Ethical approval
The study protocol was approved by the Ethics Committee Research UZ / KU Leuven (reference no. S64919). Informed consent was waived as the data gathered did not exceed what was required for the purpose of safeguarding public health.
Setting
The study was conducted in the setting of a dedicated testing and contact tracing program targeting around 50,000 higher education students at KU Leuven Association in Leuven, Belgium. A test and trace program was implemented, specifically targeting this student population. Students could attend for a PCR test at no cost and even in the absence of specific test criteria. The test centre closely cooperated with the contact tracing team to improve the flow of information.
New Year’s Eve 2022 (NYE) was a national celebration which encouraged many students to briefly meet with friends and family, before returning to a generally solitary period of studying for their January tests. On a national level, the Omicron BA.1 strain was dominant and rising at the time21. Government-mandated restrictions resulted in limited event sizes (Supplementary Figure 2)22.
Inclusion and exclusion
In the days after NYE, we observed a surge in case numbers and initiated further investigation. We attempted to interview each case with a positive result on the TaqPath COVID-19 assay at the university test centre between 25 December 2021 and 14 January 2022. Cases with a previous positive test result in the 60 days preceding their sample collection were excluded, as well as samples with only a weak positive signal (Cq value for SARS-CoV-2 N-gene target above 30). To construct the epidemic curve, we plotted symptomatic cases according to self-reported disease onset date. Based on national genomic monitoring, S-gene target failure of the TaqPath COVID-19 assay was used as a marker of the Omicron BA.1 lineage21. S-gene positive samples were considered as Delta before and Omicron BA.2 from 17 January 2022 onwards, corresponding to the first week when the share of new Omicron BA.2 cases overtook the share of Delta cases on a national level (Supplementary Figure 1)21.
Effective reproduction number
For the local student population, Reff was estimated from the epidemic curve as described by Cori et al. with a one day sliding window17. We used a parametric gamma distribution of the serial interval with a mean of 2.75 days (SD: 2.54 days)23.
We inferred Reff on a national scale from publicly reported case numbers19. As their symptom onset dates were not available to us, we inferred Reff from sampling dates using a 7-day sliding window in the R package EpiInvert. This method corrects for reporting biases such as the “weekend effect” but smooths the estimated value of Reff20. We did not use the option to ignore the effect of public holidays on Reff, because they are the subject of this study. The serial interval was assumed normally distributed with a mean of 2.75 days (SD: 2.54 days). We did not correct for delays between symptom onset and sampling.
Definition of superspreading event
A SSE was defined as an event with more transmissions than would originate from 99% of infectious cases, when assuming a simple Poisson distribution of infections24. Using our estimated Reff of 3.6 for Leuven students on NYE, events with at least nine transmissions were considered superspreading events.
Contact tracing
At least one attempt was made to contact each student with a positive PCR test for a telephone interview, in an effort to trace the source of their infection and their close contacts during the contagious period. Close contact was defined as direct contact, or an interaction with at least 2 of the following characteristics: duration longer than 15 minutes, distance shorter than 1.5 metres and absence of face masks.
A NYE event was defined as any social gathering occurring between noon on 31 December 2021 and noon on 1 January 2022. The contact tracers determined, based on experience, which cases were possibly and likely infected on NYE. Cases were classed as “possibly infected on NYE” if they had attended an NYE event (with or without other known cases in attendance), developed symptoms 0 to 14 days afterward, and – in the period from 1 week before NYE to symptom onset – had no close contact with a potential source case who was not present at the NYE event. Cases fulfilling these criteria were additionally considered “likely infected on NYE” if they were in contact with another known case during the NYE event, but not on any of the preceding or subsequent seven days.
Phylogenetic analysis
Our phylogenetic analysis included, first of all, the genomic sequences obtained from students possibly or likely infected on NYE.
In addition, we included all Belgian Omicron sequences available on GISAID, sampled between 1 November 2021 and 20 January 2022, and analysed for the purpose of baseline genomic surveillance. This was determined by searching for the word “baseline” in the “sampling strategy” metadata field. If this field was blank, the sequence was excluded.
Finally, a selection was included of other European Omicron sequences from the same period, labelled as “high coverage” by GISAID. “High coverage” here indicates less than 1% of undefined bases, less than 0.05% unique amino acid mutations, and no insertions or deletions unless verified by the submitter. The selection was obtained through proximity sampling – genetic similarity to the included Belgian sequences – and limited to 20 genomic sequences per country and per month.
Sequences with fewer than 27,000 known bases or not belonging to the BA.1 lineage were excluded.
The Nextstrain pipeline with its default parameters was used to construct a time-scaled phylogenetic tree25. Multiple sequence alignment was accomplished using MAFFT, with the WIV04 genome (GISAID accession number EPI_ISL_402124) as the reference sequence26. A phylogenetic tree was constructed using Augur, calling IQ-TREE with a general time reversible model, and time-scaled using TreeTime27–29.
As our analysis was focused on the Omicron BA.1 lineage, only the corresponding Nextstrain clade (21K) was visualised using the R package ggtree30.