We included full-text publications of randomised controlled trials (RCTs) and confounder-adjusted observational studies that enrolled patients with stroke from African countries and reported results on the efficacy of tertiary interventions for studies that measured our predefined primary (global disability) or secondary (neurologic function scales) outcomes. We included randomised and non-randomised studies to provide a comprehensive overview of the existing high-quality research given the limited number of existing literature. Observational studies were restricted to prospective cohort studies with a control of most important potential confounders (eg, age, stroke severity, time after onset) to minimise biased treatment effects.33–35 We differentiated between acute stroke (≤48 hours after stroke onset), subacute stroke (>48 to ≤6 months after stroke onset) and chronic stroke (>6 months after stroke onset). Tertiary prevention interventions involve physiotherapy, electrophysiologic treatment, psychotherapeutic interventions, community rehabilitation and transitional care following the current American Heart Association (AHA) guideline (table 1).24

We performed a systematic search in nine electronic bibliographic databases (MEDLINE, Embase, the Cochrane Library, the Cumulative Index to Nursing & Allied Health Literature, CAB Direct, Physiotherapy Evidence Database), including specialised African databases (African Journals Online, African Index Medicus) and in the International Clinical Trials Registry Platform until January 2022, with an update using MEDLINE in August 2023, according to our search strategy.32 The search strings included Medical Subject Headings, keywords describing stroke (eg, cerebrovascular, cerebral or intracranial accidents, infarction, disease, haemorrhage or ischaemia), Africa, a list of all 55 African countries and keywords related to eligible study designs. A detailed search strategy is provided in online supplemental file 2. We additionally screened the reference lists of all included full-text papers, contacted specialists and asked all first authors of the included studies for similar ongoing studies. No language limits were applied.

All references were imported to Rayyan,36 which is a web and mobile app for systematic reviews. After exclusion of duplicates, a title and abstract screening was performed independently by two authors (LA, ESK). Then, a full-text screening was independently performed by the same two authors. In all steps, disagreements were resolved in discussion with a third author (SU). Next, data extraction was performed by one author (LA), including the main characteristics and results of all included studies. A second author (BYW) checked all extractions.

We narratively synthesised the data to cover the included studies comprehensively. We also performed a meta-analysis for electrophysiologic interventions using a random-effects model on metric data described by mean differences (MDs) and standardised MDs (SMDs). A negative SMD for the Barthel score and the modified Rankin scale was regarded as a benefit for the patient. All treatment effects are reported with their corresponding 95% 95% CIs. We used SMDs to summarise the results of comparable interventions on our primary outcome, global disability across different measurement scales (online supplemental table 1). SMDs above 0.8 indicate a high effect of the interventions.37 Data were analysed using Comprehensive Meta-Analysis V.3.3 and Review Manager V.5.3.

We merged material diverse in methodology, participants, setting, interventions or outcomes studied. We assessed I² statistics to differentiate between considerable, substantial, moderate or not important heterogeneity according to the Cochrane Handbook.38 Subgroup analyses were planned to describe differences in methodologic quality of studies (eg, RCT vs non-RCT), setting (eg, urban vs rural) and timepoint of treatment (acute vs subacute vs chronic care). Due to the sparse data found, we performed only a subgroup analysis describing heterogeneity due to varying interventions (figure 2).

Risk of bias was described and judged independently by two authors on the basis of the revised Cochrane Risk-of-Bias tool for RCTs39 in five different domains (randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcomes, selection of the reported results) as ‘low risk’, ‘some concern’ and ‘high risk’. Discrepancies were resolved by discussion between the authors or with a third author. Publication bias was investigated via funnel plots (online supplemental figure 1).

No patients were involved.

None of the studies fulfilled the requirements for an overall low-risk assessment. However, most of the studies reported methods leading to low risk of bias in domains 1–3 (figure 3). Studies scoring with high bias did not report data for all participants analysed or did not provide information about the number of randomised participants41 42 50 52 56–58 67 (figure 3). Data from more than 90% of randomised participants were analysed in 11 studies, whereas in 4 studies data from less than 90% of randomised participants were analysed. Studies scoring with a high bias in domain 4 either had no blinding of the outcome assessor or did not report it.43 47 49–51 57 59 60 Nearly all studies scored with some concern in domain 5 because they did not report a prespecified analysis plan. A funnel plot on reporting of our primary outcome showed an under-representation of small studies, with negative results indicating a possible publication bias (see online supplemental figure 1).

The available evidence on tertiary prevention stroke research stems from only 6 out of 54 African countries, with most studies set in countries with a comparably low score on the human development index (11 from Nigeria, 1 from Ivory Coast, 1 from Benin and 1 from Uganda). The great majority of the participants were from urban areas. The tenuous enrolment of patients from rural areas with still the highest population percentage might reflect organisational challenges recruiting patients for rehabilitation interventions and hint at the socio-economic barriers of the patients that can complicate access to rehabilitation.68–70 Almost half of the studies were published by two authors, indicating a successful research community interested in tertiary prevention interventions in Nigeria (five studies by Abdullahi and colleagues)45–49 and Egypt (six studies by Khedr et al).59 60 62 63 66 67 Establishment of strong and productive research clusters in new settings is highly necessary and can be informed by such best-practice examples.

Our review includes 973 patients with stroke without a remarkable gender bias. Many participants were younger than 55 years because African patients with stroke are generally younger in accordance with the findings of Yuyun et al11 compared with HICs, which implicates a high potential for recovery.71 72 Therefore, physiologic functional improvements of the patients, which take place anyway, might be even more accentuated and the comorbidity burden might be lower.72 Thus, the applicability for elderly patients still needs to be addressed.73 A total of eight studies with a total of 154 participants included patients with chronic stroke, while the best time point to start rehabilitation is the subacute phase, during which only 12 studies were conducted. These studied interventions might have shown improved patient outcomes if they had included patients earlier.24 74 75

Most studies exclusively included patients with a first-ever stroke, possibly leading to an under-representation of patients with recurrent stroke. Recurrent stroke rates are estimated to be up to 22.5% within 5 years.5 76 77 Patients suffering from recurrent stroke and thus having more impairments are in greater need of rehabilitative interventions than first-ever patients with stroke.

Besides, the diagnostic procedure was not sufficiently described in many studies. An uncertainty remains concerning the inclusion of patients with conditions that can mimic stroke such as epilepsy, subarachnoid haemorrhage, demyelinating diseases, migraine, functional disorders and others. Other neurologic diseases have a higher tendency to regression than stroke. This is likely due to an overall lack of radiologic diagnostic equipment.78

The research spans from the adaptation of established concepts such as physiotherapy with and without medical appliances43 50–52 54 57 61 to the African context to cutting-edge research on transcranial magnetic stimulation with international recognition. Studies by Khedr et al were incorporated in European guideline development.79 Approved tertiary prevention interventions seem feasible with comparable benefits in African countries as in HICs. Therefore, research should focus on measures to implement established evidence-based tertiary prevention interventions. This is consistent with evidence from non-African low-income countries.80 81 Known factors influencing rehabilitation uptake include level of education, travel burden, cost and religious beliefs.3 82 Several rehabilitation approaches are being explored in low-income countries, not just in African countries, such as integrated care pathways, community health workers and the use of digital health technologies such as mobile health, video-based instructions and virtual reality.81 However, most evidence is still of comparatively low quality.83

Most of the trials included physiotherapists, occupational therapists and speech therapists. Educational costs for these professions are usually lower than they are for physicians.84 Task shifting and working in multidisciplinary teams could solve some allocation and educational shortcomings in low-resource settings, as has been recommended by the WHO85 86 and multiple studies.87 However, the availability of therapists in low-income countries in general is often low.82 88 Therefore, task shifting to caregivers can also be a promising approach.81 This process can be supported by telerehabilitation concepts.55

The positive patient outcomes of the included transitional care studies indicate a high potential of promoting self-sufficiency and QoL of patients with stroke in line with the WHO promotion of community-based rehabilitation98 and task shifting,86 as well as the South African stroke guideline for under-resourced areas.99 Additionally, the AHA guideline mentions benefits in ambulation, mobility, self-care, functional independence and patient and caregiver satisfaction.24 100–103

The outcome measures primarily focused on scales of neurologic functioning (17 out of 25) and do not report QoL or global disability. Thus, it often remains unclear if the benefits in these scales translate into an improvement in real life.104 None of the included studies reported on cost-effectiveness, so statements regarding resource allocation are not possible.

We included high-quality studies based on a comprehensive systematic search that included specifically African databases and a thorough literature search with reference screening and author inquiry. We did a study registration and published a study protocol ensuring transparency of the research methods used. Additionally, we worked in close collaboration with researchers from Ethiopia and Cameroon. A first limitation is the lack of inclusion of non-English publications that we addressed by reference screening and contacting of study authors to close the gap. Second, owing to the heterogeneity of the studies we could only perform a small meta-analysis, so to merge some results of the interventions, we therefore narratively summarised the existing data. Moreover, almost half of our studies stem from two research groups, which makes it difficult to draw general conclusions. Third, the questionable diagnostic accuracy of stroke due to the lack of radiologic equipment might limit the validity of the results.

Finally, none of the trials we considered provided a low risk of bias. Especially, prespecified analysis plans or protocols have not been reported. Only four studies were double-blinded. However, it must be considered that blinding participants to their intervention in rehabilitation studies often is impossible due to the nature of the interventions.46 55 56 Positively, most of the studies reported details on their randomisation process, did not deviate from the planned intervention and reported results on most randomised patients, thereby implicating reliable data.

As the adherence to reporting guidelines was insufficient in most studies, efforts to improve the adherence to standardised reporting guidelines are advisable.104 105

While the existing research shows the effectiveness of physiotherapeutic interventions in African patients, uptake of tertiary prevention after hospital discharge is a major barrier. Health services research on how to effectively implement tertiary prevention interventions in real-world setting is necessary.3 The WHO’s community-based rehabilitation guidelines could be used as a starting point.106

Not applicable.

Not applicable.