Introduction

Sample pooling is an effective technique for increasing testing capacity and reducing costs in diagnostic settings, particularly for detecting infectious diseases such as SARS-CoV-2. This method involves combining multiple patient samples into a single test, with positive pools undergoing subsequent individual testing.1,2 Pooled testing is commonly considered a more important approach, whereby a negative result implies that all individuals in the group are negative, whereas a positive result suggests that at least one person is infected with SARS-CoV-2.3,4 Specimen pooling has emerged as a potential solution to meet the rising demand for SARS-CoV-2 testing amid inadequate infrastructure. Studies have demonstrated that pooling can significantly increase testing capacity while maintaining sensitivity, particularly at low pool depths.5 This approach is most effective in low-prevalence settings, where it can conserve PCR reagents and increase throughput.6 Pooled testing has emerged as a viable strategy to expand COVID-19 screening capacity and conserve resources during the pandemic. Although pooling has shown to increase the efficacy of testing, it also has its own drawbacks, one of which being dilution effect where low viral‑load samples are pooled into a large number of pools as pooling too many samples (eg, 20 or 30) risks missing borderline-positive cases or report high false negative.7 Conversely, with high prevalence rate in the tested population, there is a need for repeating tests, as the result of which its efficacy drops, incurring more cost.8

Most sample‑pooling studies have focused on screening large asymptomatic populations with low prevalence, where primarily highly infectious samples with the lowest Cycle Threshold (CT) values are targeted for detection.9–11 In low-resource countries such as Ethiopia, where individuals commonly seek care at health centers after experiencing COVID-19 symptoms, evaluating the performance of pooling techniques among symptomatic individuals with varied viral loads and CT values would be practically meaningful. This approach can provide valuable information about the feasibility and effectiveness of pooling strategies in real-world healthcare settings.

Optimal pool sizes depend on factors such as prevalence and viral load.12,13 While pooling can significantly reduce the number of tests needed, potentially saving more than 80% of resources,13,14 there are concerns about decreased sensitivity for samples with low viral loads (high CT values). Studies have shown that pooling up to 6 samples does not affect the sensitivity of detecting SARS-CoV-2 when CT values are less than 35.14,15 However, other experimental studies have shown that pools of up to 11 samples can effectively detect positive cases even with low viral loads.12 Some researchers have reported successful detection in pools of up to 20 samples,16 whereas others found pools of 10 to be effective.17 Mathematical models, such as the Dorfman equation, can help determine optimal pool sizes on the basis of prevalence.12 However, it is crucial for laboratories to validate their own pooling protocols, considering factors such as test sensitivity and local infection rates.12,16 In this regard, as far as we know there is only one study from Ethiopia, which was conducted on samples collected before 2021. In fact, more such studies may have benefited the country during the early period of the pandemic if applied to different regions since its prevalence varied ranging from 1.37% reported in West Gondar to 39.6% reported in all the regions of Ethiopia, highlighting how heterogenous it was.18

The aim of this study was to determine the performance of two commonly utilized assays, the Cepheid Xpert® Xpress (Xpert) and Cobas® (Cobas) SARS-CoV-2 assays, for diagnosing SARS-CoV-2 from pooled nasopharyngeal samples and to assess the agreement between the two assays and identify the number of pooled samples that are most effective under these assays by comparing the gene targets and CT values detected in each assay.

Materials and Methods Study Setting, Design and Period

A laboratory-based cross-sectional study was conducted utilizing positive samples collected between June 1 and July 30, 2022. These samples were stored in a calibrated and daily monitored −80°C deep freezer at the Ethiopian Public Health Institute (EPHI) until the testing time on August 12, 2022. Fifty SARS-CoV-2-positive samples representing varying viral load levels (on the basis of their previously determined CT values from the Cobas testing method) were purposively selected for pooling to simulate real-world conditions. This was done by randomly selecting 15 positive nasopharyngeal (NP) samples from the very low group whose CT-values fell within the range of 35.0 to 39.9; 15 positive NP samples from the low group with CT-values ranging from 30.0 to 34.9; 10 positive NP samples from the medium group with CT-values ranging from 25.0 to 29.9; and 10 positive NP samples from the high group with CT-values below 24.9. Testing was performed via Genexpert® equipment (Cepheid, Sunnyvale, California, USA) and Cobas® (Roche Diagnostics, Basel, Switzerland), both of which are available at the Ethiopian Public Health Institute.

Pooling Procedure

To facilitate pooling, we categorized the selected 50 confirmed SARS-CoV-2-positive nasopharyngeal samples mentioned above into four groups on the basis of their initial CT values from previously conducted tests. We then pooled 350 confirmed SARS-CoV-2-negative nasopharyngeal samples with the positive samples. This categorization included 15 samples with very low viral loads (VL) (initial CT values between 35.0 and 39.9), 15 samples with low viral loads (L) (initial CT values between 30.0 and 34.9), 10 samples with medium viral loads (M) (initial CT values between 25.0 and 29.9), and 10 samples with high viral loads (H) (initial CT values below 24.9) (Figure 1). The number of positive samples in the VL and L viral load groups was increased to enhance the representation of the naturally positive samples from these viral load groups since the potential for a low detection rate during further pooling becomes high. All the chosen samples were pooled into four different pools (pool-4, pool-6, pool-8, and pool-10) as follows: Pool-4, nasopharyngeal samples pooled by mixing one positive sample with three negative samples; Pool-6, nasopharyngeal samples pooled by mixing one positive sample with five negative samples; Pool-8, nasopharyngeal samples pooled by mixing one positive sample with seven negative samples; and Pool-10, Nasopharyngeal samples pooled by mixing one positive sample with nine negative samples.

Figure 1 Positive samples grouping for pooling and pooling work flow.

Note: This work flow was followed for each of the four viral load groups (VL, L, M, and H).

A total of 200 pooled samples were prepared using 50 positive samples (15 each from very low and low, and 10 each from medium and high viral load group). To pool the samples, a sterile 7 mL plain tube was used for each pool, and each selected positive sample was subjected to vortexing for 20 seconds before pooling to ensure even distribution of the nucleic acid (NA).

To create a pool-n, a sample was transferred from n-1 negative samples into a labeled plain tube, and a positive sample was then added to the same tube. Accordingly, pool-4 was created by combining 300 µL from each of three distinct negative samples into a labeled tube, along with 300 µL from a vortexed positive sample. For pool-6, a volume of 200 µL was transferred from each of the five different negative samples, and an additional 200 µL of the positive nasopharyngeal sample was added to the same tube. Similarly, for pool-8 and pool-10, the same formula was used. In the case of pool-4, a sample volume of 300 µL was used to increase the volume of the mixture available for downstream applications via the Xpert and Cobas methods, as the number of diluent negative units was only three. For the other pool sizes, 200 µL was enough to test it with both platforms. Each test group was represented in multiple replicates as described above (15 each for the VL and L viral load groups and 10 each for the M and H viral load groups).

Laboratory Testing Procedures

Each of the pooled samples was vortexed for 20 s before testing to ensure proper mixing. Then, 300 µL of this mixture was added to the Xpert cartridge and tested using the XVI Module Xpert PCR machine within 30 minutes of sample addition. Simultaneously, 800 µL of the same mixture was tested with the Cobas assay. In pooling strategies using the Cobas system, a “presumptive positive” result was treated as a confirmed positive result. This is because the pooling of samples can dilute the viral load, making a strong positive result appear weaker, which might result in a “presumptive positive” reading.

Data Quality Assurance

Existing standard operating procedures (SOPs) for each of the Xpert SARS-CoV-2 assays and the Cobas SARS-CoV-2 assay were followed. To maintain quality assurance, laboratory personnel running pooled samples were blinded to individual sample test results, and 10 negative pooled samples were run along with the 200 pooled samples as negative controls. Daily assessments of the completed forms were conducted to ensure data completeness and internal consistency. To ensure the accuracy of the data, 15% of the data were entered a second time by another data clerk in the EPHI.

To ensure the quality of pooling and prevent cross-contamination, all necessary precautions were taken throughout the procedure. These precautions included performing all steps within a Class II A2 biosafety cabinet, which provides a controlled and sterile environment. The transfer of samples was carried out using a calibrated micropipette with sterile pipette tips. Only after pooling the negative samples were the positive samples dispensed into the designated containers. Caution was maintained to avoid creating bubbles, foam or aerosols while mixing. Furthermore, we conducted all the essential steps and sample processing sequentially, starting with samples with VL followed by L, M, and H load groups. This sequential approach was implemented to prevent potential contamination from higher viral load samples to lower viral load samples.

Moreover, the laboratory personnel who conducted the tests on the two PCR machines were blinded to each other’s results. Second, the pooled samples were analyzed immediately to prevent RNA degradation. The selection of samples was aimed at mimicking real-life situations, where the distribution of CT values in the population may vary. This approach allows for a more comprehensive evaluation of the performance of pooled testing in diverse scenarios, capturing the challenges and limitations that may arise when testing samples with different viral load levels.

Data Management and Interpretation Data Entry and Analysis

The data that were entered into Epi Data version 3.1 were exported to Statistical Product and Service Solutions (SPSS) version 28 (IBM Inc., Chicago, USA). Descriptive statistics were used to summarize the study variables, and the overall agreement between positive and negative test results was calculated. The data were then presented in the form of tables, graphs and figures. Cross-tabulation was used to compare the individual final results with the pooled final results. To compare the individual mean CT value with the pooled mean CT value, a paired-t test was utilized. Pearson correlation was applied to determine the extent of linear association between the individual final result and the final result after pooling the samples. The chi-square (χ2) test was used to evaluate the relationship between the final pooled result and the number of pools, as well as the viral load group, for both the Xpert and Cobas SARS-CoV-2 assays. The crude odds ratio was used to assess the association between the number of pools and the final pooled result for both the Xpert and Cobas SARS-CoV-2 assays. Since none of the two assays was treated as a gold standard, sensitivity, specificity, Positive Predictive Value, and Negative Predictive Value were not used for primary comparison. Instead, Overall Percent Agreement (OPA), Percent Positive Agreement (PPA), and Percent Negative Agreement (PNA), with Cohen’s kappa (κ) were used to evaluate the level of agreement between the assays and the pooled results. To ensure long-term storage and preservation of the data collected during the study, two repositories were utilized: the EPHI National Data Management Center and the Research Data Repository of Addis Ababa University.

Operational Definitions

Test agreement refers to the degree of concordance between the two assays (Xpert and Cobas SARS-CoV-2 assays).

CT value: In PCR (polymerase chain reaction), the CT value is the number of cycles required for the fluorescent signal emitted by amplified DNA to cross a set threshold above the background fluorescence, which is directly related to the viral load.

Viral load: refers to the amount or concentration of a virus present in a sample fluid. In this study, the viral load was determined indirectly from CT values as described above. Each designated viral load group is defined in the methodology section. For example, very low viral load (VL) means that samples have CT values between 35.0 and 39.9.

Pooling is an approach that increases the number of individuals who can be tested using the same amount of resources by mixing several samples together in a batch or pooled sample.

Pool size: total number of samples (positive + negative) used for pooling.

Ethical Considerations

This study was conducted according to the Declaration of Helsinki. Ethical approval (DRERC/003/22) was obtained from the Departmental Research Ethics Review Committee (DRERC) of the Department of Microbiology, Immunology and Parasitology, College of Health Sciences of Addis Ababa University. A waiver for informed consent was asked and granted by the EPHI, as the study was conducted on stored samples. A written letter was prepared, and permission was obtained from EPHI to conduct the laboratory tests (Cobas and Xpert).

Results Pooled Nasopharyngeal Specimens Tested via the Xpert SARS-CoV-2 Assay

Among the total 200 pooled samples, 182 (91.0%) samples tested positive according to the Xpert assay. Within each pool size group consisting of 50 pooled samples, the highest detection rates were found in Pool-4 and Pool-6, each with a 47/50 (94.0%) positivity rate. In the VL viral load group consisting of 15 pooled samples for each pool size (Pool-4 to pool-10), only 10/15 (66.7%) and 9/15 (40%) samples were positive from Pool-8 and pool-10, respectively (Table 1). The difference in the positivity rate between the Pool-8 (p=0.005) and pool-10 (p=0.008) groups was statistically significant compared with the initial individual sample. On the other hand, the change in positivity from individual to pooled was not statistically significant for pool-4 or pool-6 (p=0.105) (Table 1).

Table 1 Xpert Assay Results of Different Pool Size Across Viral Load Groups

Importantly, all pooling combinations among the M and H viral load groups yielded positive results regardless of pool size (Figure 2). The overall observed differences in the proportion of positive samples between the different pool sizes among all the viral load groups were not statistically significant, and the p‑value was computed to be 0.463.

Figure 2 Final pooled Xpert SARS-COV-2 assay results in different viral load groups.

Pooled Nasopharyngeal Samples Tested via Cobas for SARS-CoV-2 Detection

Among the 200 pooled samples analyzed via the Cobas assay, the assay was able to detect SARS-CoV-2 in 162 (81%) samples, failing to detect in 38 (19%) of the tests (Table 2). While this assay detected SARS-CoV-2 in all pool sizes of the H viral load group (40/40; 100%), it failed to detect it in one sample from each of the M (39/40; 97.5%) and L (59/60; 98.3%) groups (Table 2 and Figure 3). However, in the VL viral load group, the negative test result rate was significantly high (p values <0.001) among all the pool sizes, the worst result being among the pool size 10, where 9/15 (60.0%) individual samples became negative (Table 2). In fact, the overall difference in the association of the positive detection rate between the viral load groups and pool size was statistically significant (p value <0.001) in the Cobas assay.

Table 2 Cobas Assay Results of Different Pool Size Across Viral Load Groups

Figure 3 Final pooled Cobas results in different viral load groups.

Moreover, the positivity detection rate by the Cobas assay for pool-4, pool-6, and pool-8 was 41/50 (82.0%), whereas 78% (39/50) of the pool-10 group tested positive irrespective of the viral load. However, these differences among the pool sizes (pool-4, pool-6, pool-8, and pool-10) were not statistically significant (p value = 0.943).

Comparison of the Detection Performance of the Xpert and Cobas Assays for SARS-CoV-2 in Pooled Samples

For both of these two assays, an inverse relationship was observed between the positive detection rate and pooled sample size, where with an increased number of pooled samples, there was a gradual decline in the percentage of positive results, indicating a potential decrease in sensitivity as the pool size increased (correlations ranging from 0.93 to 0.99 for Xpert targets and from 0.94 to 0.98 for Cobas targets).

Compared with the Cobas assay, the Xpert assay detected more positive results for both of its gene targets (Table 3). For example, in pool-4 and pool-6, the Xpert assay detected the N2 gene in 47 out of 50 tested samples, whereas the Cobas assay detected the ORF1a target in only 36 out of 50 samples in the same pools. Interestingly, in pool-8, the Cobas assay yielded positive results for the ORF1a target in 37 out of 50 samples. In contrast, there was little difference in the detection of the commonly targeted gene E by either assay, where the detection rate among pool-4 by Xpert was 40/50, whereas it was 41/50 by the Cobas assay.

Table 3 Comparison of Median CT-Values and Correlation for Gene Targets of the Test Assays Across Different Pool Sizes

Analysis of the CT values in the Xpert N2 gene probe for each positive pooled assay revealed median CT values ranging from 33.40–34.00, whereas the median E gene value ranged from 30.15–31.10 across the tested pools. On the other hand, the Cobas ORF1a gene consistently had median CT values between 29.73 and 29.97 across the pools, whereas the Cobas E gene ranged from 30.61–31.24. All four genes show a statistically significant increase in CT value when using pooled samples compared to individual samples as computed from paired-t test (all P-values <0.001) (Table 4). The median difference in the CT values between pooled and individual samples for the Xpert target N2 and E genes ranged from 2.25–3.5 and from 2.25–3.35, respectively, while the variation in the CT values of the Cobas target genes, ORF1a and E, ranged from 1.78–2.15 and from 2.08–2.34, respectively (Table 3). Similarly, comparison of Median CT-values across Pool Sizes show that all four gene targets show a statistically significant difference in Pool median CT values across the four pool sizes (p < 0.05 for all). Coba E gene shows the strongest evidence of difference (p = 0.0186). Moreover, no statistically significant correlation was found between Individual and Pool median CT values for any gene target (all p = 0.3333) (Table 3).

Table 4 Summary of CT-Value Differences of the 4 Target Genes Between Individual and Pooled Samples Regardless of Pool Size

As shown in Table 5 and Figure 4, of the 200 pooled samples tested, 161 samples (80.5%) were positive, and 17 samples (8.5%) negative in both assays. Therefore, the overall percentage agreement (OPA) for all pooled samples was 178/200, which was 89% (95% CI, 86.31% to 91.52%). Moreover, Percent Positive Agreement, Percent Negative Agreement and Cohen’s kappa were calculated for method comparison, where overall they were 93.6%, 60.7%, and 0.553, respectively, with McNemar’s test showing significant asymmetry in discordant pairs (χ2 = 16.41, exact p < 0.001) (McNemar exact p-value <0.001). For individual pool sizes, the OPA for pool-4 was 84% (95% CI, 75.93% to 90.46%); for pool-6, it was 88% (95% CI, 80.90% to 92.12%); for pool-8, it was 90% (95% CI, 87.45% to 94.41%); and for pool-10, it was 86% (95% CI, 78.33% to 93.17%), with κ=0.552, regardless of the viral load group.

Table 5 Comparison of Xpert and Cobas Assays for Pooled Testing of SARS-CoV-2

Figure 4 Comparison of Xpert and Cobas SARS-COV-2 assays for different pool sizes. Positive by both; Negative by both; Xpert positive Cobas Negative; Xpert negative Cabas positive.

Discussion

In this study, our objective was to assess the performance of the Cepheid Xpert and Roche Cobas SARS-CoV-2 assays in detecting SARS-CoV-2 in pooled nasopharyngeal swab samples. The results indicate that pooling samples for SARS-CoV-2 testing can be an effective strategy, as a high proportion (182; 91.0%) of pooled samples tested positive by Cepheid Xpert assay, regardless of their vial load groups and pool sizes. When confirmed positive samples were pooled with confirmed negative samples in pools 4 and 6, there was 94% (47/50) agreement between pooled and confirmed individual positive tests in the Xpert assay, demonstrating a high level of concordance. This level of Xpert SARS-CoV-2 assay sensitivity on pool-4 was also demonstrated by another similar study conducted elsewhere19 despite differences in sample types (both naropharyngeal and oropharyngeal in the latter case), demonstrating the reliability of pooling samples on this specific pool group with the Xpert assay. Similarly, another study reported 100% agreement for both pool-4 and pool-6.20 However, although the sample types and methodology used in the latter study were similar to ours, the sample size was smaller (7 positive and 24 negative samples).20 In contrast, our study increased the sample size (50 positive and 450 negative samples), which may have led to a 6% difference, unlike the 0% difference reported in the above cited study. However, both studies suggested that testing samples in pools of four or six via Xpert retains the accuracy of the test irrespective of the CT value (relative RNA copy number) of the individual sample. Additionally, the Xpert assay successfully detected SARS-CoV-2 in samples diluted into pools of up to 6, with a detection limit below 100 copies/mL.21 This suggests that pooling positive samples into both four and six samples may yield results that are acceptably comparable with those of individual samples for the Xpert assay.

In terms of the effectiveness of pooling by the Xpert assay at a higher pool level, the current study revealed a high level of concordance for pool-8 and pool-10, with agreements of 45/50 (90%) and 43/50 (86%), respectively, irrespective of viral load. This finding is consistent with the FDA’s findings, which indicated 90% agreement for pool eight,22 and another study reported 85% agreement for pool-10 via the Xpert assay.23 This suggests that pooling positive samples even into pool sizes of eight and ten may yield acceptable results, as pooling into lower pool sizes may yield acceptable results, although this may depend on the viral load of the positive samples. This is because the level of positivity in pooled samples was directly influenced by the initial individual sample viral load,19 as was observed in the present study, where higher viral loads, especially in the medium and high groups, resulted in a 100% positivity rate, whereas the low and very low groups showed slightly lower rates of positivity, with 98.3% and 72.0%, respectively. The statistically significant differences in the proportions of positive samples among the viral load groups highlight the influence of the viral load on the accuracy of the SARS-CoV-2 test. Taken together, these findings suggest that testing pooled samples by the Xpert assay up to pool-6 can be a reliable and effective strategy to increase testing capacity; however, pools 8 and 10 may not be suitable, as there might be false negative results in cases where the individual positive samples have low or very low viral loads, as is the case for asymptomatic individuals.

The effect of pooling on the level of viral load reduction in this study was demonstrated by the increase in the median CT values after pooling compared with those at the individual level, which was influenced by the target gene. For example, there was a slight increase in the median CT values as the pool size increased, which was particularly more pronounced for probe E than for probe N2, suggesting that the nonspecific E gene was more affected by dilution than was N2. These results also support previous studies reporting that the N2 gene target appears to be more sensitive (easily diluted) than the E gene target in pooled samples.23,24

With respect to the effectiveness of the Cobas platform in detecting pooled positive samples, our study assessed pooling across four different pool sizes instead of only pool-5, which many studies employ.25,26 The results suggest that pooling for SARS-CoV-2 detection via the Cobas assay remains effective, as the proportion of positive samples (41/50; 82%) showed no variation across pool sizes except for pool-10. However, the performance of the assay in pooling may be influenced by the viral load of the initial positive samples. When pooling at pools 4 and 6, we observed 82.0% agreement (in both pool sizes) between the pooled and individual testing methods via the Cobas assay, indicating reasonable concordance. Barat et al reported 94% sensitivity for detecting positive samples in pool-5 from saliva samples, surpassing our observed sensitivity even compared with pool-4.25 This discrepancy may stem from differences in sample types, where the viral load from saliva is sometimes reported to be greater than that from nasopharyngeal swabs.27,28 This finding shows that pooling strategies, testing methodologies, or population characteristics may influence the performance of the Cobas assay (and Xpert assay for that matter) in pooled samples.

The sample selection method used in our study imitates real-world conditions allowing for a thorough evaluation of pooled testing performance across diverse scenarios, where the viral load level (CT values) in the population varies from high viral load, as found in symptomatic patients, to very low viral load levels, as found in asymptomatic patients, enabling the study to detect the challenges and limitations associated with testing such diverse samples. In this context, while our finding with the Cobas assay from the overall pool-6 samples (regardless of viral load) indicated a sensitivity of only 82%, McMillen et al achieved a 100% sensitivity rate from the same pool-6 approach for nasopharyngeal samples, indicating successful detection of all positive samples.29 This variation in sensitivity between our study and that of McMillen et al could be attributed primarily to differences in sample selection criteria, where McMillen et al selected samples predominantly from CT values below 34 for both the ORF1a and E gene targets, which is indicative of higher viral loads,29 whereas our study included more than half of the samples with CT values between 30 and 39.9. This likely contributed to the higher sensitivity reported by McMillen et al in contrast to our findings. This influence of the initial positive sample viral load on the performance of the Cobas assay was supported by other similar studies. For example, Lee’s study revealed a significant effect on Cobas assay performance when the pool-5 strategy was employed,26 which was particularly evident among samples with low viral loads (CT >30), similar to our findings. A 40% false negative rate was noted among samples with CT > 30, specifically those with lower viral loads, indicating a notable compromise in Cobas performance, particularly among higher CT values or lower viral load initial positive samples.

Taken together, our findings highlight the advantages of pooled sampling, particularly in increasing testing capacity during viral epidemics, especially with platforms such as the Cobas and Xpert assays. Its high throughput and simplicity make it ideal for large-scale deployment, minimizing sample handling requirements and enhancing efficiency. While pooling reduces testing time and workload, considerations regarding the impact of variations in the viral load on assay accuracy across different pool sizes are crucial. Finding the optimal pool size and setting accurate criteria to exclude low-viral-load cases from pooling ensures a balance between efficiency and assay sensitivity.

The findings of this study suggest that pooling samples can be an effective method for detecting SARS-CoV-2 via both the Cobas and Xpert systems by considering the viral load. However, previous studies did not specifically assess the agreement of the assays on pooled samples, which could be an area deserving further assessment. Our study revealed an overall percentage agreement (OPA) of 89% for all pooled samples, indicating moderate agreement between the two assays according to the kappa statistic interpretation of Mary L. McHugh’s interrater reliability.30 The Xpert assay from our study had a higher detection rate than the Cobas assay did, indicating that the Cobas assay may have reported false negative results, especially from pooled samples with high CT values or low viral loads. The superiority of the Xpert assay lies in its incorporation of the N2 target gene instead of the ORF1a gene. This can be seen from the observation that most positive pooled samples detected only by the Xpert assay were primarily positive through N gene amplification. This could be because detecting the smaller sequence of the nucleocapsid gene via the Xpert assay may be more effective than targeting the larger ORF1a gene via the Cobas assay, as stipulated by Tham et al.10 Moreover, Poon et al reported that the N2 gene persists for a longer duration than fragments from the ORF1a gene.24

One of the limitations of this study was that it had limited pool sizes—four, six, eight and ten—due to sample insufficiency and resource constraints. Therefore, more studies and optimization of pooled testing strategies are necessary to further increase the sensitivity and accuracy of pooled testing approaches in various settings and population groups. Another drawback was that the study exclusively analyzed nasopharyngeal samples and utilized an older version of the Xpert CoV-2 assay that targeted only the N and E genes. The use of the newer FDA-approved version of the Xpert CoV-2 assay, which incorporates the RNA-dependent RNA polymerase (RdRp) gene as a third target for SARS-CoV-2 detection, is desirable.31

Conclusion

Pooling up to six nasopharyngeal swab samples was effective with the Xpert assay regardless of the viral load and dilution effects. However, the Cobas assay was impacted by lower viral load samples, regardless of the number of samples pooled. In general, there was moderate agreement regarding the positive detection rate from pooled samples between the two assays. This underscores the importance of considering both pool size and a patent’s state of disease (symptomatic where the viral load is expected to be high vs asymptomatic where the viral load is expected to be low) for assay choice. Overall, these findings suggest that implementing sample pooling can be particularly beneficial in low-resource settings where testing resources are limited, enabling an increase in overall testing capacity, especially with the Xpert assay.

Acknowledgments

We wish to acknowledge Addis Ababa University, the Department of Microbiology, Immunology, and Parasitology, along with the dedicated staff at Ethiopian Public Health Institute. Our sincere gratitude extends to the participants whose samples were the cornerstone of this study.

Disclosure

The author(s) report no conflicts of interest in this work.

References

1. Eberhardt JN, Breuckmann NP, Eberhardt CS. Multi-stage group testing improves efficiency of large-scale COVID-19 screening. J Clin Virol. 2020;128:104382. doi:10.1016/j.jcv.2020.104382

2. Daniel EA, Anbalagan S, Muthuramalingam K, et al. Pooled testing strategies for SARS-CoV-2 diagnosis: a comprehensive review. Diagnostic Microbiol Infect Dis. 2021;101(2):115432. doi:10.1016/j.diagmicrobio.2021.115432

3. Reilly M, Chohan B. Pooled testing for SARS-CoV-2, options for efficiency at scale. Bull World Health Org. 2021;99(10):708–14. doi:10.2471/BLT.20.283093

4. Majid F, Omer SB, Khwaja AI. Optimising SARS-CoV-2 pooled testing for low-resource settings. Lancet Microbe. 2020;1(3):e101–2. doi:10.1016/S2666-5247(20)30056-2

5. De Salazar A, Aguilera A, Trastoy R, et al. Sample pooling for SARS-CoV-2 RT-PCR screening. Clin Microbiol Infect. 2020;26(12):1687.e1–1687.e5. doi:10.1016/j.cmi.2020.09.008

6. Chong BSW, Tran T, Druce J, et al. Sample pooling is a viable strategy for SARS-CoV-2 detection in low-prevalence settings. Pathology. 2020;52(7):796–800. doi:10.1016/j.pathol.2020.09.005

7. Polage CR, Lee MJ, Hubbard C, et al. Assessment of an online tool to simulate the effect of pooled testing for SARS-CoV-2 detection in asymptomatic and symptomatic populations. JAMA Network Open. 2020;3(12):e2031517. doi:10.1001/jamanetworkopen.2020.31517

8. Grobe N, Cherif A, Wang X, et al. Sample pooling: burden or solution? Clin Microbiol Infect. 2021;27(9):1212–1220. doi:10.1016/j.cmi.2021.04.007

9. Arevalo-Rodriguez I, Buitrago-Garcia D, Simancas-Racines D, et al. False-negative results of initial RT-PCR assays for COVID-19: a systematic review. PLoS One. 2020;15(12):e0242958. doi:10.1371/journal.pone.0242958

10. Tham JWM, Ng SC, Chai CN, et al. Parallel testing of 241 clinical nasopharyngeal swabs for the detection of SARS-CoV-2 virus on the Cepheid Xpert Xpress SARS-CoV-2 and the Roche cobas SARS-CoV-2 assays. Clin Chemistry Lab Med. 2021;59(2):e45–8. doi:10.1515/cclm-2020-1338

11. Mutesa L, Ndishimye P, Butera Y, et al. A pooled testing strategy for identifying SARS-CoV-2 at low prevalence. Nature. 2021;589(7841):276–280. doi:10.1038/s41586-020-2885-5

12. Jeong H, Lee J, Cheon S, et al. Experimental and mathematical optimization of a pooling test for detection of SARS-CoV-2 in a population with low viral load. Infect Chemother. 2021;53(1):118. doi:10.3947/ic.2021.0005

13. Williams B. Optimal pooling strategies for laboratory testing. arXiv. 2010.

14. Wacharapluesadee S, Kaewpom T, Ampoot W, et al. Evaluating the efficiency of specimen pooling for PCR-based detection of COVID-19. J Med Virol. 2020;92(10):2193–2199. doi:10.1002/jmv.26005

15. Wijesuriya P, Muthugala M, Herath H. Evaluation of sample pooling method for SARS CoV-2 RNA detection: a laboratory based simulated study. Sri Lankan J Infect Dis. 2022;12(2):1–6.

16. Deka S, Kalita D, Mangla A, et al. Analysis of multi-sample pools in the detection of SARS-CoV-2 RNA for mass screening: an Indian perspective. Indian J Med Microbiol. 2020;38(3–4):451–456. doi:10.4103/ijmm.IJMM_20_273

17. Abid S, Ferjani S, El Moussi A, et al. Assessment of sample pooling for SARS-CoV-2 molecular testing for screening of asymptomatic persons in Tunisia. Diagnostic Microbiol Infect Dis. 2020;98(3):115125. doi:10.1016/j.diagmicrobio.2020.115125

18. Mulu A, Alemayehu DH, Alemu F, et al. Evaluation of sample pooling for screening of SARS CoV-2. PLoS One. 2021;16(2):e0247767. doi:10.1371/journal.pone.0247767

19. Iem V, Xangsayarath P, Chittamany P, et al. Pooling samples to increase testing capacity with Xpert Xpress SARS-CoV-2 during the Covid-19 pandemic in Lao People’s Democratic Republic. PLoS One. 2022;17(9):e0275294. doi:10.1371/journal.pone.0275294

20. Graham M, Williams E, Isles N, et al. Sample pooling on the Cepheid Xpert® Xpress SARS-CoV-2 assay. Diagnostic Microbiol Infect Dis. 2021;99(2):115238. doi:10.1016/j.diagmicrobio.2020.115238

21. Becker MG, Taylor T, Kiazyk S, et al. Recommendations for sample pooling on the Cepheid GeneXpert® system using the Cepheid Xpert® Xpress SARS-CoV-2 assay. PLoS One. 2020;15(11):e0241959. doi:10.1371/journal.pone.0241959

22. cobas®SARS-CoV-2 & Influenza A/B - instructions for use. Silver Spring (MD): US Food and Drug Administration; 2020. Available from: https://www.fda.gov/media/141887/download. Accessed August14, 2026.

23. Procop GW, Tuohy M, Ramsey C, et al. Asymptomatic patient testing after 10:1 pooling using the Xpert Xpress SARS-CoV-2 assay. Am J Clin Pathol. 2021;155(4):522–526. doi:10.1093/ajcp/aqaa273

24. Poon KS, Tee NWS. Comparing the performance of two commercial assays for detection of SARS-CoV-2 in samples with low viral load. J Lab Precis Med. 2021;6:23. doi:10.21037/jlpm-21-17

25. Barat B, Das S, De Giorgi V, et al. Pooled saliva specimens for SARS-CoV-2 testing. J Clin Microbiol. 2021;59(3):e02486–20. doi:10.1128/JCM.02486-20

26. Lee HR, Wang FY, Li JL, et al. Real-world evaluation of a sample pooling strategy for large-scale rapid COVID-19 testing. J Clin Virol. 2022;149:105133. doi:10.1016/j.jcv.2022.105133

27. Gan Y, Du L, Faleti OD, et al. Sample pooling as a strategy of SARS-COV-2 nucleic acid screening increases the false-negative rate. Epidemiology. 2020.

28. Dash GC, Rout UK, Nanda RR, et al. Pooled testing for SARS-CoV-2 infection in an automated high-throughput platform. Clin Lab Analysis. 2021;35(7):e23835. doi:10.1002/jcla.23835

29. McMillen T, Jani K, Babady NE. Evaluation of sample pooling for SARS-CoV-2 RNA detection in nasopharyngeal swabs and salivas on the Roche Cobas 6800. J Clin Virol. 2021;138:104790. doi:10.1016/j.jcv.2021.104790

30. McHugh ML. Interrater reliability: the kappa statistic. Biochem Med. 2012;22(3):276–282. PubMed PMID: 23092060; PubMed Central PMCID: PMC3900052. doi:10.11613/BM.2012.031

31. Cepheid. Xpert Xpress SARS-CoV-2 Plus CE-IVD GeneXpert system with touchscreen: instructions for use. Sunnyvale (CA): Cepheid; 2022. Report nos.: 302–8404. Available from: https://web-support.cepheid.com/Package%20Insert%20Files/Xpert%20Xpress%20SARS-CoV-2%20Plus%20CE-IVD%20GeneXpert%20System%20With%20Touchscreen%20302-8404-EN%20Rev%20A.pdf. Accessed August14, 2026.