In this study, self-reported data were collected from 5919 participants aged 18 to 89, with 63% of participants being women. The data were obtained through the second wave of the International COVID Sleep Study survey (ICOSS-2), as described in detail in reference [24]. The survey was conducted anonymously between May and December 2021 and was accessible through various online platforms, including RedCap and Qualtrics. To reach a diverse audience, the survey was advertised on university web pages, newspapers, television, Facebook, and Twitter (since 2023 named X). Additionally, it was made available in multiple languages, including German, Portuguese, Brazilian Portuguese, English, French, Bulgarian, Croatian, Chinese, Finnish, Hebrew, Italian, Japanese, Norwegian, and Swedish. After applying the exclusion criteria outlined in Supplemental Table S1, 5919 participants who reported receiving two doses of mRNA SARS-CoV-2 vaccine (Moderna and BioNTech/Pfizer) had complete data for analysis. During the survey period in 2021, a third dose of SARS-CoV-2 mRNA vaccine was not yet standard. Therefore, we only inquired about participants receiving up to two doses of SARS-CoV-2 mRNA vaccine.

This study adhered to the principles outlined in the Helsinki Declaration and obtained either ethical approval or waivers in all participating countries, in accordance with their respective national research governance and regulations. Notably, ethical approval was not mandated by national law in Austria, Brazil, Finland, France, Norway, and Sweden due to the anonymous nature of the survey collection. The anonymity of the data was preserved throughout the study. Prior to accessing the questionnaire, participants were required to provide their consent to participate, with a minimum age requirement of 18 years. No monetary compensation was provided to participants. Further details regarding ethical approval in each country can be found in Supplemental Table S2.

The BMI ranges used to define normal weight, overweight, and obesity vary for Asians compared to other ethnicities [25]. Hence, we employed the following BMI thresholds for non-Asians: normal weight (reference group) <25 kg/m2, overweight 25-29.9 kg/m2, and obesity ≥30 kg/m2. Meanwhile, for participants of Asian ethnicity, their BMI status was determined as follows: normal weight <23 kg/m2, overweight ≥23 to <27.5 kg/m2, and obesity ≥27.5 kg/m2.

A recent study [13] identified 12 symptoms that persist for at least six months post-infection with SARS-CoV-2 that can be used to assess the likelihood of experiencing long COVID, using a PASC score. Specifically, each symptom was assigned a score based on its predictive ability for long COVID: loss of or change in smell or taste (8 points), post-exertional malaise (7 points), chronic cough (4 points), brain fog (3 points), thirst (3 points), palpitations (2 points), chest pain (2 points), fatigue (1 point), dizziness (1 point), gastrointestinal tract symptoms (1 point), changes in sexual desire or capacity (1 point), and abnormal movements (1 point).

We included eight symptoms that lasted for at least six months at the time of the survey to calculate the PASC score for each participant. These symptoms were: loss of or change in smell or taste, post-exertional malaise, brain fog, palpitations, chest pain, fatigue, dizziness, and gastrointestinal tract symptoms. Based on a previous study’s proposal [13], participants with a PASC score of 12 or higher were classified as having a high likelihood of experiencing long COVID, regardless of whether they reported a previous SARS-CoV-2 infection.

To address the potential bias of misclassifying participants as unlikely to suffer from probable long COVID due to the absence of survey questions about chronic cough, thirst, changes in sexual desire or capacity, and abnormal movements, a sensitivity analysis was performed (see statistical section for more details).

Based on participants’ sleep duration reports, we evaluated if participants usually slept less than six or more than nine hours per night (in the following, referred to as short or long sleep duration, respectively). The literature often uses these thresholds to discriminate short and long sleep duration from normal sleep duration [26, 27]. Participants’ insomnia risk was determined through the Insomnia Severity Index (ISI) [28], a validated questionnaire of seven items assessing the severity of insomnia symptoms and their impact on daily functioning. A score greater than 14 is indicative of moderate-to-severe insomnia. To assess the presence of OSA, we used the STOP scale [29]. Specifcally, participants were asked to respond on a 5-point Likert scale to the following four questions: (a) Do you snore loudly, surpassing the volume of talking or being audible through closed doors? Response options ranged from “Not at all” to “Every night/almost every night.” (b) Do you frequently experience daytime tiredness, fatigue, or excessive sleepiness? Response options spanned from “Not at all” to “Every day/almost every day.” (c) Has anyone ever witnessed you ceasing to breathe or choking during your sleep? Response options varied from “No, never” to “Every day/almost every day.” Participants’ answers were dichotomized into two categories for the first three questions: 0 = Less than three nights per week and 1 = Three nights per week or more. (d) We also surveyed whether participants currently had or received treatment for high blood pressure. A “Yes”-response was counted as one score. The risk of OSA was considered high if participants scored two or greater on the STOP scale.

Data are presented as mean ± SD unless otherwise specified. Group characteristics were compared using the Chi-Square test for categorical variables and generalized linear models for continuous variables. Logistic regression analyses were conducted to examine the associations between BMI group as a predictor and probable long COVID status as a dependent variable using SPSS 28.0 (IBM Corp., Armonk, NY, USA).

In addition to conducting an unadjusted logistic regression analysis, we employed one additional regression model to examine the robustness of the association between BMI status and probable long COVID. The adjusted analysis incorporated self-reported positive SARS-CoV-2 test, age, sex, race/ethnicity, smoking status, the time elapsed since the first vaccination ( ≤ six vs. > six months), urbanicity, and weekly physical activity level score (ranging from 0 to 7; higher score indicating higher physical activity; for more details, see [22]). We additionally considered a medical history encompassing hypertension, type 2 diabetes, depression, and attention deficit hyperactivity disorder. These were defined as instances where individuals had been diagnosed with or received treatment for these conditions either prior to or at the time of the survey. This inclusion was motivated by the recognition that each of these conditions commonly co-occurs with obesity [30,31,32,33].

To ensure the robustness of the hypothesized association between BMI status and probable long COVID, we conducted several sensitivity analyses:

1. We examined individuals who reported testing positive for SARS-CoV-2 before the survey (n = 515).

2. We separately analyzed data for men and women because previous findings suggest that the risk of long COVID is higher among women than men [14]. In this context, we assessed multiplicative interactions between BMI status and sex.

3. Participants from the USA were excluded from the analysis as they were significantly younger than participants from other countries (Supplemental Table S3).

4. We excluded individuals whose PASC score fell between 3 and 11, as our study did not survey four of the twelve symptoms used to calculate the PASC score in a previous study [13]. As mentioned earlier, if experienced for at least six months, the total sum of these symptoms corresponds to 9 points.

5. We excluded participants (n = 412) who were underweight from the normal weight reference group. Underweight was defined as having a BMI < 18.5 kg/m2 across all ethnicities.

To assess the potential variability in the risk of inadequate sleep associated with BMI and probable long COVID status, we conducted logistic regression analyses, both unadjusted and adjusted. Our binary outcome variables included moderate-to-severe insomnia, a high risk of OSA, short nighttime sleep duration (less than 6 hours), and long nighttime sleep duration (more than 9 hours). In the adjusted regression model, both BMI group and probable long COVID status were entered together to account for mutual adjustment. Additionally, for the sleep outcomes, we assessed multiplicative interactions between BMI group and probable long COVID status. Overall, a P value less than 0.05 was considered significant.