Background: Podoconiosis is a type of tropical lymphoedema that causes massive swelling of the lower limbs. The disease is associated with both economic insecurity, due to long-term morbidity-related loss of productivity, and intense social stigma. Reliable and detailed data on the prevalence and distribution of podoconiosis are scarce. We aimed to fill this data gap by doing a nationwide community-based study to estimate the number of cases throughout Rwanda. Methods: We did a population-based cross-sectional survey to determine the national prevalence of podoconiosis. A podoconiosis case was defined as a person with bilateral, asymmetrical lymphoedema of the lower limb present for more than 1 year, who tested negative for Wuchereria bancrofti antigen (determined by Filariasis Test Strip) and specific IgG4 (determined by Wb123 test), and had a history of any of the associated clinical signs and symptoms. All adults (aged ≥15 years) who resided in any of the 30 districts of Rwanda for 10 or more years were invited at the household level to participate. Participants were interviewed and given a physical examination before Filariasis Test Strip and Wb123 testing. We fitted a binomial mixed model combining the site-level podoconiosis prevalence with continuous environmental covariates to estimate prevalence at unsampled locations. We report estimates of cases by district combining our mean predicted prevalence and a contemporary gridded map of estimated population density. Findings: Between June 12, and July 28, 2017, 1 360 612 individuals—719 730 (53%) women and 640 882 (47%) men—were screened from 80 clusters in 30 districts across Rwanda. 1143 individuals with lymphoedema were identified, of whom 914 (80%) had confirmed podoconiosis, based on the standardised diagnostic algorithm. The overall prevalence of podoconiosis was 68·5 per 100 000 people (95% CI 41·0–109·7). Podoconiosis was found to be widespread in Rwanda. District-level prevalence ranged from 28·3 per 100 000 people (16·8–45·5, Nyarugenge, Kigali province) to 119·2 per 100 000 people (59·9–216·2, Nyamasheke, West province). Prevalence was highest in districts in the North and West provinces: Nyamasheke, Rusizi, Musanze, Nyabihu, Nyaruguru, Burera, and Rubavu. We estimate that 6429 (95% CI 3938–10 088) people live with podoconiosis across Rwanda. Interpretation: Despite relatively low prevalence, podoconiosis is widely distributed geographically throughout Rwanda. Many patients are likely to be undiagnosed and morbidity management is scarce. Targeted interventions through a well coordinated health system response are needed to manage those affected. Our findings should inform national level planning, monitoring, and implementation of interventions. Funding: Wellcome Trust.
We did a population-based cross-sectional survey on podoconiosis targeting all 30 districts in Rwanda. In July, 2017, trained community health workers did a census of the communities within the selected sectors (a sector being the smallest administrative unit in Rwanda). In November, 2017, expert clinical diagnostic teams verified all suspected cases of lymphoedema (listed by community health workers) within these sectors. Rwanda is a densely populated, landlocked country of about 26 000 km2 in central eastern Africa. The country has a population of 12·1 million people (as of 2018)13 and is administratively divided into five provinces, 30 districts, and 416 sectors.13 Important country indicators include maternal mortality of 210 per 100 000 livebirths and child mortality of 50 per 1000 livebirths.14 The average life expectancy is 66·6 years.13 All adults (aged ≥15 years) who lived in any of the 30 districts were included in the study. Exclusion criteria were terminally ill patients who could not respond to the interview and patients with a mental health condition that would make interviewing difficult and results unreliable. Ethical approval was obtained from the Rwanda National Ethics Committee and the Brighton and Sussex Medical School Research Governance and Ethics Committee, Brighton, UK. Written informed consent was obtained from all respondents, except for illiterate respondents who provided their thumbprint and a signature from a literate witness. Individuals younger than 18 years provided assent and a parent or guardian provided written consent. A summary of the protocol is available in the appendix. Between June 12, and July 28, 2017, 282 trained community health workers did a census of the communities within the selected sectors. They registered residence, sex, and age, and recorded the presence of leg swelling of any type. All individuals with swelling of one or both lower limbs were documented as suspected cases during an exhaustive house-to-house census and case listing. Community volunteers were provided with case definitions and pictures of podoconiosis-related morbidity to identify the suspected cases. Between Nov 19, and Dec 6, 2017, expert clinical diagnostic teams verified all suspected cases of lymphoedema (listed by community health workers) within the randomly selected sectors, on the basis of procedures previously used in Cameroon.15 Each team included four health workers, a medical doctor, a nurse, a laboratory technician, and a team leader. To ensure effective community engagement, members of the team were recruited from the targeted sectors. Questionnaires were translated into Kinyarwanda language and data were collected using the LINKS software package (version 1.4.2; Secure Data Kit, Atlanta, GA, USA) installed onto Android smartphones.16 Individual data on age, sex, education, occupation, place of residence, shoe wearing, and foot hygiene practices were recorded, as were household data on water, sanitation, and hygiene. Geographic coordinates from surveyed communities were taken using smartphones. Data collectors could not proceed to the next question without completing all required fields. We used automated skip patterns to maintain the quality of the data. Each patient with suspected podoconiosis underwent a complete physical examination in a private room at the nearest public health centre. We used data collection methods used in previous similar surveys done in other countries.15, 17 Key questions were age at onset of swelling, family member (living or dead) with history of leg swelling, type of swelling (ascending or descending), and self-reported chronic illness, such as heart disease, kidney disease, or diabetes. Ascending swelling refers to swelling that starts from the foot and progresses up the leg; descending swelling is swelling that starts from the upper leg or groin area and progresses down. Suspected patients were also asked about previous clinical diagnoses of known causes of lymphoedema (such as congenital disorders, leprosy, and postoperative lymphoedema) and about the co-occurrence of swelling in other parts of the body, such as the hands, face, and scrotum (hydrocele). All lymphoedema cases were screened for circulating Wuchereria bancrofti antigen, using Filariasis Test Strips, and circulating specific IgG4 antibody, using Wb123 tests. Screening was done by trained laboratory technicians according to the manufacturer’s instructions.18, 19 Test results with the individual’s unique ID number were recorded both on the card, and on each individual’s data questionnaire. Briefly, for the rapid test, we used positive and negative controls for quality control of test batches before starting the daily activity. The patient’s third or fourth finger was cleaned with 70% alcohol and punctured using a sterile lancet. The initial drop of blood was removed using a cotton swab, and sufficient fresh blood was obtained to fill a 75-μl capillary tube. The blood sampled was transferred from the capillary tube to the pad on a Filariasis Test Strip card. The result of each card was read at 10 min exactly. A positive result was two lines, and a negative result was a single line. The tests were repeated if the control line was not shown. For the detection of IgG4 antibodies against W bancrofti, 10-μl capillary blood was transferred from the capillary tube to the pad of a Wb123 card and four drops of assay diluent were dispensed vertically into the square assay diluent well. The result of each Wb123 card was read at 30 min. A positive result was two lines, and a negative result was a single line. The following covariates were used in our analysis in a gridded format (raster datasets): precipitation, day land surface temperature, elevation, enhanced vegetation index, distance from closest waterway, clay content, silt content, night light emissivity, and distance to stable night light (appendix). These factors are associated with podoconiosis and have been included in previous modelling studies.20 Details on the source and processing are available in the appendix. Input grids were resampled to a common spatial resolution of 1 km2 using the nearest neighbour approach and clipped to match the geographic extent of a map of Rwanda, and eventually aligned to the map. Raster manipulation and processing was done using the raster package in R v3.3.2 and final map layouts created with ArcGIS software (version 10.5; Esri, Redlands, CA, USA). Geographic coordinates of each community were used to extract estimates from the aforementioned covariates. We defined a podoconiosis case as a person residing in the surveyed district for at least 10 years who had bilateral, asymmetrical lymphoedema of the lower limb lasting for more than 1 year, negative Filariasis Test Strip (Alere; Scarborough, ME, USA) and Wb123 tests, and a history of any of the signs and symptoms associated with podoconiosis.17 We collected occupational data in precoded categories: government employee, non-government employee, subsistence farmer or fishing, self-employed, full-time student, at home doing housework, unemployed (but able to work), unemployed (unable to work), and retired. Collected data were screened every day by the field supervisors and immediate feedback was given to the enumerators; field team leaders gave overall feedback and supervision. Final assessment of the full database was done after data collection every day to identify inconsistencies and missing items. We used a system for grading the clinical stages of podoconiosis that had been developed and validated in Ethiopia previously.21 The system has five stages: stage 1, swelling reversible overnight (ie, the swelling is not present when the patient first gets up in the morning); stage 2, below-knee swelling that is not completely reversible overnight, with knobs or bumps below the ankle only (if present); stage 3, below-knee swelling that is not completely reversible overnight, with knobs or bumps above the ankle; stage 4, above-knee swelling that is not completely reversible overnight, with knobs or bumps at any location; and stage 5, swelling at any place in the foot or leg, and the ankle or toe joints become fixed and difficult to flex or dorsiflex—these symptoms can be accompanied by apparent shortening of the toes. The survey used a cluster sampling design. We included all 30 districts of Rwanda. In each district, at least two sectors were randomly selected. The survey was powered to generate prevalence estimates for podoconiosis with a precision of 0·003% at the national level, assuming a conservative prevalence estimate of 42 cases per 100 000 from a surveillance report (M Jean Bosco, Rwanda Biomedical Center–Ministry of Health, personal communication). Under a design effect of 1·6 (calculated from a mapping survey in Cameroon)15 and a community participation rate of 80%, we estimated a sample size of 986 376 individuals to be screened from 60 sectors (two sectors per district). To adjust for sector size, sectors were assigned to districts proportional to the number of sectors per district. The adjustment resulted in 80 of 416 sectors being selected, with independent selection in each district. In the selected sectors, a census of all eligible individuals was done. All individuals aged 15 years and older, who had lived in the area for at least 10 years before the survey (to exclude individuals who might have acquired lymphoedema elsewhere) were included. We described the individual characteristics and prevalence of lymphoedema and podoconiosis with 95% CIs. We estimated district-level prevalence and number of cases across Rwanda using a binomial mixed model, accounting for fixed effects (covariates) and random effects. We chose this method over geostatistical and Bayesian frameworks on the basis of the absence of spatial structure on podoconiosis prevalence, explored by fitting an empirical variogram (appendix). Spatial dependence is a prerequisite for such methods. Our model used the prevalence estimates and covariates for smoothing the prediction. Briefly, let Yi denote the random variable associated with the number of positively detected cases of podoconiosis at a community location xi. We then modelled Yi using a binomial mixed model with probability of having podoconiosis p(xi) such that: where the dj(xi) are georeferenced covariates and Zi are independent and identically distributed zero-mean Gaussian variables with variance σ2. We fitted the model using the lme4 package (version 3.4) in R software.22 Using the approach described by Bates and colleagues,23 we tested the presence of residual spatial correlation by generating the 95% CIs for the variogram of the random effects Zi under the assumption of spatial independence. As the variogram that was based on the estimated Zi using the original data fell within the 95% CIs, we concluded that there was no evidence of residual spatial correlation. Therefore, a binomial non-spatially explicit mixed model was constructed. This model was used to produce continuous predictions of podoconiosis prevalence at 1 km2 spatial resolutions. We also developed a probability map of exceeding 0·1% prevalence. For a given location x, we obtain where βj hat is the maximum likelihood estimate of the regression coefficient βj for j=0,1,…,9. To compute the probability of exceeding 0·1% prevalence, we used the multivariate Gaussian approximation of the maximum likelihood estimator.24 Gridded maps of population density and age structure were obtained from the WorldPop project.25, 26 We used this gridded population surface to compute the estimates of affected population by pixel, by multiplying prevalence of podoconiosis in 1 km2 area with the corresponding population at the same spatial resolution. We used this surface to extract the aggregate number of people with podoconiosis by district. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author and the last author had full access to all the data and had final responsibility for the decision to submit for publication.