significant. We assumed that the number of TB patients in each Fokontany fits with the Poisson distribution. The identified Fokontany associated with spatial clusters representing TB focal points of epidemics were mapped using QuantumGis 2.8® (QGIS Development Team, 2013). Areas representing focal points of epidemics with highly genetically related isolates (clustered by spoligotyping) were considered to be potential TB hotspot transmission areas.

A total of 523 TB patients were recruited from 14 of the 20 DTCs included in the study. The recruited patients were localized to 142 of the 192 Fokontany of Antananarivo (1–15 patients/Fokontany). Fifty-six patients were excluded (the addresses of 46 patients were confirmed outside the study area and residence’s Fokontany of 10 patients could not be identified). Among the 467 remaining included patients, 427 (91.4%) had positive bacterial cultures identified from the LJ-growth media (Fig. 2).

We obtained 88 individual spoligotype profiles from 394 clinical isolates (84.4% of the 467 recruited patients). We could not type 73 (16%) clinical isolates as 40 cultures were negative and 33 spoligotypes were non-interpretable. Of the individual 88 spoligotype profiles, 47 were unique and 41 were clustered, totaling 347 clustered isolates, found in clusters of 2 to 39 clinical isolates (Additional file 1: Table S1).

Clinical isolates belonging to the Euro-American/T spoligotyping family (included in Lineage 4; 43.1%) predominated, followed by the EAI lineage (Lineage 1; 12.7%), and the CAS lineage (Lineage 3; 10.4%), whereas the Beijing lineage (Lineage 2), the LAM, and the H sublineages (both belonging to Lineage 4) did not exceed 10% of the studied clinical isolates (Fig. 3). The 14 most prevalent SITs constituted 70.0% of all spoligotypes (276/394). Most prevalent SITs (more than 30 cases) were: SIT86 (T family), SIT1 (Beijing lineage), SIT156 (T family), SIT109 (EAI lineage), SIT21 (CAS lineage) (Additional file 1: Table S1).

Residences of the included patients (n = 467) were distributed in 133 of the 192 Fokontany (Fig. 1c). Spatial observations showed that the highest number of TB cases was observed in Fokontany Andohatapenaka II in the 1st district (n = 15) whereas the district with the highest number of TB cases was the 5th district (112/467; 24.0%). The relatively most populous Fokontany at the periphery of Antananarivo had the higher TB cases number compared to the center of the city (3rd district) (Fig. 1b, c).

With the spatial scanning method of Kulldorff, we observed one significant spatial cluster of TB cases (p = 0.039) formed by 6 Fokontany (localized in the 1st and the 6th neighboring districts) (Fig. 4a). The 394 patients with genotyped clinical isolates were distributed in 133 Fokontany, The patients with genotypic clustered isolates (n = 347) were distributed in 127 of these 133 Fokontany (1 to 13 isolates per Fokontany); and when scanning these 347 patients, four significant spatial clusters were observed (Fig. 4a): a first cluster formed by one Fokontany (Antohomadinika Afovoany) in the 1st District (p = 0.001), a second formed by eight Fokontany of the 4th District (p = 0.001), a third cluster in one Fokontany (Andohatapenaka II) in the 1st district (p = 0.002) which overlapped with the spatial cluster of all TB cases (Fig. 4a), and a 4th cluster formed by three Fokontany in the 2nd district (p = 0.043). These spatial clusters represent potential areas of TB transmission. The spoligotype diversity in each spatial clustering is presented in Fig. 4b. The first area with potential TB transmission included two SIT21 isolates. There were seven SIT156 isolates in the second area, followed by SIT21 (n = 5) and SIT109 (n = 5). SIT156 (n = 4) and SIT109 (n = 2) predominated in the 3rd area with potential TB transmission and SIT78 (n = 3) and SIT59 (n = 3) predominated in the 4th area (Fig. 4b). These TB hotspots could not be detected on the spatial distribution of the 11 lineages (Fig. 3b).