Article Text
Abstract
Objective Estimates of the prevalence of rheumatic heart disease (RHD) in many endemic countries are limited to samples of children attending schools, which generate an incomplete picture of disease burden in communities. The present study conducted household-based RHD screening in a representative community in Gulu district, Uganda.
Methods Members of households identified through a two-stage cluster-sampling approach between the ages of 5 years and 50 years were invited to undergo limited cardiac testing with a handheld echocardiogram to assess for the presence of RHD. Suspicious cases underwent confirmatory echocardiogram with a fully functional machine.
Results Of the 2453 community members screened, 2.45% (95% CI 1.87% to 3.14%) showed echocardiographic evidence of RHD with 1.26% (95% CI 0.860% to 1.79%) having definite RHD. The overall prevalence of RHD among participants <20 years was 2.52% (95% CI 1.78% to 3.45%), with a borderline prevalence of 1.97% (95% CI 1.33% to 2.82%) and a definite prevalence of 0.544% (95% CI 0.235% to 1.07%). Prevalence rates among youth increased with age and peaked in the age group of 16–20 years. The overall adult prevalence (>20 years) of RHD was 2.34% (95% CI 1.49% to 3.49%). The majority of definite cases were mild (81%) and marked by mitral regurgitation and associated morphological valve changes (71%).
Conclusion Our data reveal a high prevalence of undiagnosed RHD within an endemic community and fill a critical gap in RHD epidemiology in African adults.
- valvular heart disease
- global disease patterns
Statistics from Altmetric.com
Introduction
Rheumatic heart disease (RHD) remains endemic in sub-Saharan Africa and is the leading cause of cardiovascular morbidity and mortality among young people.1 The disease results from cumulative exposure to group A streptococcus and subsequent episodes of acute rheumatic fever (ARF), causing cardiac inflammation, scarring and valvular dysfunction. The Global Burden of Disease 2015 Study estimated that 33 million people are currently affected, resulting in 320 000 deaths annually.2 However, these aggregate estimates of RHD prevalence have a high degree of uncertainty due to heterogeneity in primary studies, many of which have been conducted using different methodology in non-representative populations.2 Major limitations of the literature include a focus on school children and lack of prevalence data among adults.3 4
In Uganda, no primary data are available on fatal RHD and limited data on prevalent cases stem from two screening studies, demonstrating a prevalence between 2% and 4%.5 6 These data were limited to school children aged 5–15 years old, an age group thought to represent only 15%–20% of RHD cases within an endemic population.7 Although school-based data have been supplemented by individual tertiary hospital reports,8 these data are not useful for estimating prevalence as they represent only a fraction of cases, usually the most severe.8
There is a critical gap in knowledge of disease epidemiology between echo-detected RHD of childhood and symptomatic RHD presenting with advanced disease. Obtaining epidemiological data to fill this void has been highlighted as a priority in RHD research and a critical guide to policy decisions.9 10 We hypothesised that a significant share of RHD cases in Uganda lie unknown in the community. The present study sought to estimate the prevalence of RHD across the lifespan in Gulu, Uganda, where we previously demonstrated a high prevalence in school children.6
Methods
Study area
Gulu district (figure 1) is located in Northern Uganda, 320 km from the Uganda Heart Institute, the national referral centre for cardiovascular care. The district is divided into 16 subcounties, 70 parishes and 780 enumeration areas, which are home to an estimated 275 000 people from 55 000 households. Despite economic growth following the end of a civil war in 2006, the area remains poor, with only 20% of inhabitants having access to electricity and over half relying on subsistence farming for livelihood.11
Sample size
To obtain a representative sample of the population in Gulu, we used a two-stage cluster-sampling approach.12 For practical and logistical reasons, we opted for 16 clusters (4 weeks of sampling 4 days a week, with one cluster screened per day). Using a conservative prevalence estimate of 2.0%, based on prevalence studies in the region,6 we calculated a sample size of 2100 for a two-sided alpha of 0.05 and a precision of ±0.6%. With an average of five eligible family members per household an anticipated response rate of 80%, 480 household would achieve enrolment, with a resulting cluster size of 30 households.
Sample selection
The sample was selected at two stages: (1) 16 enumeration areas were drawn from Gulu with probability proportional to size and (2) in the second stage, 30 households were drawn using simple random sampling. A new household listing was undertaken by staff from the Uganda Bureau of Statistics for each selected enumeration area prior to the start of the project. Households were notified of selection 1 week before screening day by their village chairman. Eligible household members who were not present on screening day were invited to Gulu Regional Referral Hospital on the following Saturday to participate. If a household refused, or no eligible members were present on screening day, the first house to right of the originally sampled house was approached to participate. If no household could be identified after three tries, the study team moved to their next assigned house. Households were paid in local currency the equivalent of US$5.
A household was defined as people living and sharing meals together at least four times a week for the last month. The investigators obtained written informed consent from all participants and guardians, as well as written assent from children 8–17 years.
Data collection
Demographic data were obtained from each participant with the use of a questionnaire. Health professionals (cardiologists, general physicians and nurses) trained in echocardiography performed a limited echocardiogram using a handheld device (Philips Lumify, Best, Holland) on each participant. Two-dimensional and colour Doppler images were obtained in the parasternal long axis view, parasternal short axis view and 4 and 5 chamber apical views and stored in DICOM format. All images were interpreted at the end of each screening day by a cardiologist with expertise in the World Heart Federation (WHF) criteria (SC, JS and MTD). Results were delivered to participating households by the village chairman, indicating either an abnormal or normal screening. If the reviewing cardiologist deemed the screening images inadequate to assess for RHD, the participant was asked to attend the follow-up clinic for repeat imaging. Any participant with an abnormality on screening echocardiogram, including suspicion of RHD, congenital heart disease, pericardial effusion or abnormal function, was asked to attend a follow-up clinic at Gulu Regional Referral Hospital staffed by a cardiologist from the Uganda Heart Institute (TA, EO and IS) for a repeat echocardiogram using a standard portable machine (Vivid Q, General Electric, Milwaukee, Wisconsin, USA) for confirmatory diagnosis. Following confirmation of RHD at follow-up, all patients were linked to care at the existing cardiac clinic at Gulu Regional Referral Hospital. All patients with RHD were invited to participate in the Uganda National RHD Registry, a previously approved Registry established in Gulu in 2015 supported by ongoing research initiatives, the Uganda Heart Institute and outside funding.
DICOM images were later blindly reviewed by a cardiologist with expertise in the 2012 WHF criteria (CS). Any disagreement was adjudicated by a third cardiologist (AB). Participants were classified as normal, borderline or definite RHD (using WHF 2012 criteria,13 box 1), congenital heart disease or other. Left ventricular size was assessed qualitatively, as well as a colour Doppler of the tricuspid valve in the apical four chamber view, for all subjects who returned for follow-up. Continuous wave spectral Doppler of the tricuspid valve was performed in subjects with visible tricuspid regurgitation on colour Doppler, and peak velocity was recorded when measurable.
World Heart Federation diagnostic criteria for rheumatic heart disease (RHD)
Echocardiographic criteria for individuals aged >20 years
Definite RHD (either A, B, C or D):
Pathological Mitral Regurgitation and at least two morphological features of RHD of the Mitral Valve.
Mitral Stenosis mean gradient ≥4 mm Hg.
Pathological Aortic Regurgitation and at least two morphological features of RHD of the Aortic Valve, only in individuals aged <35 years.
Pathological Aortic Regurgitation and at least two morphological features of RHD of the Mitral Valve.
Echocardiographic criteria for individuals aged <20 years
Definite RHD (either A, B, C or D):
Pathological Mitral Regurgiation and at least two morphological features of RHD of the Mitral Valve.
Mitral Stenosis mean gradient ≥4 mm Hg.
Pathological Aortic Regurgitation and at least two morphological features of RHD of the Aortic Valve.
Borderline disease of both the Aortic Valve and Mitral Valve.
Borderline RHD (either A, B or C):
At least two morphological features of RHD of the MV without pathological MR or MS.
Pathological MR.
Pathological AR.
Statistical analysis
Frequencies are reported as absolute numbers and percentages; prevalence estimates and demographic parameters are reported as proportions with associated 95% CIs and as means with SD where appropriate. P values are based on the χ2 test for the comparison of proportions and unpaired two-tailed t-test for comparison of means. All statistical analysis was performed with STATA V.15.0.
Results
Study population
Of the 480 households selected to participate, no members were available in eight households and five households refused. Replacement households were successfully added in all cases per study protocol. Within selected households, there were a total of 2949 eligible members. The overall mean household size was 7.8±3.9, (range 1–29) with a mean of 6.1±3.1 (range 1–20) members per household of eligible age (table 1).
Of the eligible household members, 2453 (83%) individuals elected to participate. Four hundred and eight-five (16%) individuals were not screened due to unavailability and 11 (0.4%) refused. The mean age of screened participants was 20.0 (±11.9) and 55% of participants were female. The population unable to be screened did not differ in age 20.5 (±10.6, p=0.33) but had significantly less females (41%, p<0.001).
Screening and Follow-up
We performed 2453 screening echocardiograms. Suspicious screening echocardiograms were found in 163 subjects who were invited for follow-up. Of these, 154 (94%) returned for a complete echocardiogram with a final diagnosis of normal in 76 (47%), RHD in 56 (36%), congenital heart disease in 7 (4.5%) and other heart disease in 15 (9.5%). The screening echocardiogram was reviewed for nine participants that failed to return for follow-up (final diagnosis of normal in five and RHD in four). Additionally, screening images from five participants were deemed inadequate by the reviewing cardiologist and invited to follow-up clinic for a repeat echo. In all five cases, the echo was normal (figure 2).
- Download figure
- Open in new tab
- Download powerpoint
Prevalence, severity and type of RHD
The community prevalence of RHD was 2.45% (95% CI 1.87% to 3.14%, 60% female). If limited only to definite RHD, the overall prevalence was 1.26% (95% CI 0.860% to 1.79%). The overall prevalence of RHD among children ≤20 years was 2.52% (95% CI 1.78% to 3.45%, 55% female), with a borderline prevalence of 1.97% (95% CI 1.33% to 2.82%) and a definite prevalence of 0.544% (95% CI 0.235% to 1.07%) (figure 3). The prevalence of RHD did not differ significantly by gender in any subgroup.
Of the 480 households surveyed, 424 (88% of households) had no cases of RHD, 52 (11%) had one case and 4 (1%) had two or more cases. On average, participants with RHD came from more populated homes (9.4 vs 7.6 members, p=0.002) and lived in rural areas (58% vs 43%, p=0.04). There were no other notable differences in these populations (table 1).
In order to evaluate the potential impact of clustered sampling on prevalence estimates and power, post hoc analyses were conducted to assess correlations within cluster. Intraclass correlations (ICCs) were found to be low for both region (ICC=0.005) and family (ICC=0.038), resulting in a modest design effect of 1.15. Using this design effect, the CI associated with the overall community prevalence was 2.45% (95% CI 1.73% to 3.16%).
A wide spectrum of disease severity was detected, with the majority of cases being mild. Among the eight children with definite RHD, two had severe findings, while the remaining six were mild (table 2). Similarly, among the 23 adults with RHD, 19 were mild, 1 moderate and 3 severe (table 2). The majority of moderate and severe cases of RHD were aged 21–30 years and five of the six cases were from rural areas. All participants with moderate to severe RHD presented with a history of symptoms: six (100%) with both chest pain and fatigue, five (83%) with palpitations, two (33%) with dyspnoea and one (17%) with leg oedema.
The distribution and pattern of valvular involvement was less diverse. Among both children and adults, the majority of RHD was isolated to pathological mitral regurgitation with or without (borderline) associated morphological mitral valve abnormalities (86% of children, 87% of adults). Adult participants showed more diversity, with three cases of mitral stenosis in a pattern of mixed-mitral valve disease, as well as mixed aortic and mitral disease.
Left ventricular dilation was present in five (four mild, one severe) subjects with definite RHD. Measurable tricuspid regurgitation was present in two subjects with definite RHD with a >3.0 m/s, indicating pulmonary hypertension.
A known history of ARF was rare and suspected in only two paediatric participants with severe RHD. One participant reported seeking care for chest pain, fever and palpitations at a private hospital in March 2015. They were treated for suspected pneumonia and discharged; however, the patient’s chest pain remained constant for 2 years. The second participant described repeatedly seeking care at a local health centre for joint pain, fever and palpitations 6 months prior to screening. The child was given topical ointment for the joint pain, which continued for 3 months. Neither of these children was diagnosed with ARF nor referred to a cardiologist. Subject’s prior awareness of their RHD was also rare; we only found one (paediatric) participant who had been diagnosed with mild definite disease in a previous school screening and had been prescribed penicillin prophylaxis.
Discussion
Following on a landmark paper published in 2007 on the prevalence of RHD in Mozambique and Cambodia,14 scores of studies have been published on the prevalence of RHD in low-income and middle-income countries. Rothenbühler and colleagues15 systematically reviewed this literature in 2014, but the number of studies continues to grow. Most studies have included children aged 5–19 attending schools, predominately in urban areas. Very few prevalence studies have been conducted among adults, who are expected—on the basis of population demographics and disease natural history—to comprise the majority of cases. Of the studies that have been conducted among adults, none have looked at participants beyond 35 years of age, even though our data suggest that the highest prevalence may be found in members 41–50 years of age, and none have assessed prevalence across the lifespan in a single population. Furthermore, the literature’s focus on children in schools potentially leads to biased estimates of RHD, which emerging evidence suggests could be more common among children too ill or poor to attend school.4 The present study thus advances the literature on RHD prevalence in three several important dimensions: use of survey sampling techniques to ensure representative populations, focus on communities rather than schools and inclusion of adults and children within the same population.
Our data suggest that the recent global burden of disease estimate may underestimate the true burden of RHD in Uganda. Even limiting included cases to those with definite disease, our data suggest 13 per 1000 persons ages 5–50 years are affected compared with 11 per 1000 persons across all age groups. The current global burden of disease estimate does not include subclinical cases of RHD, even though these cases can progress to clinical disease.16 A recent study conducted in Ugandan children found that nearly 50% of moderate to severe RHD cases, 26% of mild definite cases and 9.8% of borderline cases progressed over the 2-year study period.17 Risk of progression in adults with subclinical disease remains unknown and warrants further investigation. Additionally, the prevalence of definite RHD in adults used in the modelling was lower than the prevalence in adults discovered in this study and peaked later, further highlighting the need for additional data to understand disease prevalence and distribution patterns to inform to this estimate.
As was shown in a community-based screening in Ethiopia,4 the prevalence of RHD increased with age across the paediatric population, peaking in those between 16 years and 20 years. In this age group, 1.34% (95% CI 0.59% to 3.5%) of those screened had definite RHD and 4.84% had either definite or borderline RHD (95% CI 2.89% to 7.53%). Practical and logistical reasons have led most school-based RHD prevalence studies, including those in Uganda, South Africa and Ethiopia,5 6 18 to exclude children >16 years. However, our data suggest this age cohort may have the highest burden of disease, necessitating creative, community-engaged solutions to capture adolescents in screening programmes. This is additionally supported by findings from a study in New Caledonia, demonstrating that the prevalence of RHD nearly doubled in participants who were 16–18 years of age, compared with patients aged 5–15 years.19
Not surprisingly, the distribution of valve disease in this study is similar to the pattern of clinical RHD presenting to hospitals in low-income countries.20 Community RHD was primarily characterised by pure mitral regurgitation until age 20 years, followed by an increase of mixed mitral valve disease and mitral and aortic valve disease. However, compared with hospital-based registries, community data revealed a higher percentage of patients remaining with pure mitral valve disease, most commonly mitral regurgitation, with less aortic valve involvement and valvular stenosis. We speculate that this community pattern is likely driven by both symptoms (less) and survival (more) in patients with isolated mitral regurgitation.17 Mitral stenosis has high case-fatality in Uganda and could be underdetected in our study as a result.21 Additionally, all but one case of RHD was mild in participants over 30 years of age. Given that the mean age of death in patients presenting with clinical RHD to the National Referral Hospital in Uganda is 29 years21 and 50% of moderate to severe cases in this study were discovered in participants between 21 years and 30 years, it is possible that persons who do not progress beyond mild disease after a certain age are not at substantial risk for progression to clinical disease, though this is speculative. Future studies are needed to capture risk of progression, as well as elucidate protective and predictive factors in adult populations.
Lastly, only one participant with RHD had been previously diagnosed and two participants gave convincing histories of ARF. While not our primary aim, these findings suggest that recognition of ARF and RHD by healthcare providers in the community is poor.
This survey has several limitations. First, while we had a high response rate and believe the findings to be generalisable to this region of Uganda, the findings may not be generalisable to the rest of the country. Additionally, these data may not accurately assess prevalence in males as there were significantly more males absent from screening. Countrywide, regional and continental data are needed. If broader screening were to be conducted throughout Africa, ethical questions remain for countries where screened cases do not have access to tertiary cardiac services.
Second, the prevalence in participants <20 years for borderline and definite RHD was lower than previous reports in this region.6 While possible, we do not believe access to primary healthcare or average socioeconomic status (SES) has substantially modulated RHD risk in this region between 2014 and 2017. More likely, school-going children, which represent nearly half of household members who not included, were not fully represented. School enrolment and type of schooling (boarding/day) was not captured, but it is a reasonable assumption that children who were not present were in school, many of whom may have been away boarding. While some have suggested non-school-going children may be at higher risk due to lower average SES,22 it is also possible that school-going children, in particular boarding children, may be at higher risk due to overcrowding, peer exposure and higher streptococcal transmission.5 Further study is needed to reach definitive conclusions.
Finally, as these data were originally intended to serve as pilot data for a larger national survey, the initial sample size estimate did not incorporate an intracluster correlation. Given the slightly higher than anticipated response rate, higher mean number of eligible persons per household and post hoc analyses resulting in a design effect of 1.15 and low ICC for both household and enumeration area overall, it is unlikely that the clustered sampling had a material impact on study estimates.
Conclusions
This study fills a critical knowledge gap in the contemporary epidemiology of RHD. We demonstrate a high prevalence of asymptomatic RHD in children and adults in community settings, underscoring the need for future studies to look at RHD in groups other than school children. Our use of survey methods ensures a representative, generalisable population sample that can produce a gold standard set of prevalence estimates for Uganda and could be replicated in other endemic countries. Our data confirm that there is a wide spectrum of undiagnosed RHD within endemic communities in Uganda, advancing the epidemiological research agenda and raising new questions and challenges for the diagnosis and management of RHD in adults living in limited-resource settings.
Key messages
What is already known on this subject?
Rheumatic heart disease (RHD) affects roughly 33 million people worldwide. Echocardiography is an effective screening tool for case detection in endemic areas.
What might this study add?
This study provides the first community RHD data from Uganda and is the only study to determine prevalence for both children and adults (up to age 50 years), furthering epidemiological understanding through primary source data collection in a single population.
How might this impact on clinical practice?
Our findings suggest that there is a large paediatric and adult burden of latent, undiagnosed RHD in the community, under-represented by the 2017 Global Burden of Disease estimates. These data suggest that active case detection could play a critical role in improving the current paradigm: RHD only coming to clinical attention when patients have advanced disease and are symptomatic.
Further research is needed to understand the likely age-dependent risk to persons with mild RHD and if initiation of secondary prophylaxis improves outcomes.
Acknowledgments
The authors would like to thank Jennipher Kamarembo, Rose Akech and Francis Odong for their logistical support and dedication to this project.
References
Footnotes
Contributors All authors contributed to study design and oversight of data collection. AS, CS, and AB analysed and interpreted the data and drafted the manuscript. All authors critically revised the manuscript for important intellectual content. AB, EO, and PL supervised the study.
Funding This study was supported with a grant from the Karp Family Foundation and Children’s National Global Health Initiative.
Disclaimer The funders did not contribute to the design of the study. The funders did not take part in study design, data collection or writing of this manuscript.
Competing interests None declared.
Patient consent Not required.
Ethics approval The Makerere University College of Health Sciences Research and Ethics Committee, The Ugandan Bureau of Science and Technology and the Children’s National Health System Institutional Review Board approved the research protocol.
Provenance and peer review Not commissioned; externally peer reviewed.
Correction notice Since this paper was first published online, the affiliations 3 and 4 have been corrected.