Rapid antigen test as a screening tool for SARS-CoV-2 infection: Head-to-head comparison with qRT-PCR in Ethiopia

Africa’s reported COVID-19 burden appeared milder than anticipated, but limited surveillance and testing likely underestimated true infection rates. High seroprevalence among healthcare workers and communities in several African countries, including Ethiopia, indicates widespread transmission. Despite this, early detection and treatment remain critical, particularly in hospital settings. qRT-PCR is the global confirmatory test, but it is resource-intensive and has a ≥24-hour turnaround. Antigen rapid diagnostic tests (Ag-RDTs) can perform well during early infection in individuals with high viral load and in higher-incidence settings. WHO recommends Ag-RDTs meeting ≥80% sensitivity and ≥97% specificity, primarily for symptomatic individuals or high-risk asymptomatic contacts where qRT-PCR capacity is limited. Ethiopia adopted Ag-RDTs as point-of-care tests without comprehensive local validation. This study’s research question was to determine the diagnostic accuracy of Panbio COVID-19 Ag-RDT versus qRT-PCR among patients attending referral hospitals in Oromia, Ethiopia, to inform appropriate use in this setting.

The introduction references multiple seroepidemiological studies across Africa reporting high SARS-CoV-2 seroprevalence (e.g., 41.2% in DRC, 45.1% in Nigeria, and high rates among South African blood donors), and escalating seroprevalence in Ethiopia (from 10% to nearly 60% in 2020–2021). It outlines WHO guidance on Ag-RDT use and performance thresholds (≥80% sensitivity, ≥97% specificity), emphasizing their lower sensitivity versus qRT-PCR, especially among asymptomatic individuals, and recommending their use in symptomatic cases or high-risk asymptomatic contacts in settings with limited molecular testing. Prior evaluations have shown Ag-RDTs perform best in early infection at higher viral loads (Ct ≤25). The Ethiopian context included rapid adoption of Ag-RDTs for point-of-care screening without local validation of diagnostic performance, motivating this head-to-head comparison.

Study design and setting: Cross-sectional study conducted August 9–23, 2021, across six high-caseload referral/university teaching hospitals in four towns in Oromia Regional State, Ethiopia (Jimma University Specialized Hospital; Wollega University Teaching and Referral Hospital; Nekemte General Hospital; Ambo University Teaching and Referral Hospital; Ambo General Hospital; Adama Hospital Medical College). Oromia had 15 qRT-PCR testing centers and >60 Ag-RDT sites during the pandemic. Participants and sampling: A total of 1,721 in- and outpatients presenting for various conditions, including suspected COVID-19 cases, inpatients, pre-operative patients, and general outpatients, were recruited after consent. Two nasopharyngeal swabs were collected per participant for parallel Ag-RDT and qRT-PCR. Due to specimen quality and labeling issues, 1,364 participants with both test results were included in the head-to-head analysis. Median age was 32 years (IQR 24–50); 52% female; 40.9% had underlying medical conditions. Ag-RDT procedure: Panbio™ COVID-19 Ag Rapid Test Device (Abbott). NP swabs were processed per manufacturer’s instructions at point-of-care; results read at 15 minutes. Manufacturer-reported performance: specificity 99.8% (95% CI 98.65%–100%), overall sensitivity 98.1% (95% CI 93.2%–99.8%), with highest sensitivity in first 3–7 days of symptom onset. qRT-PCR procedure: Abbott RealTime SARS-CoV-2 assay on NP swabs in 2 mL VTM. Samples transported within 48 hours at 2–8°C to Adama and Nekemte Public Health Reference Laboratories (BSL-2). Ct values retrieved and used to categorize viral load. Manufacturer clinical performance: Positive Percent Agreement 95.9% (95% CI 86.0%–99.5%), Negative Percent Agreement 100% (95% CI 93.5%–100%). Data collection and quality: Demographic and clinical data collected electronically (CSPro/ODK), with daily supervision and central server checks. Data collectors trained; tools translated to local languages (Afaan Oromo, Amharic). Laboratory procedures adhered to PPE and biosafety standards; cold chain maintained. Statistical analysis: Data cleaned in Excel and analyzed in Stata 14. Calculated sensitivity, specificity, PPV, NPV, overall agreement, Cohen’s kappa, ROC/AUC. Positivity rates assessed across clinical subgroups and exposure histories. 95% CIs reported; p<0.05 considered significant. Ethics: Approvals from Jimma University Institute of Health IRB (Ref No-JUIR/IRB/333/23) and Oromia Health Bureau ERB. Written informed consent obtained; positive cases managed per national protocols.

- Sample included 1,364 participants with paired Ag-RDT and qRT-PCR results for head-to-head analysis. - Positivity rates: qRT-PCR 17.74% (242/1364); Ag-RDT 6.5% (89/1364). Higher positivity in those with contact with confirmed/suspected cases and those with recent COVID-19 symptoms (all p<0.001). - Diagnostic accuracy of Ag-RDT vs qRT-PCR (overall): • Sensitivity: 33.3% (95% CI 31.0%–36.0%) • Specificity: 99.3% (95% CI 98.8%–99.7%) • PPV: 91.0% (95% CI 89.5%–92.5%) • NPV: 87.4% (95% CI 85.6%–89.1%) • Overall agreement: 87.6%; Cohen’s kappa: 0.66 (substantial agreement) • AUC-ROC: 0.67 (95% CI 0.63–0.69); improved to 0.76 (95% CI 0.71–0.80) among clinically suspected cases with typical symptoms. - Subgroup sensitivities (specificity ≥98% across): • Shortness of breath: 73.3% • Fever + cough + dyspnea: 71.4% • Myalgia/arthralgia: 65.9% • Fever: 57.4%; Cough: 51.1%; Loss of taste/smell: 50.0%; Fatigue: 55.0%; Headache: 50.7%. - Viral load correlation: Mean Ct among qRT-PCR positives was 24.6. Ag-RDT positivity was 51.7% for Ct <25, 3.4% for Ct 25–30, and 0% for Ct >30 (p<0.001). 96.3% of Ag-RDT positives had Ct <25. - qRT-PCR detected 91% (81/89) of Ag-RDT positives, whereas Ag-RDT detected 33.3% (81/243) of qRT-PCR positives.

The study demonstrates that Panbio COVID-19 Ag-RDT has high specificity but substantially lower sensitivity than qRT-PCR for detecting active SARS-CoV-2 infection in a mixed outpatient and inpatient population in Ethiopia. Sensitivity improves in symptomatic individuals with typical features (e.g., shortness of breath, fever, cough) and in those with higher viral loads (Ct <25), aligning with the known performance characteristics of Ag-RDTs. Despite high specificity meeting WHO minimum performance thresholds, the overall sensitivity falls below WHO-recommended standards (≥80%), especially when including asymptomatic individuals or those outside the early symptomatic window. The findings indicate that Ag-RDTs are best leveraged as point-of-care screening tools in higher-incidence settings and among symptomatic, lower-risk patients, while qRT-PCR should remain the primary diagnostic, particularly for high-risk cases or when clinical suspicion is high. The performance metrics and ROC analysis further support selective, algorithm-based use of Ag-RDTs to optimize diagnostic yield and resource allocation in low-resource settings.

Ag-RDTs (Panbio) offer operational advantages for point-of-care screening but exhibit low overall sensitivity versus qRT-PCR, risking false negatives and potential misdiagnosis. Their use should be targeted to settings with high incidence and to symptomatic patients at low risk for severe disease. For high-risk patients and definitive diagnosis, qRT-PCR should be prioritized or performed in parallel with Ag-RDT. These findings guide more judicious deployment of Ag-RDTs within diagnostic algorithms in resource-limited contexts.

The primary limitation was the lack of data on timing of symptom onset at sample collection, precluding classification of participants as pre-symptomatic, symptomatic, or late symptomatic. This may have affected sensitivity estimates given the time-dependent performance of Ag-RDTs.