Diagnostics Are Clinical Research Too: Raising the Standard for Validation in African Populations

Nov 19, 2026 | Blog

A diagnostic test earns its place in care by providing a trustworthy result that informs a useful decision. Validation must establish both where it works and what happens when it is used.

Before a diagnostic test is purchased, one question deserves more attention: what will change for the patient when the result arrives?

A positive result might trigger treatment, a referral or further investigation. A negative result might prevent unnecessary prescribing, or falsely reassure a clinician. The value of the test depends on its accuracy, but also on the decisions and services surrounding it.

This makes diagnostic evaluation a clinical research responsibility, not simply a laboratory checkpoint. Analytical studies establish whether a test reliably detects or measures its target. Clinical accuracy studies establish how well it identifies a condition in the intended population. Clinical utility studies examine whether using it improves management or outcomes. These are connected questions, but they require different evidence.

The strategic task is to connect them before a product enters routine care, rather than discover the gaps afterward.

Local validity is a scientific question

“African populations” should never be treated as a single validation category. Relevant differences may exist between countries, between facilities in the same district, and between patients attending the same clinic.

Sensitivity and specificity can vary with the clinical characteristics of the people tested and the circumstances of testing. A study dominated by advanced disease may not establish performance in people with early or less typical presentations. Likewise, performance against one set of circulating pathogens may not hold against another.

Malaria provides a concrete example. A study published in 2021 enrolled 12,572 symptomatic participants at health facilities in three Ethiopian regions. Investigators estimated that rapid tests targeting histidine-rich protein 2 would miss 9.7% of Plasmodium falciparum infections across the study sites because of deletion of the pfhrp2 gene. The estimate varied substantially between regions. This was a finding about particular populations and parasites, not a continent-wide failure rate (1).

Testing strategies also need to reflect disease prevalence. As the condition becomes less common among those tested, false-positive results can account for a larger share of all positive results, even when sensitivity and specificity remain unchanged. WHO’s 2024 HIV misdiagnosis guidance recommends a standard three-test strategy for all countries, alongside quality assurance measures, to maintain a positive predictive value of at least 99% (2).

The implication is not that every country must repeat every study. It is that using evidence from elsewhere requires a reasoned assessment of what is comparable and what remains uncertain. Geography alone is neither proof of applicability nor a reason to reject useful evidence.

Study the decision, not only the device

The TB-NEAT trial illustrates why better detection and better outcomes must be examined separately. Conducted in primary-care facilities in South Africa, Zimbabwe, Zambia and Tanzania, it compared nurse-performed, point-of-care Xpert MTB/RIF testing with smear microscopy.

Xpert increased same-day diagnosis and treatment initiation. However, the trial did not demonstrate lower tuberculosis-related morbidity at two or six months. The investigators attributed this partly to substantial empirical treatment among patients with negative smear results (3).

The lesson is not that faster, more sensitive testing lacks value. It is that the benefit of a new test depends partly on what clinicians already do and what the new result enables them to do differently.

Diagnostic research should therefore follow the pathway from specimen collection to action. Does the result arrive before treatment is chosen? Does it reach the person responsible for acting on it? Is confirmatory testing, treatment or referral available? Does the patient need another visit, and can they complete it?

Not every diagnostic validation study needs to demonstrate improved patient outcomes, and not every test requires a randomized trial. The evidence should be proportionate to the intended claim and the consequences of error. An accuracy study should not, however, be presented as proof of clinical benefit.

Build the evidence around intended use

A useful validation plan begins with the clinical role: screening, confirming a diagnosis, ruling out a dangerous condition or guiding treatment. That role determines which errors matter most and what performance is acceptable.

Researchers should define the intended population, user, specimen, setting and decision before selecting performance targets. Recruitment should reflect the patients likely to encounter the test, rather than only clearly positive cases and healthy controls. Planned analyses should examine clinically important differences without implying certainty where subgroup numbers are small.

The reference standard needs equal scrutiny. It should be justified, consistently applied and interpreted independently of the test under evaluation wherever possible. Where no sufficiently accurate reference exists, the study should explain how that limitation is handled rather than conceal it behind precise-looking estimates.

Specimen collection, transport, storage and traceability belong in the protocol. So do the handling of invalid results, missing data and participants who do not complete testing. STARD 2015 provides a reporting framework that makes these choices visible and helps readers judge bias and applicability; it does not substitute for sound study design (4).

Studies also need appropriate ethics review, consent arrangements and safeguards for clinically important findings. Participants are not merely sources of specimens.

Make validation useful to purchasers and health services

Evidence for market authorization and evidence for a purchasing decision overlap, but they do not answer identical questions. A health service must understand not only whether a test performs adequately, but whether it can deliver a reliable service with the staff, infrastructure and funding available.

Operational evaluation should measure invalid-result rates, turnaround time, training needs, maintenance requirements and supply interruptions. Environmental and infrastructure requirements should be tested against the intended setting. Economic evaluation should look beyond the purchase price to the costs of repeat testing, staff time, patient travel and follow-up.

These questions give practical meaning to the World Health Assembly’s 2023 resolution on strengthening diagnostics capacity. The resolution urges countries to consider national diagnostics strategies, health technology assessment and essential diagnostics lists adapted to local needs, alongside investment in research and production (5).

Research funders, manufacturers and procurers should agree early on which evidence will inform adoption. Funding a product evaluation without budgeting for workflow assessment or patient follow-up risks leaving the most consequential questions unanswered.

Shared, well-characterized specimen panels and common protocols can support more consistent comparisons across sites. They should complement, not replace, prospective evaluation in intended users and patients. A specimen panel cannot show whether a result reaches a clinician in time.

Trust must be maintained after introduction

Validation is not a permanent guarantee. Products, users, disease patterns and service conditions can change.

WHO distinguishes manufacturer-led post-market surveillance from regulator-led market surveillance, both concerned with continued safety, quality and performance (6). Diagnostic services should contribute through clear reporting routes for suspected false results, device failures and recurring operational problems. Those reports need an agreed response, not merely a place in a database.

For ACRN, the priority should be to connect these stages: rigorous clinical evaluation, dependable specimen and data systems, assessment across relevant settings, and evidence that informs actual care decisions. A research network’s contribution should be judged by the questions its studies resolve, not simply by the number of sites participating.

The standard is not whether a diagnostic has been tested somewhere in Africa. It is whether the evidence supports its use for a defined purpose, in the people and services expected to rely on it, with a plan to identify when that evidence no longer holds.

That is what makes diagnostic validation clinical research. And it is what turns a test result into something patients and clinicians can trust.

References

  1. Feleke SM, Reichert EN, Mohammed H, Brhane BG, Mekete K, Mamo H, et al. Plasmodium falciparum is evolving to escape malaria rapid diagnostic tests in Ethiopia. Nature Microbiology. 2021;6(10):1289–99.
  2. World Health Organization. Preventing HIV misdiagnosis: implementation guide. Geneva: World Health Organization; 2024 2024/11/07. Report No.: 9789240092136.
  3. Theron G, Zijenah L, Chanda D, Clowes P, Rachow A, Lesosky M, et al. Feasibility, accuracy, and clinical effect of point-of-care Xpert MTB/RIF testing for tuberculosis in primary-care settings in Africa: a multicentre, randomised, controlled trial. The Lancet. 2014;383(9915):424–35.
  4. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ : British Medical Journal. 2015;351:h5527.
  5. World Health Assembly. Resolution WHA76.5: Strengthening diagnostics capacity. Geneva: World Health Organization; 2023.
  6. World Health Organization. Guidance for post-market surveillance and market surveillance of medical devices, including in vitro diagnostics. Geneva: World Health Organization; 2020.