This decentralized, prospective observational study was completed via collaboration between the Scripps Research Translational Institute (SRTI) and eMed, which implemented the National Institutes of Health Home Test to Treat Program, a 24/7 virtual health platform that provides proctor-guided at-home testing with public health reporting through a digital Clinical Laboratory Improvement Amendments (CLIA)–waived laboratory. Eligibility criteria included a positive rapid antigen test for COVID-19 within 48 hours prior to study enrollment, residing within the United States, speaking English, and aged at least 18 years. eMed users who tested positive were offered NPR at the discretion of an eMed provider. NPR was the only COVID-19 oral antiviral therapy offered through the eMed platform. There were two study recruitment cohorts: (1) from August 2022 through December 2023, eligible eMed users that tested positive for COVID-19 through an eMed telehealth appointment were invited via follow-up email, text, and/or phone call; and (2) from August 2023 through December 2023, social media campaigns and SRTI blog posts invited people who self-tested positive to consider participation. Participation required a virtual meeting with an eMed provider to confirm a positive COVID-19 rapid antigen test and for further eligibility screening.

Interested individuals were sent to a landing page with more information on the study, an eligibility screener survey, and e-consent forms. Upon providing consent, participants reported baseline information (self-reported demographic characteristics, preexisting conditions, and vaccination history) and completed a symptom survey prior to requesting a study kit. Study kits were shipped overnight and included detailed instructions for the 15-day acute study phase and eMed live-proctor–enabled COVID-19 test-to-treat rapid antigen test kits (containing Abbott BinaxNOW Home tests). Participants who enrolled after August 2023 (cohort 2) were additionally sent 2 at-home blood collection devices with biohazard shipping materials. Cohort 2 was instructed to return blood samples and testing swabs to SRTI for serological analyses and viral genome sequencing. Results from blood sampling and sequencing are forthcoming in a future report.

Participants were asked to complete a survey (capturing symptoms and NPR use) and a proctored eMed COVID-19 rapid antigen test on days 2, 5, 7, 9, 11, 13, and 15 from enrollment (acute study phase). To complete a rapid antigen COVID-19 test, participants were instructed to scan a QR code on their test box to initiate a telehealth proctored visit where they were guided through the testing procedure and test results validated. If a participant tested positive during the visit and had not already received care either in person or through a telemedicine visit, they were offered an optional eMed telemedicine visit that could provide a prescription for a 5-day course of NPR (provided under the brand name Paxlovid). Postacute COVID-19 symptom surveys were sent to all participants at 1-, 3-, and 6-months after enrollment, with results forthcoming. Participants received up to 3 notification reminders per task using their preferred mode of communication (email, text, and phone) unless they chose to withdraw early. Research Electronic Data Capture (REDCap) tools were used for survey and notification administration [16,17]. Participants were offered an incentive of electronic gift cards totaling up to US $200, which were issued over the study period according to completion of required tasks.

The primary outcomes were cumulative incidence of viral and symptom rebound in NPR and control groups within 15 days of enrollment. Viral rebound was defined as a positive rapid antigen test after previously testing negative. Symptom rebound was defined as reporting at least one symptom on a survey after reporting zero symptoms on a previous survey. Secondary outcomes included time from enrollment to initial negative rapid antigen test and to symptom resolution (viral and symptom clearance, respectively), and rebound probability among patients with clearance by day 15. Exploratory outcomes included comparing characteristics of rebound and nonrebound groups and symptom frequency during the acute study phase. Reported symptoms were aggregated into binary categories of systemic, respiratory, gastrointestinal, neurological, and other symptoms.

The study was approved by the Scripps Institutional Review Board (22‐7978). All participants provided electronic informed consent. SRTI performed safety oversight.

Of 917 consented participants, 669 (73%) were eligible for inclusion in analyses (Figure 1); of these, 443 (66.2%) reported taking at least one dose of NPR (NPR group) and 226 (33.8%) did not (control group). Within the NPR group (n=443), 197 (44.5%) reported taking all 5 days of their NPR prescription. The average age was 46.1 (SD 12.9) years, and 62.6% (419/669) of participants were female. The most common preexisting condition was high blood pressure (132/669, 19.7%), and half of participants (340/669, 50.8%) did not report any preexisting conditions. While sex was similar between groups, there were differences in age (56/67, 83.6% of those older than 65 years opting to take NPR compared with 189/258, 73.3% of those aged 45-65 years and 198/344, 57.6% of those aged 18-44 years), race (351/491, 71.5% of White participants opting to take NPR compared with 85/168, 50.3% of non-White participants), and the presence of preexisting conditions (Table 1). The proportion of participants in the NPR group was higher in the first cohort (181/251, 72.1%) than the second cohort (262/418, 62.7%; Table S1 in Multimedia Appendix 1).

The 15-day cumulative viral rebound incidence was 12.3% (82/669) overall (NPR: 70/443, 15.8% and control: 12/226, 5.3%). The 15-day cumulative symptom rebound incidence was 13.8% (92/664) overall (NPR: 73/443, 16.5% and control: 19/226, 8.4%). Kaplan-Meier curves showed significantly increased probability of viral test rebound by day 15 of 16.2% (95% CI 13.1%‐20.1%) for the NPR group and 5.6% (95% CI 3.2%‐9.6%) for the control group (Figure 2; log-rank P<.001). Probability of symptom rebound by day 15 was 17.2% (95% CI 13.9%‐21.1%) and 8.8% (95% CI 5.7%‐13.4%) for NPR and control groups, respectively (Figure 2; log-rank P=.004). Secondary analyses examining incidence across different eligibility and timing criteria (Table S2 in Multimedia Appendix 1), whether a partial (1‐4 days) or full (5 days) prescription of NPR was taken (Table S3 in Multimedia Appendix 1), and adjusted for baseline demographic and clinical characteristics using Cox proportional hazards regression (Tables S4 and S5 in Multimedia Appendix 1) did not change conclusions.

This COVID-19 Rebound Study found that participants who independently decided to take at least one dose of NPR had triple the 15-day cumulative incidence of viral test rebound and double the cumulative incidence of symptom rebound than the control group. These findings persisted in our secondary analysis after adjusting for baseline characteristics that may confound the association between NPR and COVID-19 rebound. Furthermore, when restricting the analysis to participants who achieved initial clearance before day 15, NPR use was significantly associated with a higher probability of both viral rebound (19.1% vs 7.2%; P<.001) and symptom rebound (47.7% vs 16.9%; P<.001) at study completion.

The estimates in our study are comparable to other studies, which have reported cumulative incidence of viral rebound in NPR treatment populations from 4% to 27%, and symptom rebound from 0.8% to 32% [21-23]. Studies with lower estimates tended to be secondary analyses of randomized controlled trials or retrospective analyses of electronic medical record data, which may only capture more severe cases [12,22-26]. Studies with higher estimates tended to use more sensitive measures of viral load, and in some cases, they did not require full resolution of symptoms for participants to be eligible for symptom rebound [21,27]. A strength of our study is that it was designed with rebound as the primary outcome and used frequent telehealth, proctor-verified home testing, and symptom surveys. Furthermore, in contrast to most other studies, we did not exclude participants who took partial doses of NPR, which was more common than taking the full 5-day prescription (n=226, 33.7% none, n=246, 36.7% partial, and n=197, 29.4% full), allowing for more accurate real-world estimates of rebound.

We found minimal overall differences in time to initial viral or symptom clearance between the NPR and control groups, suggesting that further investigation was needed into the discordance between how NPR reduces the potency of the virus to ultimately reduce the progression to severe COVID-19, as found in previous studies, and the identified increased rates of rebound or persistent virus. Time to initial symptom clearance was longer than viral clearance, consistent with another prospective study [21]. The tendency for symptoms to remain after testing negative was further seen in examinations of consistency between positive tests and symptoms across the 15-day acute phase follow-up (Table S6 in Multimedia Appendix 1). Of note, we captured self-reported presence or absence of symptoms, whereas symptom severity might yield additional insights.

An additional finding from our study is that at 5 days after enrollment (corresponding to up to 7 days after first turning positive), more than one-third (244/669, 36.5%) of participants remained positive on rapid antigen tests (Figure 3). Although Centers for Disease Control and Prevention guidance has largely recommended ending isolation after 5 days, our findings suggest that individuals may remain infectious for longer and should therefore remain cautious.

Numerous mechanisms have been hypothesized for why rebound occurs. The constant push and pull between virus and immunity drives “predator-prey dynamics,” where the virus’ presence triggers the activation of immune defenses, and virus clearance signals these immune defenses to shift toward a recovery mode [28]. If the shift is too early, the relaxed defenses can allow remaining virus to rapidly re-expand. Regarding NPR and increased rebound risk, it is possible that early treatment initiation pharmaceutically drives down virus growth while abrogating the need to raise a robust immunological defense. Thus, upon removal of NPR, remaining virus could replicate at a rate that outpaces immune defenses, leading to a virus rebound until defenses catch up. Consistent with many COVID-19 symptoms in an era of widespread immunity that represent activating immune defenses (ie, fever and congestion), rebound symptoms generally reflect mounting immune defenses. If immunity must increase to clear residual virus after completion of NPR treatment, many individuals would be expected to experience symptoms; this would lead to a more common symptom rebound phenomenon and a less common full virus rebound, which aligns with our results [29-31].

The COVID-19 Rebound Study was not designed to investigate the efficacy of NPR. Prior studies clearly demonstrate the impact of NPR on reducing COVID-19–related hospitalizations and severe outcomes [3-7,15]. On the basis of our rebound-related findings, our recommendation to clinical health care providers prescribing NPR is to inform patients of the higher likelihood of viral and symptom rebound, that there is insufficient evidence regarding the severity of these symptoms, and that the evidence supporting reduction in hospitalizations and progression to severe COVID-19 is strong.

This was a nonrandomized study that may be limited by selection bias, recall bias, and an unbalanced sample size. While NPR was the primary COVID-19 oral antiviral therapeutic, we did not systematically capture use of emerging alternative antiviral therapies in the control group, which may influence rebound estimates. Our definition of symptom rebound (≥1 symptom after resolution) is sensitive but may have lacked specificity, potentially overestimating clinically meaningful rebound events. Another limitation is that we required enrollment within 48-hours of a positive rapid antigen test; however, participants may have had symptoms and disease for longer durations prior to self-testing and/or asymptomatic participants may have tested at an unknown time point after infection, thereby leading to underestimates on time to viral or symptom clearance. Timing estimates were also impacted by the length of the study as not all participants experienced clearance by day 15. The relatively large numbers of participants in both groups helps to mitigate, though not necessarily eliminate, the impact of these potential biases as they pertain to relative differences between groups. The social media recruitment arm was opened partway through the study primarily to increase the number of participants who were not intending to take NPR, as those recruited through the eMed platform may have been drawn to the test-to-treat virtual care paradigm. The analysis excluded 27% (248/917) of consented participants due to missing data that precluded capture of potential rebound. Excluded participants tended to be less vaccinated, NPR users, and non-White (Table S7 in Multimedia Appendix 1). There was also nonresponse or dropout during the study, whereby 67.6% (452/669) of the analysis sample completed all acute study phase surveys.

In summary, this prospective observational study of COVID-19–positive outpatients showed that NPR treatment decision was associated with a substantially increased risk of both viral and symptom COVID-19 rebound. No difference was found in the overall time to initial clearance of viral test positivity or symptoms between the NPR and control groups, although the NPR group was older and had more preexisting conditions at baseline. With this important caveat for interpretation of our findings, there is a clear-cut need for more durable antiviral therapies or regimens and for improved knowledge regarding optimal patient selection.