Evaluation of RT-qPCR Primer-Probe Sets to Inform Public Health Interventions Based on COVID-19 Sewage Tests.

Sewage surveillance is increasingly employed as a supplementary tool for COVID-19 control. Experiences learnt from large-scale trials could guide better interpretation of the sewage data for public health interventions. Here, we compared the performance of seven commonly used primer-probe sets in RT-qPCR and evaluated the usefulness in the sewage surveillance program in Hong Kong. All selected primer-probe sets reliably detected SARS-CoV-2 in pure water at 7 copies per μL. Sewage matrix did not influence RT-qPCR determination of SARS-CoV-2 concentrated from a small-volume sewage (30 mL) but introduced inhibitory impacts on a large-volume sewage (920 mL) with a ΔCt of 0.2-10.8. Diagnostic performance evaluation in finding COVID-19 cases showed that N1 was the best single primer-probe set, while the ORF1ab set is not recommended. Sewage surveillance using the N1 set for over 3200 samples effectively caught the outbreak trend and, importantly, had a 56% sensitivity and a 96% specificity in uncovering the signal sources from new cases and/or convalescent patients in the community. Our study paves the way for selecting detection primer-probe sets in wider applications in responding to the COVID-19 pandemic.

Introduction

Surveillance of COVID-19 is a key pre-requisite for epidemic preparedness, especially when numbers of symptomatic patients have been largely contained under the impacts of the control strategy and vaccination. Sewage surveillance, which examines the pooled human excreta in particular sewersheds, is an alternative strategy to detect undiagnosed patients in a population served by the sewerage system. The year 2020 witnessed the exponential expansion of COVID-19 sewage surveillance studies, growing from the first report in the Netherlands in April 2020 to the current status of sewage monitoring programs in at least 64 countries worldwide.[1] With appropriate interpretations, sewage surveillance could supplement clinic surveillance in various ways by providing early warning signals for virus emergence,[2,3] tracking the outbreak trend,[4,5] and informing the non-circulation of the virus in the community.[6,7] A key component for the appropriate interpretation of sewage data is the selection of the RT-qPCR assays for SARS-CoV-2 determination.

The analytical sensitivity and efficiency for multiple primer-probe sets have been evaluated in clinical tests, which showed that most primer-probe sets were of highly similar sensitivities,[8] with some sets exhibiting superior sensitivity in diagnostic tests.[9,10] Unlike clinical specimens, sewage consists of various compounds that could inhibit the PCR,[11,12] like humic substances, organic salts, detergents, tannic acids, and other polyphenolic substances as well as degraded plant or tissue materials.[13−15] Such interference effects on the PCR detection of the target,[16,17] termed matrix effects, should be evaluated and considered before embarking on a sewage surveillance program. Besides, there is another dilemma associated with the sewage sample volume and the test sensitivity. Recent studies have focused only on comparing detection rates[18] and stabilities[11] of some primer-probe sets in sewage samples, but to what extent the sewage matrix would impact the RT-qPCR detection across different primer-probe sets remains to be studied.

The ongoing COVID-19 pandemic has catalyzed monitoring projects for SARS-CoV-2 in sewage across time scales and a wide range of geographies, which included examples of large-scale field tests, such as in the US,[19−24] Brail,[25,26] Austria,[27] and the Netherlands,[7] for implementing sewage surveillance to support COVID19 responses. Because most studies on wastewater-based epidemiology of COVID-19 are region-wide surveys by collecting samples at conventional sewage treatment facilities,[28] the interpretation of sewage surveillance data of upstream catchments at the neighborhood level or individual building level by taking into considerations the impacts of concentration and detection methods could help demonstrate the true value of sewage surveillance. The diagnostic performance of different primer-probe sets needs to be assessed for usefulness in sewage surveillance by interpreting sewage testing data with matched demographic, clinical, and epidemiological data across individual neighborhoods.

In this study, seven primer-probe sets[29−32] were selected based on the following considerations: (1) most frequently used in wastewater-based epidemiology for COVID-19 (N1, N2, N3, and E) (Table S1); (2) preferably targeting at the different genetic loci (genes N, E, and ORF1) (Figure a) and at the conserved regions (Figure S1) of SARS-CoV-2; and (3) regionally used in Hong Kong for clinic diagnostics of COVID-19 patients (HKU-N and HKU-ORF1b). To avoid the introduction of different variables caused by the assays per se, we used (1) the same primer-probe concentrations; (2) the same RT-qPCR reagents; (3) the optimum annealing temperature for each set; and (4) the same amplification cycle number and threshold.

Figure 1

Analytical sensitivity of seven primer-probe sets. (a) Locations of seven primer-probe sets in the SARS-CoV-2 genome. (b) Standard curves for seven primer-probe sets tested for three technical replicates with 10-fold dilutions of SARS-CoV-2 RNA. (c) PCR efficiency [left of the dashed line in (c)] and y-intercept Ct values [right of the dashed line in (c)]. Colors indicate the seven tested primer-probe sets. Dashed lines in (b) indicate the Ct value of 40. Error bar amplitude matches the mean ± standard deviation of triplicates.

The aims of our study include the following: (1) laboratory evaluation of all commonly used primer-probe sets in wastewater-based studies for COVID-19; (2) diagnostic effectiveness of these primer-probe sets using real sewage samples collected from sewersheds that had the convalescent patients and/or new cases identified; (3) usefulness of the best-performing N1 set in sewage surveillance using over 3200 samples collected from the implementation trial in Hong Kong; and (4) considerations and suggestions for primer-probe selection to achieve optimal results of informing public health interventions.

Materials and Methods

Quantification for RNA and DNA Standards

SARS-CoV-2 RNA Control 2 (MN908947.3) was purchased from Twist Synthetic. The RNA control was quantified as 7 × 106 copies per μL by ddPCR using the N1 primer-probe set and the one-step RT-ddPCR Advanced Kit for Probes (Bio-Rad). Each reaction consisted of 5 μL of 4× Supermix, 1 μL of DTT, 2 μL of transcriptase, and primers and probe with concentrations of 500 and 250 nM, respectively, topped up with RNase-free water to a final volume of 20 μL. The PCR conditions were as follows: 30 min at 25 °C, 60 min at 50 °C, 10 min at 95 °C, and then 45 cycles of 30 s at 95 °C and 1 min at 55 °C, followed by 10 min at 98 °C and holding at 4 °C.

The SARS-CoV-2 RNA extracted from the heat-inactivated SARS-CoV-2 virus was quantified as 1.84 × 106 copies per μL by the N1 set and its standard curve under the same conditions as the one-step RT-qPCR. N1 gene was synthesized to commercial pUC19 vector by BGI (Beijing) to obtain a DNA standard. The DNA standard was quantified using the Qubit dsDNA HS assay kit (ThermoFisher), and the copy number was calculated using Avogadro’s number. The standard curve was prepared by serial 10-fold dilutions of DNA standards (ranging from 10 to 105 copies per μL).

Matrix Effect Evaluation

A sewage sample was deemed negative if no viral sources were identified in its respective sewershed, and it was tested in duplicate RT-qPCR reactions as negative using seven primer-probe sets. All SARS-CoV-2 RNA-spiked sewage RNA extracts were tested in triplicate (small-volume treatment) or in duplicate (large-volume treatment) RT-qPCR reactions.

To evaluate matrix effects, a large volume (920 mL) of 16 negative sewages or a small volume (30 mL) of 30 negative sewage samples collected from community manholes or sewage pumping stations were pasteurized at 60 °C for 30 min. All the inactivated samples were firstly centrifuged at 4750g for 30 min to remove large particles in sewage samples. And then the 30 mL supernatant from small-volume sewage samples was directly concentrated by ultracentrifugation at 150,000 g for 1 h. While the 920 mL supernatant from large-volume sewage samples was concentrated at 20,000g for 1 h, and the obtained concentrated pellet was resuspended using ∼30 mL phosphate-buffered saline (PBS) for further ultracentrifugation at 150,000g for 1 h. The final concentrated pellets from small-volume and large-volume samples were resuspended by ∼200 μL PBS to extract RNA.[100]

For 30 mL sewage sample, a concentration fold of 750 (from 30 mL sewage to 40 μL RNA extracts) was applied. We spiked 3 μL 10-fold diluted SARS-CoV-2 RNA (7 × 101 to 7 × 105 viral RNA copies per μL) into 27 μL pooled RNA from 30 negative sewage samples of a small volume (30 mL per sample) to obtain SARS-CoV-2 RNA-spiked sewage RNA extracts with the concentration ranging from 7 to 7 × 104 copies per μL. The stabilities and sensitivities of primer-probe sets were tested by 10-fold dilutions (ranging from 7 to 7 × 104 viral RNA copies per μL) of SARS-CoV-2 RNA as well as the SARS-CoV-2 RNA-spiked sewage RNA extracts.

For the large-volume sewage sample, the starting volume of 920 mL and the RNA elution volume of 40 μL resulted in a concentration fold of 23,000. We spiked 3 μL 10-fold dilutions of viral RNA (ranging from 2 × 101 to 2 × 106 viral RNA copies per μL) extracted from heat-inactivated SARS-CoV-2 into 27 μL pooled RNA extracted from 16 negative sewage samples to generate SARS-CoV-2 RNA-spiked sewage RNA extracts with the concentration ranging from 2 to 2 × 105 copies per μL. The matrix effect was calculated by the following equation

Pooling Samples for Diagnostic Evaluation

We made 20 independent RNA pools from 56 “case positive” sewage samples and 13 independent RNA pools from 39 “case negative” sewage samples. All sewage samples were 3 h composite samples collected from municipal manholes at either the neighborhood level or the individual building level, with the serving population of each sampling site ranging from 357 to 207,836 during December 2020 to March 2021. Sewage samples were delivered to the lab with ice immediately after collection, followed by pasteurization, viral concentration, and viral RNA extraction using previously described methods.[3] Finally, 40 μL RNA samples were obtained from each 30 mL sewage sample. The extracted RNA samples were stored in −80 °C until further use.

We pooled RNA samples extracted from 2 or 3 individual sewage samples with equal volume to create one independent RNA pool. The details about used sewage samples are provided in Table S6. We used amplicon-based sequencing via an Illumina sequencing platform following a previously described method[33] to confirm the presence of a near-complete SARS-CoV-2 genome in one positive sample pool.

Sewage Samples for the Usefulness of the Best-Performing N1 Set

In total, 2078 samples collected from over 100 stationary sites with a population size ranging from 4500 to 289,547 were detected to analyze the utility in showing pandemic trends using N1. Besides, additional reanalyses were performed using the sub-data set (1130 samples) in our previous study,[34] which focuses on the practical effectiveness of using sewage data to take actions. These samples were collected during a COVID-19 outbreak in Hong Kong (October 13, 2020, to July 6, 2021) from community manholes and sewage pumping stations. Sewage samples were collected and analyzed for SARS-CoV-2 as described above.

Statistical Analyses

For the analysis of diagnostic performance, first, we assumed that RT-qPCR-confirmed SARS-CoV-2 infected individuals, either convalescent patients or new cases, living in the sewershed under survey will shed virus into the sewage. Second, an evaluation period of 7 days before or after the sewage sampling date was applied. Following the above assumptions, a sewage sample was deemed as “case positive” when its sewershed had the convalescent patients within 7 days before sampling or new cases within 7 days after sampling. The population size of the individual sewershed was provided by the Environmental Protection Department, The Government of the Hong Kong SAR. The case number was counted using the clinical data provided by the Hospital Authority of Hong Kong.

Sensitivity, specificity, negative predictive rate (NPV), and positive predictive rate (PPV) were calculated using the number of false positive sampling sites (sites without confirmed patients or convalescent patients incorrectly tested as positive), true positive sampling sites (TPS, sites with confirmed patients and/or convalescent patients correctly tested as positive), false negative sampling sites (FNS, sites with confirmed patients or convalescent patients incorrectly tested as negative), and true negative sampling sites (TNS, sites without confirmed patients or convalescent patients correctly tested as negative) according to the following formulas

Statistical analyses like linear regression were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). The Venn diagram was plotted using the R package “VennDiagram”.[35]

Results

Analytical Sensitivity and Amplification Efficiency of the Seven Primer-Probe Sets

The evaluation of RT-qPCR assays for SARS-CoV-2 RNA detection using the selected seven primer-probe sets was initially performed with the standard curves obtained from 10-fold serial dilutions of SARS-CoV-2 RNA standards (Figure b and 1c). We found that all the seven primer-probe sets can detect SARS-CoV-2 RNA at 7 viral RNA copies per μL with amplification efficiencies above 90%. The ORF1ab set, however, had an efficiency below 90% (∼81%), which did not meet the criteria for efficient qPCR assays[36] and may introduce a large quantification bias. The analytical sensitivity of the assays, as indicated by the y-intercept Ct values when tested with 1 viral RNA copy per μL, ranged from 34 to 38 for the seven primer-probe sets (Figure c).

Matrix Effects for Different Primer-Probe Sets

To assess the matrix effect using seven primer-probe sets in RT-qPCR determination of SARS-CoV-2 in sewage samples, the same amount of SARS-CoV-2 RNA standards was spiked into (1) water, (2) pooled RNA extracts from 30 negative sewage samples starting with a small volume of 30 mL, and (3) pooled RNA extracts from 16 samples of large-volume (920 mL) negative sewage samples. The matrix effect was assessed based on the SARS-CoV-2 RNA determination in the above three matrices.

For small-volume (30 mL) sewage samples, no significant matrix effects were observed in RT-qPCR detection, compared with the SARS-CoV-2 RNA in water solution. The difference in Ct values between RNA in the water solution and RNA in the sewage matrix was less than 1 Ct value for all primer-probe sets (Figure a). The calculated matrix effect ranged from −15 to 42% (Table S3). Also, we found that less inhibitory effects occurred for all sets in the small-volume samples. At 7 copies per μL, all sets were 100% sensitive and were able to detect the SARS-CoV-2 signal in the sewage matrix.

Figure 2

Analysis of matrix effects of real sewage samples for seven primer-probe sets. The Ct value, Ct difference, and matrix effects between SARS-CoV-2 virus RNA and virus RNA with sewage matrix in small-volume samples (a) and large-volume samples (b) for seven primer-probe sets. Colors indicate the seven tested primer-probe sets. Error bar amplitude matches the mean ± standard deviation.

For the large-volume (920 mL) sewage samples, the Ct differences between the water solution and the sewage matrix ranged from 0.2 to 10.8, leading to the calculated matrix effects from −100 to −9% (Table S4). Moreover, for all primer-probe sets, the sewage matrix in large-volume samples caused significant inhibition effects. At the detection limit of 2 viral RNA copies per μL, our data showed that only three primer-probe sets (N1, N2, and N3) were able to detect SARS-CoV-2 RNA (Ct < 45) in the sewage matrix. At 20 viral RNA copies per μL, all primer-probe sets detected signals (Figure b).

Detective Efficiency Using Different Primer-Probe Sets for Stored Samples

To evaluate the detective efficiency of SARS-CoV-2 virus in the stored sewage sample, we spiked the heat-inactivated SARS-CoV-2 virus into the negative sewage samples and then stored them at ambient temperature (25 °C) and determined viral RNA amount in 30 mL sub-samples at nine time points (day 0, 1, 2, 12, 18, 25, 32, 47, and 69) using the selected seven primer-probe sets. It was found that six out of the seven sets could detect SARS-CoV-2 signals at all the sampling points even after 69 days (T90 (the time for the reduction of 90% SARS-COV-2) was less than 3 days for all primer-probe sets), while ORF1ab could not detect any signal after 2 days (Figure S2). The decay of SARS-CoV-2 RNA in the sewage sample followed the one-phase decay model (Figure S2) (Table S5). Besides, among different primer-probe sets, N1 and N3 sets had the highest sensitivities as demonstrated by their lower Ct values (Figure S2h) and still showed better reproducibility even in the later period of the decay.

Diagnostic Performance of Sewage Tests in Finding Cases

To evaluate the diagnostic sensitivity of the selected primer-probe sets to discern true “case positive” sewage sample, we compared the percentage of sites that had agreed “sewage positive” and “case positive” using the seven primer-probe sets. Twenty independent RNA pools of sewage samples taken from 59 sampling sites and 2 positive RNA samples extracted from SARS-CoV-2 virus spiked-sewage samples were tested using seven primer-probe sets (Table S7). All these “case positive” sampling sites had the contributing sources of viral signal in corresponding sewersheds, covering residential populations ranging from 357 to 207,836. We found that most of the pooled “case positive” samples (21 out of 22) had detectable SARS-CoV-2 RNA due to at least one primer-probe set being positive (<45 Ct; Table S7). Only one sample of the “case positive” sample was tested “sewage negative” by all assays, implying a failure of the sewage test to catch the signal, probably due to many reasons, including non-matching of the time windows of shedding and sampling, dilution of the low virus concentration along the sewer, sampling randomness, viral RNA degradation in the sewer, and so forth.

There was a wide range of diagnostic sensitivities among the different assays. The highest levels for a single primer-probe set of 68% were obtained by the N1 or N3 sets. The majority of the samples had lower Ctt values using the N1 set compared to the other primer-probe sets (Figure a), showing that the N1 set was the most sensitive for virus detection among the evaluated primer-probe sets. Except for one sample that was not detected by any of the seven primer-probe sets, for the remaining 21 samples, when the virus was not detected by the N1 set (>45 Ctt), it was detected by the N3 set, suggesting that the N3 assay is a useful supplement to the N1 assay (Figure b). Indeed, the diagnostic sensitivity was increased to 86% when we combined the results of N1 and N3 sets (Table S9), providing presence-absence data for case finding with a sufficient high probability.

Figure 3

Performance comparison of primer-probe sets using “case positive” samples. (a) Ct value detected by five primer-probe sets. (b) Venn diagram displaying the relationship of five primer-probe sets based on detectable signals of the samples of “case positive”. (c) Parallel trends between the 4-day moving average line of sewage detection rate based on the N1 set and the surge and decline of clinical cases in the 4th wave of COVID-19 in Hong Kong. Transient change in social distancing rules is also indicated in the figure.

Assays using the N2, E, and HKU-ORF1b sets had a diagnostic sensitivity of 55, 50, and 41%, respectively (Table ). Results of N2 and E were comparable in terms of the sensitivity, and both provided confirmatory diagnoses of the results, when N1 being positive and N3 being negative or the other way around. Assays using HKU-ORF1b performed less well compared with N2 and E, and we found that when the HKU-ORF1b was positive for SARS-CoV-2 detection, each instance was paired with N2 detected, showing that the HKU-ORF1b detection was overlapped with the N2 results (Figure b). Both HKU-N and ORF1ab lacked sensitivity in identifying “case positive” samples. Especially, we found that only 2 out of 22 “case positive” samples were deemed “sewage positive” owing to positive signals of ORF1ab, implying its low diagnostic sensitivity in testing SARS-CoV-2 RNA extracted from sewage samples.

Table 1

Diagnostic Performance for Seven Primer-Probe Sets and Correlations between Positive Results of SARS-CoV-2 RNA in Sewage and New Cases and/orConvalescent Patients Based on an Evaluation Period of 7 Days

	seven primer-probe sets (na= 35, Pb: 22, Nc: 13)							best-performing set (nd= 1130)
analysis tests	N1	N2	N3	E	HKU-N	ORF1ab	HKU-ORF1b	N1
% sensitivity	68	55	68	50	36	9	41	56
% specificity	100	100	85	77	100	100	92	96
%PPV	100	100	88	79	100	100	90	79
%NPV	65	57	61	48	48	39	48	90
% false positive rate	0	0	15	23	0	0	8	4
% false negative rate	32	45	32	50	64	91	59	44

The number of pool samples.

The number of pool positive samples.

The number of pool negative samples.

The number of real sewage samples.

Following the same principles of diagnostic sensitivity evaluations, we also investigated whether the selected seven primer-probe sets would have differences in diagnostic specificity using 13 independent pools of “case negative” sewage samples (Table S8). From the 13 pooled sewage samples taken from 39 sampling sites without identifying the contributing sources of the viral signals, we determined the occurrence of “false positive” of different primer-probe sets. Despite the differences in diagnostic sensitivity, all evaluated primer-probe sets had good to excellent specificity based on the test results of “case negative” sewage samples. Our data showed a specificity of 100% for four primer-probe sets, including N1, N2, HKU-N, and ORF1ab (Table ).

We observed weak positive signals for “case negative” sewage samples in several assays using N3, E, and HKU-ORF1b sets. All of them were only positive by one reaction of the RT-qPCR technical duplicates for each set. Only one “case negative” sewage sample was tested “sewage positive” by two individual primer-probe sets. These marginal detections suggest the possibility of non-specific amplifications in the matrices of sewage samples. Of the 6 “false positive” results, 2 were seen in N3 and had Ct values of 37.2–37.4, 3 were seen in E and had Ct values of 39.3–40.0, and 1 was seen in HKU-ORF1b with a Ct value of 36.9. Changing the Ct cut-off value for these three sets would increase their test specificity but at the cost of lowering the percentage of “sewage positive” samples. For example, changing the positive Ct value to 39 for the E set to achieve 100% specificity decreasing its percent positivity from 50 to 36.

Usefulness of the Best-Performing N1 Set Using Sewage Samples Collected in Hong Kong

COVID-19 sewage surveillance using the best-performing N1 set showed its utility in revealing the community trends in Hong Kong. During October 13, 2020, to July 6, 2021, 2078 samples were collected from the systematic routine sewage monitoring program in Hong Kong, which covered over 100 stationary sampling sites and provided early warning signals of COVID-19 re-emergence for Hong Kong. With the use of the N1 primer set, the sewage tests had effectively caught the rising trend of clinical cases in the 4th wave of Hong Kong since mid-November 2020. Moreover, a strong correlation between the decline in the overall COVID-19 outbreak in the territory and the area-wide negative sewage testing results was observed (Figure c).

In addition, a total of 1130 samples were also collected from 485 sites of individual buildings and estates during December 14, 2020, to March 4, 2021,[34] including 97 “case positive” sites and 388 “case negative” sites. Additional reanalys is results showed that using the N1 set had a sensitivity of 56%, a specificity of 96%, a PPV of 79%, and an NPV of 90% for distinguishing and identifying “case positive” and “case negative” sampling sites (Table ). This demonstrates the value of sewage surveillance using the N1 set on informing public health interventions within the community.

Discussion

To date, no single or combined primer-probe sets have been accepted as standard assays for COVID-19 sewage testing though there are some general recommendations of N1, N2, and E.[37] When selecting an appropriate SARS-CoV-2 RT-qPCR primer-probe set, the primary consideration is the analytical sensitivity and amplification efficiency.[38] Our study described the analytical sensitivity and amplification efficiency of the seven primer-probe sets in assays with and without the presence of inhibitors from sewage samples. The assays using synthetic SARS-CoV-2 RNA showed similar results among most of the assays, except for the ORF1ab set, indicating the comparable performance of different primer-probe sets when there is no matrix effect.

The small-volume swage samples have almost no matrix effects for the seven primer-probe sets on RT-qPCR detection. For large-volume sewage samples, significant inhibitory matrix effects occurred for all sets, probably due to more complex compounds consisting of PCR inhibitors[12,14] concentrated from the large amount of sewage samples. We also found that for large-volume sewage samples, the TaqMan reagent outcompeted the Bio-Rad iTaq reagent, although the two reagent buffers performed similarly for the small-volume samples or the pure RNA (Figure S3). Because the extent of impacts posed by the sewage matrix on the RT-qPCR detection could vary with the changes in matrix components, further efforts are needed to quantify the matrix effects on RT-qPCR determination associated with other sewage types collected across spatial scales or with different conditions.

For SARS-CoV-2 in large-volume samples, assays using the N1, N2, and N3 sets, whose detection limits were 2 copies per μL, continued to perform better than the other four sets, which had a detection limit of 20 copies per reaction. N1, N2, and N3 sets targeted at three different regions in the nucleocapsid (N) gene of coronaviruses (Table S2), which encodes a structural protein that binds to the viral RNA for providing stability.[39] It was previously reported that N gene sub-genomic mRNAs were produced in a higher amount than others during coronavirus replication,[40] which might enhance the analytical sensitivity of N-gene-based RT-qPCR assays.

In addition to the analytical detection performance of the seven sets of primer-probes, in this study, we also evaluated their diagnostic performance by assessing the trade-off between diagnostic sensitivity in “case positive” sewage samples and diagnostic specificity in “case negative” samples for recommendations on sewage surveillance under different epidemiological contexts.

Assays using the N1 set were observed as the standout assay, achieving the highest test sensitivity of 68% and specificity of 100% for individual primer-probe sets. Our observation was consistent with the growing evidence that the N1 set tended to have a high sensitivity in the virus detection.[18,41,42] Besides, additional study also validated the efficiency of N1 for public health surveillance on a university campus.[43] We recommend using the N1 primer-probe set for routine monitoring of the SARS-CoV-2 circulation in the community when the outbreak is largely contained, with the following considerations: (1) the tolerance of analytical sensitivity to the matrix in sewage; (2) high diagnostic sensitivity and specificity for the initial screening; and (3) quick diagnostic test by a single primer-probe set to reduce the practical working load and being compatible with the need for adequate large-scale surveillance. Our results also indicated that the ORF1ab set had its extremely low sensitivity (9%) in discerning “case positive” sewage samples and it was not recommended.

The diagnostic sensitivity of assays could be increased to 86% by considering the combination of the results of N1 and N3 sets albeit at the expense of the test specificity. There could be a role for the combination of N1 and N3 sets, especially in settings where precautionary actions are needed to identify the hotspot places during a pandemic for follow-up public health interventions (such as compulsory individual testing via RT-PCR). Because N1 and N2 sets were specifically designed for SARS-CoV-2 and the N3 set was intentionally designed for a more universal detection of closely related viruses within the subgenus Sarbecovirus, including SARS-CoV, SARS-CoV-2, and bat and civet SARS-CoVs,[44] the combination of N1 and N3 sets could thus reduce the false negative results due to the occurrence of polymorphisms or mutations within the N1 target sequences as the virus evolves over time. In addition, the results of the E set for the use in confirming the disagreement between the results of N1 and N3 sets achieved a high sensitivity of 64% and specificity of 100%. It can be used for case ascertainment in settings where the infection incidence is low to serve the purposes like the early detection of undiagnosed cases or the indication of the non-circulation of SASR-CoV-2 in the community.

There are limitations in our study. First, our data on diagnostic performance characteristics of COVID-19 sewage tests could be biased owing to the limited number of sewersheds surveyed. Second, there are no standard methods to ascertain our expectation that all sewage samples collected from sewersheds with or without COVID-19 patients are true positive samples or true negative samples. Nevertheless, using whole-genome sequencing, we obtained near-full coverage of the SARS-CoV-2 genome in one selected “case positive” sewage sample, validating our expectation about the association (Figure S4). Third, we do not know the extent to which the results by the sewage test can be impacted by the uncertainties caused by the fecal viral shedding amount, sampling randomness in the sewer, viral RNA degradation, and so on. Our recent study has also showed that RT-qPCR from sewage samples did miss COVID-19 cases even when sewage tests were performed at the individual building level,[45] reinforcing the importance of considering such uncertainties in interpreting negative sewage test results. Fourth, the diagnostic sensitivity can be impacted by the sampling strategy of the sewage tests. Our data were all based on 3 h composite samples with 15 min interval. A sampling strategy using a longer sampling period, such as 24 h, would result in a higher probability in catching the viruses shed by the patients and higher diagnostic sensitivity.[200] Additionally, it is necessary to redesign or re-evaluate the common primer-probe sets, when the newly emerging variants carry ever more mutations.

A consensus is emerging that COVID-19 sewage tests can be used as a supplemental tool for the COVID-19 surveillance.[41,46,47] However, the best practices for implementing sewage testing and the methods to standardize test data interpretation still need to be defined. Our study rigorously compared the analytical characteristics of selected primer-probe sets for COVID-19 sewage tests using standardized or spiked-virus samples. Additionally, we performed diagnostic evaluations using real sewage samples from individual sewersheds during a COVID-19 outbreak in Hong Kong and discussed the selection of a single assay or combinations for deploying sewage tests to identify undiagnosed COVID-19 cases. Importantly, the usefulness of best-performing N1 set in catching community trends and finding cases was analyzed using over 3200 sewage samples. Our study could enhance the sewage surveillance to inform public health interventions for the control of COVID-19.