This systematic review did not require human subjects or institutional review board oversight. It is reported in accordance with PRISMA guidelines and is registered in the PROSPERO (International Prospective Register of Systematic Reviews) database (Located at http://www.crd.york.ac.uk/NIHR_PROSPERO/) (Registration #CRD42012003161).[10], [11]

Observational studies were included that reported the incidence or prevalence of AMI among any population within SSA. Published articles were required to be in English, Spanish, French or Portuguese, or have complete translations into one of those languages. The study definition of “acute myocardial infarction” was based on the universal definition of myocardial infarction, defined by a combination of clinical symptoms felt by the clinician to represent coronary ischemia within 24 hours of presentation to the hospital or during the present hospitalization and either EKG changes (new ST-segment elevation, new LBBB, new pathologic Q wave) or elevation of cardiac enzymes (troponin or CK-MB >99th percentile) with or without ST-segment elevations on EKG.[12] Our study definition did not require biomarker elevations as most studies were conducted in areas without access to these tests. In situations where more than one study was conducted on the same data set, the study with more complete data was included. Studies were included from January 1992 through October 28, 2012.

The following electronic databases were searched: PubMed, Embase, Global Health Archive, CINAHL, and Web of Science. In addition, two independent reviewers (JTH and JMR) manually evaluated the references of the included articles and performed a citation analysis of the included studies using Google Scholar. Additional proposed articles for inclusion were solicited from the authors of the included studies through email communications. These information sources will be made available to interested readers upon email request to the authors.

The initial search comprised the MeSH terms “Africa”, “Africa South of the Sahara”, “Africa, Central”, “Africa, Eastern”, “Africa, Western”, “Africa, Southern”, “Chest Pain”, “Acute Coronary Syndrome”, “Myocardial Ischemia”, “Angina Pectoris”, “Angina, Unstable”, “Myocardial Infarction” and related entry terms. The complete search strategy used for the PubMed database is shown in Appendix S1. Limits for language and/or time were not used in the initial search.

Titles and abstracts of the retrieved articles were independently evaluated by 2 reviewers (JTH and JMR), and assessed for eligibility according to inclusion criteria. Any abstract considered by either reviewer to potentially be eligible for inclusion was then subjected to a full-text evaluation by both reviewers independently to determine eligibility and documented in separate spreadsheets. Disagreements regarding study eligibility were resolved by consensus and re-review, and if disagreements persisted, a third reviewer decided (CAL).

The quality of the included studies was assessed using the Loney scoring system for evaluating incidence and prevalence studies.[13] The Loney system is based on the presence (score of 1) or absence (score of 0) of the following criteria: (1) random sample or whole population under study, (2) unbiased sampling frame or list from which subjects are drawn, (3) adequate sample size (>300 subjects), (4) standardised measures, (5) outcomes measured by unbiased (blinded) assessors, (6) adequate response rate (at least 70%), (7) confidence intervals (CI), subgroup analysis, and description of refusers provided and (8) all study subjects described. The maximum score was 8 points. Scores of 0 to 4 indicate lower quality, while scores greater than 5 suggest higher quality.

Two reviewers (JTH and JMR) independently conducted the data extraction and any disagreements in the extracted data were resolved by the third reviewer (CAL). The recorded characteristics of the studies included study design, measured prevalence of AMI, and any covariates measured with MI (lifestyle risk factors, patient demographics, and comorbid diseases).

A descriptive analysis of the included studies was performed. AMI prevalence for each study was calculated by dividing the number of patients diagnosed with AMI by the total number of patients evaluated in the study. Since studies represented highly heterogeneous populations, a meta-analysis was not performed. Formal GIS maps were constructed to show the distribution of studies across SSA. Data were processed using the open sourced software R Language version 3.0.1 and Quantum GIS (QGIS) version 1.9.0-Master.[14], [15]

A total of 2292 records were retrieved from the databases, 29 of which were selected for full-text assessment of eligibility. Six observational studies from the databases met all criteria. One additional study was proposed by the corresponding author of an included study, and was included.[16] Ultimately, seven studies were included in the systematic review, representing 92,378 patients.[16]–[22] Figure 1 demonstrates the search and article selection process.

The characteristics of the seven included studies are outlined in Table 1. Included studies came from five countries (South Africa, Sudan, Nigeria, Senegal, and Kenya), and all studies were conducted among hospitalised patients or patients presenting to the emergency department. Both prospective and retrospective studies were included. The included articles assessed the prevalence of AMI among very different study populations, such as patients admitted to the hospital with a diagnosis of acute stroke, patients with diabetes who presented to the emergency department, and intensive care unit patients.[17], [21], [22] The number of study patients ranged from 67 to 77,429 patients, and the mean age of study participants ranged from 55.6 to 63.9 years. In two studies, objective testing for AMI (such as EKG or cardiac biomarkers) was performed on all participants regardless of clinical presentation; in the other five studies, objective testing was performed at the discretion of the clinician treating the patient.

The results of the quality assessment of the included studies using the Loney score are demonstrated in Figure 2. Quality scores of the included studies ranged from 3 to 7 on an 8-point scale, with a score of 8 indicating highest quality. Two studies demonstrated particularly higher methodological quality.[17], [21] All included studies reported data from the entire population under study; no randomization techniques were used. None of the included studies reported a participant response rate or refusal rate.

The prevalence of AMI among each study population is presented in Table 2. The prevalence of AMI ranged from 0.1% in Senegal among patients presenting to an emergency department, to 10.4% in Sudan among diabetic patients who died while in hospital.[18], [20] Due to heterogeneity of study populations, we did not assess for temporal trends. Figure 3 demonstrates the geographic distribution of included studies and their respective AMI prevalences.

There were several limitations to our review. Firstly, nearly all of the studies reviewed were conducted in specific subpopulations, such as acute stroke patients and diabetic patients presenting to the emergency department, making inferences about the prevalence of AMI in the general population difficult. Secondly, by requiring either enzymes or EKG in the diagnosis of presumed AMI, we excluded several studies that diagnosed AMI based on clinical exam alone. This search strategy may have introduced bias by favoring studies conducted in medical facilities in which advanced diagnostic equipment was available and patients with more comorbidities present, resulting in potential overestimation of the population prevalence of ACS and limiting the applicability of study findings to the general population. However, we felt the accuracy of diagnosis was important and required some objective measure to be valid. Thirdly, nearly all of the included studies were conducted in urban centers, likely representing a publication bias and disproportionately sampling patients with unique behavioral and socioeconomic risk factors for coronary artery disease. The findings of these studies, therefore, would have little applicability to poorer, rural communities.

Finally, five of the included studies were subject to reporting bias as IHD was diagnosed only if the treating clinician initiated a workup for it, including EKGs and cardiac biomarkers.[16], [18]-[20], [22] As previously discussed, some clinicians might not explore potential angina symptoms with a relatively expensive cardiac workup. It is unknown what proportion of these patients present with atypical symptoms as compared with wealthier nations, since most patients with atypical symptoms would likely be missed in areas of low disease incidence. Furthermore, in a region in which emergency services and community education on AMI symptoms are often absent, patients may disregard the symptoms of AMI or fail to obtain transport to an emergency department until they are in cardiac arrest, leading to an incalculable number of unreported pre-hospital deaths due to AMI. Given the limitations in the above mentioned studies, we elected not to conduct this review in accordance with Cochrane Collaborative guidelines as we anticipated that none of the above articles would meet strict Cochrane criteria.