Participants

The protocol (Supplementary Section 5), including all amendments, was approved by the Ethics Committee of the Medical Faculty of Kiel University (file reference A107/20), and all participants provided informed consent before any study procedures (see below). COVit-1 and COVit-2 were registered at the World Health Organization primary registry German Clinical Trials Register (DRKS00021214); COVit-2 was additionally registered at ClinicalTrials.gov (NCT04751604). In the COVit-1 trial, 56 outpatients with early symptomatic COVID-19 in domestic quarantine were recruited between 6 April 2020 and 28 January 2021, and 900 outpatients were enrolled into the COVit-2 trial between 1 February 2021 and 17 January 2022. Participants were compensated for their time according to the ethics committee’s approval (up to 265.00 € for interviews and questionnaires, up to 50.00 € for stool samples and up to 120.00 € for blood samples). Vaccinated individuals were excluded to avoid confounding (for example, by inhomogeneous vaccination schedules or selection by age groups that received preferential access to vaccines). Self-reported demographics, including sex assigned at birth (male or female) or the gender option ‘diverse’ (not selected by any participant), were collected at screening. Owing to the lack of evidence for a sex-specific effect of nicotinamide in COVID-19 or similar infections, neither sex nor gender were specifically considered in the design of the trial, but were analysed in an exploratory fashion. No analysis of viral subtypes was performed, but with respect to population epidemiology in Germany, participants were almost exclusively infected by the wild-type virus in the COVit-1 trial and by the Alpha (B.1.1.7) or Delta (B.1.617.2) virotypes in the COVit-2 trial⁴³.

For COVit-1, participants were recruited through outpatient facilities surrounding the University Hospital Schleswig-Holstein, whereas screening for COVit-2 was performed using diagnostic laboratories at 71 sites in Germany (Supplementary Section 3.1). Inclusion criteria for the overall population were ≥18 years of age, a recent SARS-CoV-2 infection (≤7 days after first positive test) and at least one symptom of COVID-19 on the day of randomization (Supplementary Section 3.2). Most participants reported five or more symptoms at baseline. Symptom load and the number of risk factors were similar in both trial arms and over the trial period (Supplementary Section 3.3). Exclusion criteria were current participation in another interventional study, pregnancy, breast-feeding and current or anticipated hospitalization. Inclusion into the acute RFITT primary-analysis population additionally required at least one risk factor for severe COVID-19 (Supplementary Section 3.4). At the 6-month follow-up, the ITT group was the primary analysis population, in which subgroups of participants at risk for developing PCS and responders to the intervention were further analysed (Supplementary Section 3.4).

Randomization, masking and trial procedures

The trials were performed remotely owing to the contact restrictions, which were in place during the trial time period. For details on trial procedures and design, see Supplementary Section 3.1. Participants provided electronic written informed consent, which was confirmed by telephone. Whereas COVit-1 recruited participants through advertisements, COVit-2 identified candidates by contacting all patients who received a positive SARS-CoV-2 test result at 71 sites operated by German diagnostic laboratory service providers. After confirming eligibility, participants were computer-randomized and received the interventional product and paper questionnaires (SF-36 and FACIT-F), delivered by next-day courier. Eligible participants were randomly assigned in a 1:1 ratio to daily self-administration of either 1 g day^–1 nicotinamide (500 mg immediate-release nicotinamide and 500 mg controlled-ileocolonic-release nicotinamide (CICR-NAM, Setamer®)) or matched placebo tablets in identical primary and secondary packaging, taken with breakfast for 4 weeks. Tablets were formally released by the pharmacy of the trial sponsor, the University Hospital Schleswig-Holstein (Kiel, Germany). The intervention received by a participant was not disclosed to personnel involved in the study; personnel responsible for clinical supply and safety were unblinded. Participants and personnel involved with participant care remained blinded throughout the study, including the 6-month follow-up. Participants underwent structured telephone interviews at weeks 2, 4 and 6, as well as 6 months after baseline. Optional stool samples were collected from participants by mail (Supplementary Section 3.5).

Clinical outcomes

The original primary clinical outcome of COVit-1 was the rate of hospital admission for a minimum of 24 h of continuous oxygen therapy. Secondary endpoints included the rates of machine ventilation, intensive care and death, as well as time to resolution of symptoms (Supplementary Section 1). Owing to the results of the pilot experiment, COVit-2 focused on participant-reported COVID-19 symptom burden in the acute primary analysis RFITT population (ITT participants with at least one risk factor for severe COVID-19; Supplementary Section 3.4). The primary endpoint was restoration of physical performance at week 2. Key secondary endpoints were an improvement of the ability to perform normal activities, resolution of cough and resolution of fatigue at week 2. All endpoints were tested in participants with the respective symptoms at baseline. Prespecified subgroup analyses were performed for key risk factors. At the 6-month follow-up, the main outcome was PCS determined by a previously established PCS score, which was derived and validated in a large and prospective German cohort²³. For the complete list of outcomes and for details on trial populations, see Supplementary Sections 3.2 and 3.4.

Gut microbiome analyses

In COVit-2, 16S rDNA phylogenomic and metagenomic analyses were performed to investigate the effects of nicotinamide on gut microbial community composition and on functional metabolic capabilities of the colonic microbiome stratified by pathways, gene families and enzyme categories. Stool samples from 88 participants (70 participants for 16S and 18 participants for metagenomics) were analysed at week 0 (baseline), weeks 2 and 4 (exposure to nicotinamide or placebo) and week 6 (follow-up). For a detailed description of methods and references, see Supplementary Section 3.5.

Safety

The safety population included all randomized participants. AEs were classified into preferred terms and summarized using MedDRA version 25.1. These AEs were reported in participant interviews conducted after baseline and through ad hoc reports from participants requesting medical consultation until week 6. Serious AEs were recorded by structured interview queries, with follow-up if necessary. All hospitalizations and emergency-room visits were recorded. Serious AEs related to the underlying disease were evaluated according to the World Health Organization’s COVID-19 scale²⁸. For the safety analysis of the intervention, symptoms reported during the course of the study that were not present at baseline or were increased in severity were listed, and their frequency was compared between the nicotinamide and placebo groups using unadjusted Chi-square or Fisher’s exact tests.

Primary statistical analysis

COVit-1 included 56 outpatients recruited from the referral network of the University Hospital Schleswig-Holstein and served to establish the rationale for the larger study, COVit-2. Statistics were descriptive because the study was a pilot trial (Supplementary Section 1). The results of the COVit-1 pilot trial were kept separate from those of the main trial, COVit-2. Statistical analysis was conducted in the blinded dataset by a third-party provider with established clinical trial statistics expertise (Novustat).

The COVit-2 trial enrolled 900 participants, following the sample size assumptions detailed in Supplementary Section 3.6. A pre-planned futility analysis was conducted by the Data Management Board after 400 participants had been recruited (Supplementary Sections 3.6 and 6). The full analysis set (the ITT population) included all participants who received at least one dose of nicotinamide or placebo. The RFITT population was defined as all participants in the ITT population with at least one symptom, demographic characteristic or underlying medical condition that was previously associated with an increased risk of developing severe COVID-19. The respective per-protocol populations, PP and RFPP, excluded those who dropped out or failed to comply with the investigational product intake for at least 80% of the study duration (Supplementary Section 3.4). In the RFITT population, each analysis regarding resolution or improvement of a symptom was performed only for those participants who reported the respective symptom at baseline. In case of ordinal queries (Supplementary Section 3.2), participants with severe symptoms (a value of >3) at baseline were selected for analysis. For the FACIT-F and SF-36 questionnaires, only those with severe complaints (baseline values ≤ median) were included in the analyses.

Baseline characteristics were summarized according to trial group and overall, with the use of descriptive statistics for continuous and categorical measures.

Primary and confirmatory secondary binary endpoints were analysed by assessing changes from baseline for all weeks within each intervention group, compared using the Cochran–Mantel–Haenszel test. Post hoc analyses for each week were calculated using Fisher’s exact test, with Benjamini–Hochberg adjustment for multiple testing. The Woolfe test was performed to test for homogeneity of odds ratios across time. If significant P values were obtained from the Woolfe test, the Cochran–Mantel–Haenszel test would not be appropriate. In this case, Fisher’s exact tests for each timepoint were used instead of the Cochran–Mantel–Haenszel test. A continuity correction was applied for zero frequencies.

Continuous secondary outcomes (scales, SF-36 and FACIT-F) were analysed with the use of a mixed model for repeated measures (MMRM), with the change from baseline at each of the three scheduled post-baseline time points (2, 4 and 6 weeks) as the dependent variable and baseline value, intervention group, time and time–intervention interaction as independent variables (Supplementary Section 6). Statistical methods that do not involve imputation, such as the MMRM or the Chi-square test, were used for the analyses. The use of the MMRM model assumes implicitly that data are missing at random.

For time-to-event analyses, a Kaplan–Meier approach was used. The log rank test was performed to test whether time to event differed between intervention groups. A Cox proportional-hazards model was used to evaluate and estimate the impact of the intervention group. The assumption of proportional hazards was analysed before the model was applied.

The reliability of the FACIT-F and SF-36 questionnaires was assessed using Cronbach’s alpha coefficient (α ≥ 0.80) for internal consistency and item-to-total correlations exceeding 0.20. The average variance extracted was calculated to assess discriminant validity.

We applied multivariable generalized linear models, using a binomial family with a log link for binary endpoints and a Gaussian family for endpoints measured on a scale of complaints, to assess the impact of sex. Treatment, change from baseline, baseline value, sex and treatment–time and treatment–sex interactions were considered as independent factors. Sex-specific subgroup analyses were performed as part of the exploratory analyses. Odds ratios and 95% confidence intervals were calculated for binary endpoints, and Hedges’ g effect sizes including 95% confidence intervals were calculated for ordinal scales of complaint. Additional analyses were performed in accordance with the analyses of the entire RFITT population.

We assessed normality using the Shapiro–Wilk test as well as histograms, and homogeneity of variances using Levene’s test. The results of these tests confirmed that the assumptions of the statistical tests were met. No data points were excluded from the analyses, as we also used MMRM and generalized linear models, which account for all available data without explicit exclusions. Detailed descriptions of the study populations for specific analyses are available in the statistical analysis plans (Supplementary Sections 6 and 7).

All statistical analyses were performed with the use of R software (R Foundation for Statistical Computing, 2021) version 4.1.2 or higher. For further details on the statistical analyses, see Supplementary Sections 3.5, 3.6, 6 and 7.

Exploratory statistical analysis of occurrence of PCS

For the 6-month follow-up, the PCS score²³ served as the primary efficacy variable and was compared between the intervention groups using a t-test or a nonparametric Mann–Whitney U test. Subgroup analyses were performed to further define responders in defined risk groups. For further details on the statistical analyses, see Supplementary Section 7.

Quality-control measures

Blinded (recruiters, interviewers, study physicians, statisticians, technicians and scientists for microbiome analysis) and unblinded (study material distribution, safety) personnel were strictly separated. All personnel completed documented formal monitored training on trial procedures, and delegation logs were adapted from good-clinical-practice guidelines. Guided standard operating procedures were regularly retrained, and detailed instructions for participants were implemented to ensure the validity of assessing participant-reported outcomes through structured telephone interviews (Supplementary Sections 3.1 and 3.2). Key interview questions were redundant, and source data entry into the database was monitored. SARS-CoV-2 test results were verified. Compliance was surveyed by remote tablet count during each interview and through specific questions (Supplementary Section 3.1). Side effects were queried and coded according to MedDRA Version 25.1 (see above).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.