Efficacy of SARS-CoV-2 wastewater surveillance for detection of COVID-19 at a residential private college

Abstract Many colleges and universities utilized wastewater surveillance testing for SARS-CoV-2 RNA as a tool to help monitor and mitigate the COVID-19 pandemic on campuses across the USA during the 2020–2021 academic year. We sought to assess the efficacy of one such program by analyzing data on relative wastewater RNA levels from residential buildings in relation to SARS-CoV-2 cases identified through individual surveillance testing, conducted largely independent of wastewater results. Almost 80% of the cases on campus were associated with positive wastewater tests, resulting in an overall positive predictive value of 79% (Chi square 48.1, Df = 1, P < 0.001). However, half of the positive wastewater samples occurred in the two weeks following the return of a student to the residence hall following the 10-day isolation period, and therefore were not useful in predicting new infections. When these samples were excluded, the positive predictive value of a positive wastewater sample was 54%. Overall, we conclude that the continued shedding of viral RNA by patients past the time of potential transmission confounds the identification of new cases using wastewater surveillance, and decreases its effectiveness in managing SARS-CoV-2 infections on a residential college campus.


Introduction
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic clearly poses significant challenges for many economic sectors. Prior to the introduction of vaccines, and even postintroduction, many businesses and institutions that rely on inperson meetings struggled to identify best practices that could allow for continued operation. In some instances, especially during government-mandated 'lockdowns,' institutions transitioned to remote operations. However, in many cases, remote operation was either not sustainable or failed to deliver satisfactory outcomes, leading these institutions to consider means for returning to inperson operation. These challenges continue to be felt acutely among institutions of higher education, particularly those with a sizeable residential component, as the high population density of dormitories, especially those with shared bathroom facilities, create conditions that can facilitate rapid transmission of respiratory pathogens.
As data emerged that SARS-CoV-2 can infect cells in the gastrointestinal tract (Lee et al. 2020, and that SARS-CoV-2 RNA can be detected in patient stool and other bodily fluids that may be deposited in wastewater , Crank et al. 2022, including from patients prior to the onset of clinical symptoms ), many groups began investigating whether detection of SARS-CoV-2 RNA in wastewater could be used as a means of monitoring communities for emerging infections, or of directing public health responses. Epidemiological wastewater surveillance had previously been utilized for poliovirus, norovirus, and rotavirus (Zurbriggen et al. 2008, Lodder et al. 2012, Santiso-Bellón et al. 2020, and gave strong indica-tion that similar approaches could be beneficial, either to directly assess disease burden in a population, identify areas where the SARS-CoV-2 was newly introduced, or monitor changes in transmission status over time and thereby assess the effects of community interventions (Sims and Kasprzyk-Hordern 2020). The utility of these approaches became even more evident as data emerged correlating levels of SARS-CoV-2 RNA in wastewater with number of cases in the sewershed (Nemudryi et al. 2020, Wu et al. 2020, Weidhaas et al. 2021, with wastewater RNA levels increasing several days prior to an increase in positive test results indicative of SARS-CoV-2 infections (Medema et al. 2020, Peccia et al. 2020. Based on these reports, in the lead up to the Fall 2020 academic semester, many colleges and universities in the US sought to implement wastewater surveillance as a tool to manage the pandemic, either as a means to detect the presence of cases within the campus population, to monitor infection levels on campus, or to target particular residence halls for individual or pooled patient testing to directly identify infected individuals (Betancourt et al. 2021, Gibas et al. 2021, Harris-Lovett et al. 2021, Karthikeyan et al. 2021, Scott et al. 2021, Sharkey et al. 2021. Since essentially all of these systems were implemented de novo, there was a lot of variability in how they were set up and used. Some institutions monitored wastewater via campus or municipal wastewater treatment plants, whereas others implemented systems to capture samples from individual residence halls or sewer lines. Some institutions collected samples that were then processed by commercial laboratories, whereas others handled their samples in house. Among the strategies communicated by various institutions, there was considerable variation in the timing and frequency of collection, and in the number of campus locations that were sampled. Given this variability, it is important to assess the efficacy of the various approaches, to identify best practices and help inform continued development and implementation of wastewater surveillance strategies as a pandemic response (Harris-Lovett et al. 2021).
In the summer of 2020, we developed and implemented a wastewater surveillance system as one piece of a comprehensive plan to facilitate the return of students to campus residences at Colgate University. Wastewater was assayed via 24-hour composite samples taken 2-3 times weekly at each of seven campus locations (sewer lines or sewage lift stations) and the samples were tested for the presence of SARS-CoV-2 RNA by quantitative reverse-transcriptase PCR (qRT-PCR). Simultaneously, students were regularly tested for SARS-CoV-2 infection by anterior nares swabs, using a commercial PCR-based testing provider. We obtained de-identified information on student positive cases, the Ct value of the positive tests, and the time of removal and return to the residence hall for isolation, and analyzed this data with respect to the detection of SARS-CoV-2 in the waste stream. In the fall and spring academic semesters, Colgate identified a total of 107 student cases of SARS-CoV-2 infection. Almost 80% of these cases were associated with positive wastewater tests, resulting in an overall positive predictive value of 79.31%. However, half of the positive wastewater samples occurred in the two weeks following the return of a student to the residence hall following a 10-day isolation period (starting from either date of symptom onset or date of positive test for asymptomatic infections). Upon reanalysis of the data following exclusion of these samples, the positive predictive value of a positive wastewater sample was determined to be 53.85%. Overall, we conclude that the continued shedding of viral RNA by patients past the time of potential transmission confounds the identification of new cases using wastewater surveillance, and decreases its effectiveness in managing SARS-CoV-2 infections on a residential college campus.

Study site
Colgate University is a private residential undergraduate college with around 3000 students, situated in the rural community of Hamilton, NY (population 3814). For the 2020-2021 academic year, students had the choice of remote or in-person instruction, and over 2600 students chose to be in residence, with 2333 students in campus residence halls in the fall semester and 2280 students in campus residence halls in the spring semester (Table 1), and the rest in off-campus (private) housing in the village of Hamilton.

Wastewater sample collection
Wastewater samples were collected at seven locations on the Colgate University campus (Table 1; Fig. 1) using three ISCO GLS compact composite samplers (Teledyne ISCO, Lincoln, NE). Five of the seven locations (GH, SPW, BAE, CD, BC) were in sewer access holes, while two were at lift stations (PL, TH). A 24 h composite sample (50 ml every 20 min, maintained at ambient air temperature) was collected at around 10 AM each weekday, and the samplers were rotated among the collection points at that time, resulting in each collection point being sampled an average of 2.1 times per week. Three of the access points (GH, BC, and TH) required external ground-level storage of the sampler; at these locations, a single draw was obtained during the times when the ambient air temperature was too low to permit composite sampling (intermittently between Jan. and Mar.). Composite samples were mixed by agitation and 250 ml of the total volume of each sample was collected and returned to the lab, where it was stored at 4 • C until processing. The interior tubing of the samplers was rinsed with 10% bleach followed by distilled water after each sample collection. Dedicated collection hoses and filters remained at each location and were rinsed with distilled water after each collection and with 10% bleach on a regular basis.

Ultracentrifugation and RNA extraction
RNA was extracted following the method of Wilder et al. 2021(Wilder et al. 2021. All steps prior to RNA extraction were conducted in BSL2 conditions, as approved by the Colgate University Institutional Biosafety Committee. Briefly, 26 ml of composite sample was layered over 12 ml of sucrose cushion (50% sucrose in 20 mM Tris-HCl (pH 7.0), 100 mM NaCl, and 2 mM EDTA) in a 25 × 89 mm ultracentrifuge tube (maximum capacity, 48 ml; Beckman Coulter, Brea, CA). The samples were then centrifuged in a SW28 Ti rotor (Beckman Coulter) at 150 000 x g in a L7-55 ultracentrifuge (Beckman) for 45 m at 4 • C. The supernatant was decanted and the pellets were resuspended in 200 μl of PBS and transferred to a 1.7 ml microcentrifuge tube. RNA was extracted from each 200 μl sample either using an AllPrep PowerViral DNA/RNA kit (Qiagen Biosciences) according to manufacturer's protocol (with the omission of the optional bead beating step), or using Trizol (1 ml; Invitrogen/ThermoFisher) according to manufacturer's protocol. RNA was eluted in 50 μl of RNase-free water and stored (when necessary prior to analysis) at −80 • C. Both methods yielded similar RNA concentrations, total yield, and RNA quality, as determined by NanoDrop via 260/280 ratio, based on pilot runs using duplicate samples. Samples (5 μl) were treated with DNase I (Invitrogen/ThermoFisher) according to the manufacturer's protocol. DNase-treated samples were then directly used for quantitative reverse transcriptase PCR (qRT-PCR).

Quantification of relative wastewater SARS-CoV-2 RNA levels
Levels of SARS-CoV-2 RNA in undiluted samples were detected by one-step multiplex TaqMan qRT-PCR, using TaqPath 1-step Multiplex Master Mix (ThermoFisher) according to manufacturer's protocol. qPCR reactions were performed on a QuantStudio5™ under fast conditions (25 • C for 2 min; 53 • C for 10 min; 95 • C for 2 min; followed by 40 cycles of 95 • C for 3 seconds and 60 • C for 30 s. Baseline and threshold settings were determined automatically). SARS-CoV-2 RNA was detected using the 2019-nCoV CDC RUO kit (Integrated DNA Technologies; IDT), containing primers amplifying two regions in the SARS-CoV-2 nucleocapsid (N) gene (N1 and N2). A control plasmid containing the complete nucleocapsid gene from SARS-CoV-2 (IDT), as well as a synthetic SARS-CoV-2 RNA (Applied Biological Materials, Richmond, BC, Canada), were used as positive controls for isolation procedures and for qRT-PCR reactions. In optimization experiments, standard curves using the synthetic SARS-CoV-2 RNA gave an R 2 value of 0.98, with a slope of −3.659 and an efficiency of 99.77%. Based on this analysis, we set our Cq cutoff for SARS-CoV-2 positivity at 38. A no-template control and a sample blank was included for each primer set during each qRT-PCR run; the full run was repeated if a negative control resulted in amplification above the threshold level. Extraction controls using water were also performed on a regular basis. To normalize for differences in biological material in each sample, extracted RNA was also probed using TaqMan primers specific for human RNAse P (2019-nCoV CDC RUO primers and probe kit; (IDT); as used in (Stahl et al. 2021)), and for the crAssphage bacteriophage (Wilder et al. 2021). No specific positive controls of RNAse P or crAssphage were available, so purified amplicons were used to assess the amplification, as described (Wilder et al. 2021). Initial samples showed similar trends in RNAse P Ct values and crAssphage Ct values, which did not affect the normalized SARS-CoV-2 RNA levels. Therefore, to conserve resources only RNAse P Ct values were used for normalizing SARS-CoV-2 N gene multiplex Ct values, using the Ct method, to obtain a relative value for SARS-CoV-2 RNA for each sample. This value was reported to the Colgate University Health Analytics Team.

Individual SARS-CoV-2 surveillance testing
For both academic semesters, all Colgate University students were required to submit negative results from a home PCR-based SARS-CoV-2 test taken 7 days prior to arrival at campus. Students, along with all employees, were then tested using a PCR-based test at the time of student arrival to campus. All students then entered a mandatory 14-day quarantine period. Students were tested a second time 7-9 days post-arrival and were not released from quarantine until the results from the second test had been returned. Following quarantine, the whole campus population (students and employees) were randomly selected for surveillance testing at a rate of 6% of each residence hall/area or campus employee population per week (fall semester), or scheduled for surveillance testing once every 2 weeks (spring semester). In the fall semester, 50-70 additional students were selected each week for targeted surveillance testing based on wastewater SARS-CoV-2 results, identified positive cases within specific residence halls, or other events indicating increased risk of transmission (bringing the total percentage of the student population tested up to around 10% per week). Targeted surveillance testing was not widely performed during the spring semester, given the increased frequency of baseline surveillance testing. Patient samples were collected from the anterior nares by Colgate University Student Health Services and volunteers, and were shipped to Aegis Biosciences (Nashville, TN) for PCR-based analysis. Ct values for positive tests were reported by Aegis to Colgate University Student Health Services. From 8/24/2020 to 9/15/2020, Aegis reported results of two PCR reactions, amplifying regions of the nucleocapsid (N) and ORF1 genes; a positive case required amplification (Ct <40) in one of the two reactions. After 9/15/2020, Aegis added a third PCR target, in the S gene. From this time forward, a positive case required amplification (Ct <40) in at least two of the three targets. The average Ct value for these two or three PCR reactions was calculated for each case and used for further analysis. Most cases identified through surveillance testing were asymptomatic at the time of detection; whether these students subsequently exhibited symptoms was not recorded. In both semesters, students with symptoms consistent with COVID-19 (2 in the fall semester; 18 in the spring semester) were diagnostically tested in-house using a rapid PCR-based assay (Cepheid Biosciences). In some cases, patient tests were performed at other local laboratories. Ct values were not available for these test results. Students that tested positive for SARS-CoV-2, and their close contacts, were identified upon receipt of test results (typically 48-72 h post-collection) and separately isolated or quarantined at a local facility (which was not subject to wastewater surveillance), or, if feasible, at their family home, until allowed to return to their residence hall as established by Madison County Board of Health and New York State guidelines, either at 10 days following symptom onset or at 10 days following the initial positive sample, for asymptomatic cases.

Data analysis
Data on positive SARS-CoV-2 test results, their Ct values (if available), and the start and return dates of isolation, devoid of patientidentifying information were obtained from Colgate University Student Health Services. Data on de-identified total residence hall occupancies were obtained from the Office of Institutional Planning and Research. Due to low case numbers in one residence (GH), data from this residence and waste stream was combined with that of BAE to eliminate potentially identifying information. Additionally, specific dates have not been included in the manuscript to also eliminate potentially identifying information. This project was ruled as exempt from the review process by the Colgate University Institutional Review Board. Data was analyzed using GraphPad Prism statistical software package (Version 9.3.0) as indicated. For temporal analysis, positive wastewater samples were categorized according to their temporal relationship with positive student case(s), in one of the following categories: those that occurred (i) within seven days prior to detection of a positive case; (ii) between the date of a positive patient sample and when the student was moved to the isolation facility; (iii) between the date a student was moved to the isolation facility and when they returned to the residence; (iv) within 5 days of the student's return from isolation; (v) within 6-14 days of the student's return from isolation; and (vi) those that did not correspond to any identified case within those temporal parameters. No positive wastewater result fit more than one of these categories.

Wastewater surveillance rationale and strateg
During the spring and summer of 2020, Colgate University engaged in a strategic planning process to determine how best to facilitate the return of students to campus for the 2020-2021 academic year, given the challenges associated with the SARS-CoV-2/COVID-19 pandemic. At that time, data was emerging regarding the efficacy of detecting the presence of SARS-CoV-2 RNA in wastewater efflux via quantitative reverse transcriptase PCR (qRT-PCR) as a means of population surveillance. SARS-CoV-2 RNA can be detected in the wastewater stream as early as a week prior to identification of patients with COVID-19 based on screening of symptomatic individuals (Peccia et al. 2020, and based on this information, we decided to implement wastewater surveillance as one part of a comprehensive plan to facilitate the return of students to campus residences for the academic year. Colleagues at other local academic institutions were working on methods to quantify SARS-CoV-2 RNA in wastewater (Wilder et al. 2021) and graciously assisted in developing protocols for wastewater surveillance at Colgate University.
The University purchased three compact composite wastewater samplers and we consulted with the Facilities Department to determine locations for sampling. We identified seven locations on campus that would allow as many student residences as possible to be sampled, while minimizing the overlap with effluent waste from academic and administrative buildings (Fig. 1). However, due to the integrated nature of some administrative spaces within residence halls, it was impossible to exclude these entirely at some of the sampling points. In total, the sampling locations captured waste from thirteen individual residence halls or residential complexes, which housed approximately 73% of Colgate students living in campus-owned residences ( Table 1). The largest population sampled was a complex with 370 students, while the smallest was a residence with 73 students, with a mean of 242 residents sampled at each collection point. Due to containing potentially identifying information, data from the smallest residence (GH) was combined with that for BAE for analysis. For wastewater samples, the relative level of SARS-CoV-2 RNA was determined using the Ct value of multiplex qRT-PCR reactions targeting the SARS-CoV-2 genome, normalized to the amount of biological material in the sample, as determined by the Ct value of a qRT-PCR reaction to detect human RNase P mRNA.

Individual surveillance testing
Concurrently, Colgate University engaged in individual surveillance testing of students and employees, as outlined. During the fall semester, from the start of first year arrival in late August, 2020 until most students left campus in mid-November, PCR-based surveillance testing detected 44 cases of SARS-CoV-2 infection in the student population (Table 1; Fig. 2A). Of the fall cases, 36 of the 44 (81%) were detected during the two arrival testing periods, and many of these cases had Ct values of >35, suggesting low viral RNA levels. The mean Ct value of the positive tests in the fall was 34.54 (Fig. 3A). Thirty-one of the cases (70.5%) occurred in residences from which wastewater samples could be collected. In the spring semester, from arrival for most students starting in late January, 2021, through the end of the semester in early May, 63 cases of SARS-CoV-2 infection were detected (Table 1; Fig. 2B).  Again, a large number of cases (28; 44%) were detected during the two arrival testing periods. The mean Ct value of the positive tests in the spring was 26.59, which was significantly lower (higher viral RNA levels) than in the fall ( Fig. 3A; P < 0.0001). The percentage of cases that occurred in residences from which wastewater samples could be collected (48 of 63; 76.2%) was not significantly different than in the fall (P = 0.51 by Chi-square analysis). Cases in residences from which wastewater samples were unable to be collected were excluded from further analysis.

Wastewater results
In the fall, SARS-CoV-2 RNA was detected in 12 wastewater samples, from 6 out of the 7 collection points (Table 1; Fig. 2A).
During the spring semester, 48 wastewater samples tested positive for SARS-CoV-2, from all 7 collection points (Table 1; Fig. 2B). The relative, normalized level of SARS-CoV-2 RNA detected in wastewater samples was not significantly different between the two semesters (Fig. 3B). We analyzed the data for each wastewater sample location (Fig. 4), plotting the relative wastewater SARS-CoV-2 levels (black circles; left Y-axis); as well as the Ct value of each positive patient sample (red circles; right Y-axis). We also plotted positive patient samples for which no Ct value was available (red stars), as well as the dates the positive cases were moved from the residence hall into the isolation facility (yellow squares) and the dates the cases moved back to the residence hall following isolation (green squares). From these data, we categorized each positive wastewater sample according to its temporal relationship with positive student case(s), as described.
Overall, 14 of the positive samples (23.3%) were collected in the seven days prior to the identification of an infected individual in a residence that contributed to the sampled waste stream, or in the time between when patient specimen was collected and the time the individual moved to quarantine (Fig. 5A). Thirty of the positive samples (50%) were collected from locations in the two weeks following the return of an individual from isolation into that waste stream. Only two positive samples were collected in the period between when an infected individual was removed to the isolation facility and when they returned to the residence, which could result from lingering biological material in the waste stream. About 12 positive samples (20%) were not linked temporally with any student cases in that waste stream. There was no significant difference in the normalized wastewater SARS-CoV-2 RNA levels in samples from these categories ( Fig. 5B; P = 0.2 by Kruskal-Wallace test). We also analyzed the wastewater results according to the number of student cases identified in the residences contributing to the waste stream at the time of sample collection. Again, no significant differences were observed ( Fig. 5C; P = 0.17 by Kruskal-Wallace test).
A similar analysis was conducted on the Ct values of the patient specimens based on the temporal proximity of positive wastewater samples (Fig. 5D). In this analysis, 21 of the identified student cases in sampled residences (21.5%) were not linked temporally with any positive wastewater sample. There was a significant difference in Ct values between categories, with significantly higher Ct values (lower viral RNA levels) found in patient samples that were not linked with any positive wastewater samples than in those found in patients in which a positive wastewater sample was detected in the seven days prior to specimen collection (P < 0.001 by Kruskal-Wallace test with Dunn's multiple comparisons).

Positive and negative predictive values of wastewater testing
Examining the ability of a positive wastewater test to predict the presence of a case in a residence, we found that 46 positive wastewater samples were collected from residences in which a student case was identified, in the period from 7 days prior to the positive patient sample(s) being collected to 14 days after the student(s) returned to the residence following isolation. In contrast, 12 positive wastewater samples were collected from residences in which no student case was identified (i.e. false positives), resulting in an overall positive predictive value of 79% (Table 2; Chi square 48.1, Df = 1, P < 0.001). Out of 350 negative wastewater samples, 120 were collected in the period from 7 days prior to a positive case identification to 14 days after the student(s) returned to the residence (i.e. false negatives), resulting in an overall negative predictive value of 66%. From this data, the sensitivity of the wastewater testing was 28%, while the specificity was 95%. The fact that many of the positive samples occurred in the 14 day period following students returning to a residence following isolation was not unexpected, as SARS-CoV-2 RNA can be detected in patient stool for several weeks following infection (Jones et al. 2020). However, this could confound the use of wastewater samples to identify new cases in a residence hall. To assess the predictive capacity for an isolated positive wastewater sample to identify new cases in residence halls with no recent cases, we excluded all samples collected in the 14 days following the return of a student to a residence from the analysis. We identified 14 positive wastewater samples collected in the period 7 days prior to when a positive patient sample was collected, in contrast to 10 negative samples collected in the same period, giving a sensitivity of 58% (Table 3; Chi square 64.61, Df = 1, P < 0.001). The positive predictive value Red stars indicate positive student tests for which no Ct value was available. Yellow and green squares represent the time of isolation start (when individual student cases were removed from the indicated residences) and the time of return to the residence following isolation, respectively.
of these positive samples, based on the 12 false positive samples, was 54%, while the negative predictive value was 96%. Overall, this data suggests that wastewater testing deployed in this manner can be largely predictive of SARS-CoV-2 infections in college residence halls; however, the analysis of a positive result can be complicated in situations where infected students return to a residence following isolation, as SARS-CoV-2 RNA may continue to be detected in the waste stream for at least two weeks after return.

Discussion
Municipal wastewater testing for SARS-CoV-2 RNA can provide a robust indicator of community transmission (Peccia et al. 2020, Wu et al. 2020, and this information can be used to inform public health responses. Many colleges and universities implemented wastewater surveillance testing during the 2020-2021 academic year (Harris-Lovett et al. 2021); however, the efficacy of these efforts in identifying infection clusters and lead-ing to mitigation of transmission has not been fully investigated. In this study, we combine data on wastewater SARS-CoV-2 detection from discrete campus residences with information on student cases, including RNA levels and timing of exit and return to the residence following isolation, to provide insight into the efficacy of wastewater surveillance as a public health strategy at a residential undergraduate college. Overall, we found that while around a quarter of positive wastewater samples indicated the presence of SARS-CoV-2-infected individuals prior to their identification by individual surveillance testing, half of the positive wastewater samples occurred within 14 days after students returned to residence halls following isolation. The capacity for individuals to continue to shed SARS-CoV-2 RNA in stool past the point of likely transmission is a potential confounding factor for the use of wastewater surveillance in a college environment where students leave and return to residences following infection.
Our results, demonstrating the capacity to use wastewater surveillance to identify the presence of SARS-CoV-2 in student populations, are generally concordant with other reports of wastewater surveillance programs on college campuses. However, our determined positive predictive value for isolated wastewater positive samples (excluding dormitories with students returning from isolation), of 54% is lower than some reported values, such as at the University of Arizona (82%; (Betancourt et al. 2021)), or at UC San Diego, where wastewater surveillance allowed for the early diagnosis of 85% of cases on campus (Karthikeyan et al. 2021). However, these programs operated on a very different scale, with, for example, the UCSD system employing 68 samplers with automated sample processing of daily samples. Similar to other institutions, we could identify single cases in a residence hall from wastewater samples (Gibas et al. 2021, Scott et al. 2021, although a large number of negative samples from residences with student cases decreased the overall sensitivity. Groups that sampled on a weekly basis concluded that that was too infrequent for effective identification of cases, and suggested that daily testing would be most appropriate (Sharkey et al. 2021). We found that our strategy, where most residences were sampled 2-3 times a week, was generally effective while not becoming impractical to manage for a small academic laboratory.
We observed significantly higher average Ct values (lower viral RNA levels) among cases in the fall semester than the spring semester (Fig. 3A). This is most likely a reflection of changes in the stringency of the cutoff for the identification of a positive case used by the PCR testing provider that occurred in mid-September October 2020. Initially, a positive case was identified by a defined Ct value of <40 in one of two PCR reactions using different primer pairs. Starting in October, a third PCR reaction was added to the protocol, with a positive case identified with a defined Ct value of <40 in at least two of the three reactions. This increased stringency may have helped exclude potential false positives, especially when the positive predictive value of the PCR-based test was low due to low prevalence in the population. These data could also potentially be explained by students having been infected during the summer and continuing to shed low levels of SARS-CoV-2 RNA in their nasopharynx at the time of the arrival tests in August. If this were the case, however, we would have expected the pre-arrival test that students took ahead of coming to campus to identify these individuals. Alternatively, students may have been infected during travel to campus, and therefore have been tested at a time prior to when they reached peak RNA levels. However, if this were the case, we would have expected that positive cases identified in the second round of arrival testing would have lower Ct values (higher RNA levels), which was not observed. Whatever the reason for this difference, it has implications for our wastewater results, as cases with higher Ct values (lower RNA levels) were significantly less likely to be associated with a positive wastewater result (Fig. 5D).
We observed no significant difference in the normalized SARS-CoV-2 RNA levels in wastewater based on the proximity of the wastewater sample to when a positive student case was identified ( Fig. 5B), or based on the number of cases contributing to the waste stream (Fig. 5C). These results are contrary to other reports, in which wastewater SARS-CoV-2 RNA levels were observed to tightly correlate with case numbers (Peccia et al. 2020. We attribute this discrepancy to the relatively small number of people contributing to each waste stream in our study (hundreds vs. several thousand individuals in some municipal sewer sheds), and to the variable and sporadic nature of the flow rate at each sampling location. Composite samples from the lift stations tended to be more consistent than from the sewer access holes, where the flow rate changed dramatically on a minute-byminute basis, and where the structure of the sewer caused pooling in some lines more than others. This caused marked differences in the amount and density of samples collected, and this variability could potentially lead to differences in the amount of biological material collected between the locations, or between samples from the same location. Additionally, during periods of inclement weather, some sampling locations could only be sampled with a single draw, rather than a 24-hour composite. These factors could therefore impact the relative concentration of SARS-CoV-2 RNA present in each sample, despite the use of RNase P RNA levels to normalize the SARS-CoV-2 RNA levels. Furthermore, the variable flow rate, as well as the lack of equipment to accurately measure this type of flow, made absolute quantification of SARS-CoV-2 genome copies per milliliter, (the standard used by several other studies; (Peccia et al. 2020, Wu et al. 2020), impossible. We conclude that under these conditions, detection of SARS-CoV-2 RNA in the waste stream should be considered on a positive/negative basis, rather than attempting to use RNA quantification as a predictor of case load or timing. Additionally, increasing the cutoff threshold for classifying a wastewater sample as positive could also increase accuracy, by potentially reducing the level of false positives.
One limitation of the study is that, due to the layout of campus infrastructure, several of the sample collection points did not only capture waste from student residences. In some cases, other campus buildings contributed to the waste stream (though every effort was made to minimize this), whereas in other cases, the residence halls also house additional administrative or classroom spaces. Additionally, custodial and facilities maintenance staff routinely accessed the buildings and may have contributed to the waste stream at the sampling locations. In one instance in the fall semester, a positive SARS-CoV-2 RNA signal in a wastewater sample was used to guide targeted PCR-based testing of students in residences, which did not identify any positive student cases. Concurrently, however, an employee that spent time in the space(s) contributing to that waste stream did test positive for SARS-CoV-2 infection, although whether they used restroom facilities in those location(s) is not known. Therefore, it is possible that other positive wastewater samples that did not coincide with a positive case may not be 'false positives,' per se, but rather reflect other non-residents that contributed to the waste stream. It is also possible that these positive wastewater samples reflect infected students that were not identified by individual surveillance testing, either because they were not tested during their infection or their test results were falsely negative. Finally, it is also possible that some of the positive wastewater samples were the result of non-resident students visiting a particular residence hall. While cross-residence hall visitation was prohibited from August until October 2020, this prohibition was difficult to monitor, and students were allowed cross-residence hall visitation for the remainder of the fall semester and much of the spring semester, and so this could also have potentially confounded our wastewater testing results.
There are several additional limitations of this study that impact our ability to fully interpret the data. First, students that reported symptoms consistent with COVID-19 were tested using an in-house rapid nucleic acid amplification system that does not provide Ct values. Differences in viral dynamics at the time of testing (in surveillance tests vs. symptomatic testing) could skew our analysis of the Ct values of cases associated with positive wastewater detection of SARS-CoV-2. Second, the nature of our sampling strategy, with the collectors rotating between sampling locations on a two-to-three-day schedule, means that our samples were not evenly spaced, especially around times of arrival testing, when a majority of our positive cases were detected, and around times when students returned to residence halls following isolation. This confounds our analysis of wastewater RNA levels relative to the time of case detection or return, as it meant that some residences may have been sampled the morning after a student returned (potentially leading to a higher level of viral RNA shed into the waste stream), whereas others might not have been sampled until up to 72 h post-return, when shed RNA levels may have been lower. A sampling protocol that allowed for more continuous monitoring of wastewater RNA levels from residences following a return from isolation may have allowed for a more precise correlation between RNA levels and time from isolation return to be detected. This, however, would require more resources devoted to the project and would be beyond the scope of what our laboratory could handle. Additionally, given the limited capacity of our rotor for ultracentrifugation of samples, the sample volume used for RNA isolation was relatively low (26 ml); changes to the protocol to accommodate larger sample sizes might also increase the ability to detect SARS-CoV-2. Finally, due to infrastructure architecture and geography, only around 73% of students lived in residences that could be easily monitored via the waste stream. This decreased the overall utility of wastewater surveillance as a means for managing emerging cases on campus.
Overall, based on the data presented here and our lived experience of managing the COVID-19 pandemic during the 2020-2021 academic year at Colgate University, we found that on a practical level, wastewater surveillance had more utility in the fall semester, when a smaller percentage of the community was subject to random PCR-based individual surveillance testing. The wastewater results could then be used to direct additional targeted testing to individual residences with positive wastewater RNA samples, and, with a positive predictive value of 54% for an isolated wastewater positive sample, this could be potentially be a more cost-and resource-effective means of detecting new positive cases than large-scale individual student testing. In the spring semester, with approximately half of the students in each residence were tested each week, by the time a positive wastewater sample was detected, it was, in our experience, more practical to wait for the individual surveillance results to become available than to target students in those residences for additional testing. This was further confounded when positive student cases returned to the residences following isolation, making it difficult to distinguish between a new infection and continued SARS-CoV-2 RNA shedding in previously-infected individuals. Additionally, the presence of cases in residences in which no wastewater SARS-CoV-2 RNA was detected suggests that wastewater surveillance of this type should not be used as the only means for the detection of cases on campus. As the price of personal surveillance testing, via pooled or individual samples, continues to decrease, we conclude that this should be viewed as a more effective means of controlling SARS-CoV-2 transmission on campus than wastewater testing of individual residence halls.

Authorship contribution statement
GHH initiated the project, developed the wastewater sampling strategy and SARS-CoV-2 RNA detection protocol, contributed to sample collection, analyzed and interpreted the data, and drafted the manuscript; ML contributed to sample collection, performed the RNA extraction and qPCR, analyzed the qPCR data, and edited the manuscript; EB contributed to sample collection; EL and MM collected and assembled deidentified patient information, and edited the manuscript.

Data availability statement
All data related to this project is available at figshare.com, licensed CC by 4.0