Descriptive and univariate analysis
The characteristics of our study population are summarized in Fig. 1 and Supplementary Tables 1 and 2. It consisted of 1,542,510 adults aged 18 years or older, including a total of 17,448 swab-positive individuals: 2,971 (0, 4%, 95% confidence interval (CI). [0.4,0.4] unweighted prevalence) for wild type; 2275 (0.6% [0.6,0.7]) for Alpha; 1493 (0.7% [0.6,0.7]) for Delta and 10,709 (4.4% [4.3,4.5]) for Omicron variants (Supplementary Table 1).
The proportion of swab-positive individuals reporting any of the 26 symptoms (symptoms listed in Supplementary Table 1) was higher in those infected with BA.2 (75.9% [74.4,77.2]compared to 70.0% [68.3,71.6] in those with BA.1, 63.8% [61.3,66.2] in those with Delta, 54.7%, [52.7,56.8] in those with Alpha and 45.0% [43.3,46.8] in those with wild type) (Table S2). Background prevalence of symptoms was also highest during January-March 2022, when Omicron dominated: 21.9%, [21.7,22.0] of all respondents reported one or more symptoms, compared to 13.5% [13.4,13.5] during the wild-type period (Supplementary Table 1).
Those infected with BA.2 reported an average of 6.0 (95% CI: 5.8, 6.2) symptoms in the week prior to PCR testing, compared with 2.70 (2.6, 2 .8), 3.4 (3.2, 3.6), 4.6 (4.4, 4.9) and 4.6 (4.5,4.8) for wild type, Alpha, Delta and BA.1 respectively (Table complementary 2). A higher proportion of people with BA.2 reported that their symptoms had affected their ability to perform daily activities “a lot” (17.6% [16.3,18.8]) compared to those infected with BA.1 (10.7% [9.6,11.9]) or Delta (10.5%, [9.1,12.2]) (Supplementary Table 2).
All symptoms were positively associated with swab positivity for all variants (Fig. 2, Table S3). The odds ratio for ‘any’ vs. ‘none’ swab positivity of 26 symptoms was highest for BA.2 (OR = 12.9 [11.9,14.0]compared to 5.7 [4.8,5.6]6.0 [5.1,7.1]9.5 [8.6,10.6] and 9.6 [8.8,10.5] for wild-type, Alpha, Delta, and BA.1, respectively) (Supplementary Table 3, Fig. 2).
Unlike wild-type, Alpha and Delta, where the highest odds ratios for swab positivity were for loss or change in sense of smell (OR 49.7). [44.3,55.7]37.8 [28.6,50.0] and 73.4 [64.2,83.9]respectively) or taste (ORs 35.9 [31.9,40.4]38.9 [29.9,50.6] and 68.1 [59.4,78.0] respectively), for BA.1 and BA.2, cold and flu-like symptoms were relatively more predictive of swab positivity, and loss or change in sense of smell or taste relatively less. Within BA.1 and BA.2, the highest odds ratio for all symptoms was for fever: ORs were 18.4 [16.5,20.5] for BA.1 and 30.2 [27.7,33.0] for BA.2, compared to 12.9 [11.1,15.1] and 17.2 [15.1,19.5] respectively due to loss or change of the sense of smell and 16.0 [13.9,18.5] and 21.3 [18.9,24.0] respectively by loss or change of sense of taste (Fig. 2, Supplementary Table 3). In a sensitivity analysis, additional adjustment for time since symptom onset attenuated odds ratios, but patterns between variants remained consistent with the main analysis (Supplementary Fig. 7).
A pooled analysis (Methods) reinforced the findings of the univariate analysis after adjusting for SARS-CoV-2 prevalence and prevalence of background symptoms, and showed that Alpha and Delta were associated with higher symptom-specific odds ratios in most symptoms, whereas Omicron BA.1 was associated with lower odds ratios in most symptoms, and especially for loss of sense of smell or taste (Supplementary Fig. 2). Omicron BA.2 was associated with higher odds ratios versus BA.1, most notably for cold-like symptoms and chills.
Multivariate analysis for variable selection
We used least absolute selection and reduction operator (LASSO) penalized logistic regression to identify selected parsimonious symptom sets as joint and positive predictors of swab positivity for each variant (Fig. 3, Fig. S3); this method takes into account the differences in the concurrence of symptoms by variant (Figs. S4 and S5). Loss or change in sense of taste, new-onset persistent cough, and fever were selected for each variant. In particular, the cold-like symptoms of runny nose, sore throat, sneezing, and hoarse voice were only selected for Omicron (BA.1 and BA.2).
Omicron (BA.1 and BA.2)
When comparing symptoms of BA.2 with those of BA.1 using logistic regression (based on model fitting or comparison, see ‘Methods’), BA.2 infection was positively associated with chest pain, severe fatigue , runny nose, muscle aches, sneezing, fever, chills, tiredness, nasal congestion, and headache (in both sets of tests); in peer-reviewed analysis, BA.2 infection was further associated with eye pain, loss of appetite, and new persistent cough (Fig. 4).
In a subgroup of 5,598 people with BA.1 or BA.2 who were double and triple swab-positive vaccinated, people infected with BA.2 were 54% more likely to report symptoms that interfered with their ability to carry out their activities everyday. daily activities ‘a lot’ (OR 1.54 [1.16, 2.06]) vs ‘a little’, ‘not at all’ or not reporting any symptoms, after adjustment for age group, sex, vaccination count, time since most recent vaccination, prior SARS-CoV-2 infection, time since onset of symptoms and time schedule (Table 1). In the same models, men were 38% less likely than women to report symptoms that interfered with their ability to carry out daily activities ‘a lot’ (0.62 [0.52,0.73]). Vaccine booster status and time since vaccination were not associated with ability to perform daily activities. In the same subgroup, a log-linear regression of symptom counts found that those infected with BA.2 reported 14% more symptoms, on average, than those with BA.1 (OR = 1.14 [1.10,1.19]) after adjustment for the same covariates as before (Supplementary Table 5).
Ct values were lower for BA.2 than BA.1 (Supplementary Fig. 1). This may reflect the timing of sampling with respect to variant growth, as more recent infections will tend to have lower Ct values (see Supplementary Table 2 and Supplementary Figs 6 and 7, which show a positive correlation between the time from symptom onset and Ct values, and that the mean time from symptom onset was shorter for BA.2 than for BA.1). As expected, symptomatic people had lower Ct values (higher viral loads) than asymptomatic people. In linear regression models between swab-positive individuals at rounds 17 to 19 (January 5 to March 31, 2022), for each of the 26 symptoms surveyed, symptom reporting was associated with a higher Ct value. bass. The lowest adjusted Ct values were for cold- or flu-like symptoms: fever, chills, sore throat, muscle aches, runny nose, sneezing, and headache (Fig. 5), which frequently coincided ( Supplementary Fig. 4). With the exception of fever, these symptoms were also commonly reported as the first symptom among positive symptomatic swab results (Supplementary Figure 5, Supplementary Table 2).