Article Text
PDF
Abstract
Objective Digoxin is widely used in patients with rheumatic heart disease (RHD) despite a lack of data on its impact on clinical outcomes. We aimed to determine the association of digoxin use on clinical outcomes in patients with RHD.
Methods We performed a retrospective analysis of the association of digoxin use with mortality at 2 years in a large RHD registry. Secondary outcomes were recurrent heart failure (HF) and hospitalisation for any cause. We assessed associations using multivariable logistic regression in the entire cohort and in subgroups of patients with atrial fibrillation (AF) and HF. We also estimated average treatment effects from propensity-adjusted analyses using inverse probability treatment weighting.
Results Information on digoxin use at baseline was available for 98.7% (3298/3343) of patients. In the overall population, digoxin was significantly associated with mortality (OR 1.63, 95% CI 1.30 to 2.04, p<0.0001) and recurrent HF (OR 1.48, 95% CI 1.07 to 2.04, p=0.019). On propensity-weighted analyses, this effect was markedly attenuated (OR 1.05, 95% CI 1.01 to 1.09, p=0.005). Patients in sinus rhythm without HF had a higher propensity-adjusted odds of death with digoxin use (OR 1.06, 95% CI 1.01 to 1.12, p=0.015), but those with both AF and HF had lower mortality (OR 0.88, 95% CI 0.80 to 0.98, p=0.019).
Conclusion Digoxin use is associated with higher mortality in patients with RHD, but this is greatly attenuated on propensity adjustment, indicating the presence of substantial treatment bias. The adjusted estimates may therefore not be reliable, and large randomised trials are needed to determine the true effect of digoxin in patients with RHD.
- valvular heart disease
- heart failure
- atrial fibrillation
Statistics from Altmetric.com
Introduction
Rheumatic heart disease (RHD) is an important cause of morbidity and mortality in less developed countries. A recent Global Burden of Disease Study report estimates the current worldwide prevalence to be over 33 million cases and the number of deaths attributable to RHD to be nearly 320 000 annually.1 Endemic regions in low/middle-income countries account for three-quarters of the burden of disease and mortality. Mortality is higher in patients with atrial fibrillation (AF), heart failure (HF) and poor functional class. In the Global Rheumatic Heart Disease Registry (REMEDY), which recruited 3343 patients from 12 African countries, India and Yemen, over a fifth of patients were in AF, a third were in HF and nearly a quarter were in New York Heart Association (NYHA) class III or IV at presentation.2 Mortality in this population was 16.9% at 2 years.3 Markers of severe valve disease, including HF and poor functional class, were the strongest predictors of death.3 While the definitive treatment of patients with significant valve involvement is surgical or interventional correction of the valvular abnormality, many patients in low/middle-income countries do not receive timely surgery or balloon valvuloplasty.3 Patients therefore receive medical therapy for control of symptoms while awaiting definitive treatment.
Even though there are no recommendations for the use of digoxin in patients with RHD,4 5 it is commonly prescribed. In the REMEDY study, nearly 35% of patients were on treatment with digoxin.2 Clinicians use digoxin with the intent to control heart rate, both in patients with AF and also in those who are in sinus rhythm and have significant mitral stenosis, where a reduction in heart rate may translate into lower transvalvular pressure gradients and improvement in symptoms.6 Further, as in patients without valve disease, digoxin forms part of the treatment regimen of patients with RHD who are in HF. However, there are no large studies, either observational or randomised, evaluating the efficacy and safety of digoxin on clinical outcomes in this population. Among patients in HF without valve disease, one large trial indicated that digoxin had a neutral effect on mortality but significantly reduced hospitalisations for HF.7 However, more recent data from large observational studies suggest that digoxin use is associated with increased mortality.8 9 Patients with non-valvular AF are typically elderly and have a high prevalence of coronary artery disease and may be at greater risk of adverse effects due to digoxin (such as life-threatening ventricular arrhythmia). On the other hand, patients with RHD are young (median age 28 years in REMEDY), and have a lower prevalence of hypertension, diabetes and atherosclerotic cardiovascular disease, and so the effects of digoxin may be different in this population. We performed a retrospective analysis of observational data on digoxin use in the REMEDY study in order to understand its association with mortality and recurrent HF among patients with RHD.
Methods
Study design, participants and outcomes
The design, baseline characteristics and 2-year outcomes of the REMEDY study have been published previously.2 3 10 Briefly, REMEDY was a prospective, multicentre, international, hospital-based registry which enrolled 3343 patients with a clinical diagnosis of RHD, who were seen at 25 hospitals in 14 countries (12 African countries, Yemen and India). The key outcomes of interest were death, HF, stroke, transient ischaemic attack or non-central nervous system systemic embolism at 2 years.
For the purposes of this analysis, we considered patients to be digoxin users if they were on the drug at baseline. The principal outcome was mortality at 2 years. Other outcomes studied were recurrent HF, all hospitalisations and the incidence of valve surgery or percutaneous interventions for valve disease. We also assessed the effect of digoxin use on the composite outcomes of death or recurrent HF, and death, recurrent HF or hospitalisation. The severity of valve lesions and other echocardiographic variables were described using standard criteria.2 Patients were considered to have severe valve disease if at least one of the involved valves had severe disease. HF at baseline and during the study (recurrent HF) was diagnosed if any two of the following criteria were present: (1) symptoms (dyspnoea on exertion or at rest, orthopnoea,paroxysmal nocturnal dyspnoea or ankle oedema) or signs (rales, increased jugular venous pressure or ankle oedema) of congestive HF; (2) radiologic signs of pulmonary congestion; and (3) treatment with diuretics. Surgery for valve disease was defined as performance of valve repair, or replacement of any affected valve with a tissue, or mechanical prosthesis. Percutaneous valve interventions consisted of balloon dilatation of stenosed mitral, aortic or tricuspid valves. Patients who had at least one overnight admission to hospital were considered to have a hospitalisation event. Details and definitions of the baseline and outcome measures used have been published previously.2 3 10
Statistical methods
We compared the baseline characteristics of digoxin users and non-users using independent t-tests and Χ2 tests as appropriate. We performed univariable and multivariable logistic regression to assess the effect of digoxin on the following outcomes: death, recurrent HF, hospitalisation, surgical or percutaneous valve intervention and on the composites of death or recurrent HF, and death, recurrent HF or hospitalisation. The variables included in the multivariable model were decided a priori based on their relevance to prognosis and were identical to the ones used previously in our primary analysis of outcomes.3 They were: age, sex, presence of AF or atrial flutter, NYHA class, history of HF or HF at enrolment, previous heart valve surgery or intervention, a history of stroke or IE, severe disease (severe valve disease in at least one affected valve), multivalve involvement and use of secondary prophylaxis. We performed subgroup analyses to assess effect modification by the presence of AF or HF at baseline, on the association of digoxin on the principal outcomes. In addition, we tested for interactions between gender, body mass index (BMI) and concomitant beta-blocker use with digoxin, by adding appropriate interaction terms in the model. Finally, we also assessed the effect of digoxin use in the subpopulation of patients who had both AF and HF at baseline when compared with those who had neither.
Propensity-adjusted analysis
We generated propensity scores for each patient using logistic regression with digoxin as the dependent variable. The variables used in this model (listed in figure 2) were selected based on their potential association with digoxin use and clinical outcomes, in accordance with expert recommendations.11–14 After generating the propensity scores we tested for similarity of distribution, adequate overlap of the distributions (common support) (online supplementary figure S1) and ensured balance of covariates between the two groups (digoxin and no digoxin) within blocks of propensity scores. After creating a balanced propensity score, we used a weighting method (inverse probability treatment weights, IPTW) and a matching method (1:2 calliper matching with replacement, with calliper width set at 0.2×SD of the logit of the propensity score) to ensure balance of covariates between the two groups.14 Balance was assessed by the standardised percentage of bias for each covariate. An overall mean standardised percentage of bias <5%, or a standardised mean difference <0.1, is considered to indicate good balance. We estimated average treatment effects of digoxin on outcomes using the IPTW as the primary analysis and corroborated the results using those obtained from 1:2 calliper matching. Further, if statistically significant subgroup effects were present, we generated separate propensity scores for subgroups of patients who were in AF, HF or both at baseline (variables listed in online supplementary table S2) and followed similar procedures to obtain effect estimates of digoxin use on outcomes.
Supplemental material
The effect of digoxin use on outcomes was estimated using ORs and their 95% CIs. A p value of <0.05 was considered statistically significant. All analyses were done using Stata V.14 (StataCorp, College Station, Texas) and the propensity analyses were performed using the pscore, psmatch2, dr and teffects packages.
Results
Of the 3343 patients in REMEDY, 3298 (98.7%) had information on baseline digoxin use, and 1144 (34.7%) of these were on digoxin. The proportion of patients lost to follow-up among those on and those not on digoxin was similar (figure 1). Patients who were on digoxin at baseline were more likely to reside in low-income countries, were older, in AF, HF and poor functional class, and were more likely to be on diuretics and ACE inhibitors (or angiotensin receptor blockers).(online supplementary table S1) Overall, 51% (1709/3343) of patients had severe valve disease. The prevalence of clinical and echocardiographic markers of severe valve involvement was higher among those prescribed digoxin (table 1). After propensity score weighting, we achieved excellent balance between the two groups as evidenced by the small (<5%) standardised mean difference between covariates (figure 2). The variables used to generate the propensity scores and the percentage bias reduction are depicted in figure 2. Balance of covariates for the 1:2 calliper matched analysis is shown in online supplementary figures S2 and S3.
Flow of patients contributing to the digoxin versus no digoxin analysis. Numbers are for patients using digoxin at the time of enrolment into the study. REMEDY, Global Rheumatic Heart Disease Registry.
Comparison of baseline characteristics of patients who completed 2-year follow-up
Standardised percentage of bias across covariates before and after inverse probability treatment weighting. Figure illustrates the balance of covariates between treatment and comparison groups before and after inverse probability weighting. Standardised mean differences of covariates after matching (indicated by red solid circles) are well within the prespecified ±5 margin indicating good comparability between the two groups. NYHA, New York Heart Association; P/H/O, Past history of stroke.
Effect of digoxin on clinical outcomes
All patients
Among all patients, digoxin use was significantly associated with mortality (OR 1.63, 95% CI 1.30 to 2.04, p<0.0001) and recurrent HF (OR 1.48, 95% CI 1.07 to 2.04, p=0.019) at 2 years (table 2). There was no impact of digoxin on hospitalisations or the performance of percutaneous or surgical interventions. The composite outcome of death or HF, and death, HF or hospitalisations was also significantly increased with digoxin use (table 2). The association of digoxin with mortality was not modified by low BMI, female gender or the concomitant use of beta-blockers (p values for interaction 0.14, 0.08 and 0.35, respectively).
Clinical outcomes with digoxin use
On propensity score weighted analyses, the effect of digoxin on mortality was markedly attenuated, though remaining nominally statistically significant (OR 1.05, 95% CI 1.01 to 1.09, p=0.005). The effect on the composite outcomes was similarly attenuated but the association with recurrent HF was no longer statistically significant (table 2). Propensity score matched analysis by the 2:1 calliper matching method yielded identical results (online supplementary table S3).
Patients in AF
Digoxin was used by 454 (41%) patients in AF and 548 (31.5%) patients without AF. More patients taking digoxin died both among patients in AF (113, 24.9% vs 94, 14.4%) and among those in sinus rhythm (148, 27% vs 133, 10.7%), compared with those not on digoxin. In multivariable analyses, digoxin was associated with a significantly higher odds of death among patients in sinus rhythm than among patients in AF (p for interaction=0.03), which remained significant, although markedly attenuated on propensity adjustment (OR 1.07, 95% CI 1.03 to 1.11, p=0.001) (table 3). Propensity weight adjustment showed no significant effect on mortality with digoxin use among patients in AF (OR 1.01, 95% CI 0.95 to 1.07, p=0.87) (table 3). For recurrent HF, and the composite of death or HF, there was no significant difference in the effect of digoxin among those with and without AF at baseline (online supplementary table S4 and S5).
Effect of digoxin on mortality by presence of AF, HF or both at baseline
Patients in HF
About half of the patients (532, 50.2%) who were in HF, or who had history of HF, were on digoxin compared with about a quarter of those who did not have HF (481, 25.8%). A greater proportion of patients taking digoxin died, both among patients with (156, 29.3% vs 98, 18.6%) and without HF (110, 22.9% vs 130, 9.4%). Similar to the effect in patients with AF, digoxin use was associated with a greater odds of death among patients who did not have HF compared with those who had HF (table 3). On propensity adjustment, this remained significant though markedly attenuated (OR 1.08, 95% CI 1.04 to 1.12, p<0.0001). Propensity weight adjusted analysis suggested a neutral effect of digoxin use on mortality among patients with HF (OR 0.99, 95% CI 0.94 to 1.06, p=0.83) (table 3). Similar results were seen for recurrent HF, and the composite of death or HF (online supplementary table S4 and S5).
Patients with both AF and HF
A significant proportion of patients (412, 14.1%) had both HF and AF at baseline. Over half of these patients (245, 59.5%) were on digoxin. Nearly a quarter of the patients without either HF or AF (265/1151, 23%) also received digoxin. Among patients with both HF and AF, a similar proportion of digoxin users died compared with non-users (65, 26.5% vs 42, 25.2%). However, among those with neither HF nor AF, the proportion of deaths among digoxin users was greater than among non-users (58, 21.9% vs 77, 8.7%). This association persisted in multivariable analyses (OR for patients not in AF or HF 2.26, 95% CI 1.5 to 3.42; OR for patients with both HF and AF 0.71, 95% CI 0.43 to 1.17, p for interaction <0.0001). However, there was no difference in the odds of developing recurrent HF. (online supplementary table S4) Propensity score adjustment suggested persistent harm in patients with neither AF nor HF, and a protective effect of digoxin in mortality for patients who had both AF and HF (OR 0.88, 95% CI 0.80 to 0.98, p=0.019) (table 3). Results for the composite outcome of death or HF were broadly similar in this group of patients (online supplementary table S5).
Discussion
In this large cohort of patients with symptomatic RHD, digoxin use was associated with a small but nominally significant increase in mortality. This association was primarily driven by the effect of digoxin among patients without either AF or HF. By contrast, among patients with conventional indications for digoxin use such as AF or HF, propensity-adjusted analyses suggested a neutral effect on mortality, and a potential reduction in mortality in those who had both AF and HF. These results are based on a well-powered (494 deaths), robust, propensity-adjusted analysis in a cohort of well-characterised patients with RHD. However, these results need to be interpreted with caution because of the non-randomised and retrospective nature of the analyses, and the large difference between the unadjusted and propensity-adjusted estimates, which highlight the magnitude of bias.
Digoxin in RHD
To our knowledge, this is the only study relating digoxin use to clinical outcomes in patients with RHD. The few previous studies have enrolled small numbers of patients and have reported subjective or surrogate outcomes such as symptom relief and exercise capacity. In an open-label, cross-over trial, Ahuja et al randomised 10 patients with AF to receive digoxin, verapamil or metoprolol and compared the efficacy of these agents in improving symptoms and exercise capacity.6 Of the three drugs, digoxin produced the least improvement in subjective symptoms and peak treadmill exercise capacity. In the Control of Rate versus Rhythm in Rheumatic Atrial Fibrillation Trial study, although rate control regimens were not systematically evaluated, digoxin was used as add-on therapy to diltiazem in 15% of patients in the rate control arm to optimise treatment.15 However, neither of these studies was designed or powered to evaluate hard outcomes in relation to digoxin use.
Digoxin and mortality
Numerous observational studies and a few randomised trials have addressed the effect of digoxin use and mortality in patients with non-valvular AF and HF, but have yielded contrasting results. Observational studies have in general suggested that digoxin use is associated with about a 20% higher mortality.8 9 16 On the other hand, digoxin has been shown to have a neutral effect on mortality in randomised trials.7 17 Much of this difference is likely to be due to confounding by indication (prescription/treatment bias) in observational studies.18 In a metaregression of all-cause mortality according to risk of bias, Ziff et al showed an increase in the risk of mortality with increasing risk of bias.17 Propensity-adjusted analyses may help reduce confounding due to known variables to some extent. Expectedly, studies employing propensity matched analysis show a considerably attenuated risk of death associated with digoxin, with some reporting a neutral effect and others showing a smaller increase in risk than that obtained after conventional multivariable adjustment.17 In our analysis, we found an OR of 1.63 for death on multivariable adjustment which attenuated markedly to 1.05 in the propensity-weighted analysis, which is similar to the risk ratios (RR) obtained by Ziff et al (RR 1.61 and 1.18).17 The large difference in risk between the propensity-adjusted and unadjusted analyses highlights the substantial prescription bias in digoxin use in observational studies.19
Digoxin, recurrent HF and hospitalisation
Unlike in previous studies of non-valvular AF and HF,17 we did not observe any benefit of digoxin on recurrent HF or hospitalisation. There could be several possible explanations for this discrepancy. First, even if in reality digoxin reduced HF, its effects may be masked by the substantial differences in inherent risks between patients prescribed digoxin and those who were not. Second, worsening HF in patients with RHD may be due to progression of valve disease or occurrence of new valve disease which are unlikely to be affected by digoxin. We did not collect information on cause of admission to hospital during follow-up. Hospitalisation in patients with valve disease could be due to progression of valve disease, admission for performance of urgent interventions or due to complications such as stroke, infective endocarditis or recurrence of rheumatic fever, none of which are affected by digoxin use.
Digoxin use in patients without AF or HF
Unlike patients without valve disease, in whom digoxin is only used in those with AF or HF, a significant proportion of patients with RHD receive digoxin in the absence of these conditions (primarily with a view to reducing transvalvular pressure gradients and symptoms in patients with mitral stenosis). In this study, treating physicians appear to have prescribed digoxin in the absence of AF or HF to patients who had severe disease, larger left atrial and left ventricular dimensions, along with diuretics and ACE inhibitors or receptor blockers (online supplementary table S6). We found a significant increase in the odds of death with digoxin use in patients without either AF or HF. Digoxin appeared to confer a significant reduction in mortality among those with both AF and HF which persisted in propensity-adjusted analysis. These results are consistent with previous studies which showed that the benefits of digoxin are most pronounced in patients with severe HF,7 and the risk of death is highest in those with no HF and the lowest baseline cardiac risk.20 The mechanisms of increased mortality with digoxin among low-risk patients are unclear. It is plausible that among patients in AF and HF, and the accompanying neuroendocrine derangements, the salutary effects of digoxin may balance its deleterious effects such as proarrhythmia, while low-risk patients may only be exposed to its deleterious effects without deriving any benefit.20 21 These data may potentially be used to guide the choice of patient populations for inclusion in randomised controlled trials of digoxin in RHD.
Limitations
The most important limitation of our study is that this was a retrospective analysis of observational data, and was therefore subject to all the biases associated with this design, particularly confounding by indication (prescription bias).19 We attempted to mitigate imbalances between the digoxin and no digoxin groups using propensity score weighting and matching, but there were large differences in the risk associated with digoxin use between multivariable and propensity-adjusted analyses. In the presence of large biases, we believe that no amount of statistical adjustment will be sufficient to yield reliable effect estimates. The substantial attenuation of the apparent excess mortality raises the possibility that the residual excess risk may be spurious.19 We only included variables present at baseline in our analyses, and did not account for the occurrence of AF or HF, and initiation of digoxin during follow-up. Given that this study was performed in low-resource settings, we did not measure serum digoxin levels, which may be related to mortality.22 Finally, we did not collect data on cause of death. This information would have provided useful insights into the mechanisms contributing to increased mortality with digoxin.
Conclusion
Digoxin use appears to be associated with a slightly higher mortality in patients with RHD, which was mainly seen in patients who did not have AF or HF. But these estimates are based on a retrospective, propensity-adjusted analysis and may not be reliable due to the substantial biases associated with digoxin prescription and its use. However, these data suggest that patients with RHD who have AF, and/or HF may potentially derive some benefit from digoxin use. Our data provide the impetus and background information that can assist in designing large trials of the use of digoxin in patients with RHD.
Key messages
What is already known on this subject?
Among patients with non-valvular atrial fibrillation (AF), observational data suggest that digoxin may increase mortality. However, there are no data relating digoxin use to clinical outcomes in patients with rheumatic heart disease (RHD).
What might this study add?
This is the first report of the effect of digoxin on clinical outcomes in patients with RHD. These data suggest a possible association of digoxin and mortality in this population, particularly among those without AF or heart failure (HF).
How might this impact on clinical practice?
The attenuation of the association of digoxin with mortality with propensity adjustment indicates the presence of substantial treatment bias in observational analyses, and highlights the need for randomised trials of digoxin in RHD. In the meantime, it may be prudent for clinicians to avoid using digoxin in patients with RHD who neither have AF or HF.
Acknowledgments
The authors acknowledge the substantial contribution made by the following individuals: Addis Ababa, Ethiopia: Araya Gidey Desta, Bekele Alemayehu Shasho, Dufera Mekonnen Begna; New Delhi, India: Jitender Sharma, Gaurav Purohit; Nairobi, Kenya: Christine Yuko Jowi; Windhoek, Namibia: Henning du Toit, Masomi Kaaya, Liina Sikwaya, Andreas Wilberg; Abeokuta, Nigeria: Tanimowo Sunkanmi; Jos, Nigeria: Ludu Audu, Charity Durojaiye-Amodu, Ngozi Elekwa, Ogechi Maduka, Oludolapo Marcaulay, Shamsudeen Mohammed, Halim Odiachi; Cape Town, South Africa: Dylan Barth, Patrick Commerford, Rezeen Daniels, Veronica Francis, Felicia Gili, Alexia Joachim, John Lawrenson, Carolise Lemmer, Nonkululeko Koyana, Katya Mauff, Kathryn Manning, Wendy Matthiassen, Alet Meiring, Peggy Mgwayi, Lwazi Mhlanti, Simpiwe Nkepu, Mpiko Ntsekhe, Janine Saaiman, Unita September, Kathie Walker, Marnie van de Wall; Polokwane, South Africa: Priscilla Adolf, Jabulani Mbokazi, Susan Perkins; Maputo, Mozambique: Neusa Jessen; Khartoum, Sudan: Tagwa Eltahir. The authors also thank Dr Shrikant Bangdiwala (Hamilton, Canada) for reviewing the statistical methods. Dedication: The authors wish to dedicate this manuscript to Prof Bongani Mayosi, who was a dear colleague, friend and mentor. This work would not have been possible without his efforts.
References
Footnotes
Contributors Conception and design: GK, ND. Data acquisition: LZ, MEE, SR, GK. Data analysis: ND. Data interpretation: all authors. Drafting of manuscript: GK. Critical revision for important intellectual content: all authors. Approval of final version: all authors. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding REMEDY was funded principally by the Canadian Network and Centre for Trials Internationally (CANNeCTIN) to GK as part of the Clinical Research Initiative of Canadian Institutes of Health Research. The other sources of funding were the South African Medical Research Council, Lily and Ernst Hausmann Trust, the Else Kroner Fresenius Foundation, the University of Cape Town, the National Research Foundation of South Africa, Harold and Ethel Pupkewitz Heart Foundation (Namibia) and the World Heart Federation. The Jos site was funded by the Jos University Teaching Hospital, the Heart Aid Trust and Faith Alive Foundation. The Sudan sites had partial funding from Sheikan Insurance. LZ was funded in part by the Discovery Foundation, a US National Institutes of Health Fogarty International Clinical Research Fellowship, Thrasher Research Fund Early Career Award, Wellcome Trust Clinical Infectious Disease Research Initiative (CIDRI) Research Officer Award, the Hamilton Naki Clinical Scholarship and by Medtronic Foundation through support to Rheumatic Heart Disease Action. SY is funded by the Marion Burke Chair of the Heart and Stroke Foundation of Canada.
Competing interests None declared.
Patient consent Not required.
Ethics approval Ethics committees of all participating sites approved the study protocol.
Provenance and peer review Not commissioned; externally peer reviewed.
Collaborators The REMEDY Investigators: clinical investigators and sites: Egypt: Benha University: Azza Abul Fadl (Principal Investigator); Cairo University: Sahar S Sheta (Principal Investigator). Ethiopia: University of Jimma: Abraham Haileamlak (Principal Investigator); University of Addis Ababa: Senbeta G Abdissa, Dufera M Begna, Wandimu Daniel, Araya G Desta, Dejuma Yadeta Goshu (Principal Investigator), Bekele A Shasho. Kenya: University of Nairobi: Bernard Gitura, Stephen Ogendo (Principal Investigator). Malawi: University of Malawi: Neil Kennedy (Principal Investigator). Mozambique: Eduardo Mondlane University: Albertino Damasceno (Principal Investigator); Instituto Nacional de Saúde: Ana Olga Mocumbi (Principal Investigator). Namibia: Windhoek Hospital: Christopher Hugo-Hamman (Principal Investigator). Nigeria: University of Abuja: Dike Ojji (Principal Investigator); University of Jos: Ganiyu A Amusa, Fidelia Bode-Thomas (Principal Investigator), Christopher C Yilgwan, Olukemi Ige, Basil Okeahialam; Kano University: Mahmoud U Sani (Principal Investigator); University of Ibadan: Okechukwu S Ogah (Principal Investigator); Federal Medical Centre, Abeokuta: Taiwo Olunuga; University College Hospital, Ibadan: Abiodun M Adeoye, Okechukwu Ogah (Principal Investigator). Rwanda: King Faisal Hospital: Joseph Mucumbitsi (Principal Investigator). South Africa: University of Cape Town: Blanche Cupido; University of Limpopo: Phindile Mntla (Principal Investigator), Christopher Sutton (Principal Investigator), Rajeev Misra. Sudan: Alzaiem Alazhari University: Ahmed ElSayed (Principal Investigator), Ahmed S Ibrahim; Ahmed Gasim Teaching Hospital: Huda HM Elhassan (Principal Investigator). Uganda: Makerere University: Peter Lwabi, Charles Mondo (Principal Investigator), Emmy Okello. Yemen: University of Sana’a: Mohammed M Al-Kebsi (Principal Investigator). Zambia: University of Lusaka: John Musuku (Principal Investigator).