Lifestyle Changes and Weight-gain in Youth during a 4-Month COVID-19 Lockdown: A Retrospective Observational Study

Background: The coronavirus disease-2019 (COVID-19) has resulted in nationwide lockdowns, cessation of school and work, and social distancing. Reducing the negative impacts, such as weight-gain, has garnered global attention. Methods: In this dual-center retrospective observational study of 12,889 students (aged 18-20), who performed a national student physical tness battery between December 1, 2019, to January 20, 2020, and completed a follow-up online questionnaire beginning on May 1, 2020, and ending on May 23, 2020, and examined the associations between weight-gain and COVID-19 induced stress, depression and self-reported physical activity, dietary habits and sedentary time. Results: Participants gained an average of 2.6 kg (95% CI: 2.0 to 3.2 kg) for males and 2.1 kg (1.9 to 2.4 kg) for females. An increase in overweight and obese individuals according to Asian cut-off points as a demographic percentage by 4.5% and 2.7% and 4.8% and 3.4% in males and females respectively (P<0.001). Weight change was associated with sedentary time, COVID-19 induced stress, and depression score. Conclusion: These ndings suggest that limiting sedentary time during lockdowns and strategies aimed t improve lifestyle habits and psychological health may be warranted to mitigate weight-gain during the pandemic.


Introduction
The sudden emergence of coronavirus disease-2019 (COVID-19) has reverberated the world over, pushing peoples and governments into the unchartered territories of countrywide lockdowns and social distancing. On January 20, 2020, China issued a national lockdown as a measure to halt the spread of the deadly virus. While this strategy was largely successful, its negative effects will be of consequence to the health of youth, as for four months they were out of school and for the most part, stuck in the con nes of their homes. (1) Weight-gain compared to stable weight during adolescence and early adulthood is associated with a signi cantly increased morbidity of diabetes and cardiovascular disease, and an increased risk of cardiovascular and all-cause mortality. (2,3) The increase of weight is a multifactorial matter, yet existing evidence on these factors is largely limited due to a tendency in studies to do a univariate analysis of behaviors on weight-gain. Furthermore, trials that focus on weight loss tend to enlist already obese subjects and intervene with specialized diets attenuating short-term weight loss, limiting the extrapolation of the ndings to other populations. Notwithstanding, long-term observational studies and randomized controlled trials have found dietary factors, and time spent watching TV were associated with long-term weight-gain in healthy populations. (4,5) Early evidence has shown lockdown poses a real threat to psychological well-being with research citing ampli ed stress, depression and anxiety in North American, European, and Asian populations. Furthermore, in certain countries there has been a surge in unhealthy dietary habits and increased sedentary time, which could be a recipe for increased weight gain (6,7). Staying at home for a prolonged period could also support a more sedentary lifestyle and increased time watching TV or on other electronic entertainment, along with eating more calorie-dense meals, snacking and alcohol consumption.
To date, a limited number of studies have been conducted related to weight change and its associated factors during the COVID-19 lockdown period, and some of those have suggest that behavior change during the pandemic being a driving force behind weight change (6,7). This dual-center retrospective observational study assessed weight change during the pandemic lockdown through a multifactorial analysis of associated exposure variables.

Study design and participants
This retrospective observational study enrolled youth from two universities (Hunan Traditional Chinese Medical College, Hunan, China, and Medical College of Jinhua Polytechnic, Zhejiang, China) that performed the compulsory Chinese National Student Physical Fitness Standard (CNSPFS) battery between December 1, 2019, to January 20, 2020, when the government issued sanctioned lockdowns and social distancing and completed a follow-up online questionnaire beginning on May 1, 2020, and ending on May 23, 2020. A total of 14,059 university students who were free of chronic diseases and had completed the CNSPFS were screened. Of these, 13,013 participants (response rate of 93.2%) completed the online questionnaire. The participants who replied with questionnaires of poor quality were excluded (n = 124). A total of 12,889 participants were nally recruited in the study. All baseline data were extracted from the CNSPFS system, follow-up data were collected from a professional survey platform (https://www.wjx.cn). The criteria of poor quality were: (1) ID information in the CNSPFS system did not match that in the follow-up questionnaire; or (2) The 81 question survey was completed in less than three minutes.This study received ethical approval from the Medical Ethics Committee of Xiangya Hospital, Central South University (REC reference: 202005126). All participants gave electronic consent before enrolment in the study, which was conducted in accord with the principles of the Declaration of Helsinki and participants were anonymized and codi ed to protect their personal information.

Characteristics of Lockdown
The rst lockdown order in China was delivered on January 23rd, 2020 by the government of Wuhan, Hubei province, followed by the other cities and provinces of China. The main requirements included: (1) all individuals were ordered to stay home or at their place of residence, except for permitted work, local shopping or other permitted errands, or as otherwise authorized. (2) all schools, sports facilities, entertainment, and recreational venues, personal care and beauty services, and the majority of factories and markets were closed.
During the period of data collection, the participants involved would continue their studies through online learning at the beginning of their second semester of teaching. There were no imminent academic examinations in either time point of data collection.

Weight change
Baseline weight was assessed by staff members of the two respective universities. This was done in accordance with CNSPFS standards for all university students across China, using scales and after the removal of shoes/coats. Follow-up body weight was obtained via questionnaire. Concerning the determination of follow-up body weight, all participants were asked to measure their body weight in the morning right after waking up, in a state of fasting, shoes off, no large coat. Weight change was calculated by subtracting baseline weight from follow-up weight and the results were veri ed according to perceived weight change. Weight change was further classi ed as no signi cant change (-0.9 to 0.9 kg), mild decrease (-3.9 to -1.0 kg), moderate decrease (-6.9 to -4.0 kg), major decrease (≤ -7.0 kg), mild increase (1.0 to 3.9 kg), moderate increase (4.0 to 6.9 kg), major increase (≥ 7.0 kg).

Physical activity
Physical activity was classi ed as exercise and free-living activity. Exercise habits were collected through questions regarding exercise pre-and post-lockdown, these included frequency, duration, and intensity of aerobic exercise as well as strength exercise, in a modi ed version with added items of the International Physical Activity Questionnaire-Long form-Chinese (IPAQ-LC), which had shown adequate reliability and reasonable validity for use in Chinese students (8). Exercise volume was expressed in MET-hr/wk, calculated according to the American College of Sports Medicine (ACSM) metabolic equations prior to and during the lockdown.(9) Sedentary time, de ned as any waking behavior where energy expenditure is ≤ 1.5 MET's while reclining or sitting, included the number of hours a day spent using a computer or mobile phone (9) and was expressed as hr/d.

Dietary habits
In this study, food composition was not assessed due to the high participant burden of such a questionnaire and data collection limitations. The present study evaluated the dietary habits, including breakfast and lunch frequency, alcoholic drinks per week, and snacking times per day. Breakfast and lunch were categorized into three groups: less than once a week, two to six times per week, every day.
Snacking was also categorized into three groups: no snacking, snacking once a day (day or late-night), snacking twice per day (day and late-night). Alcohol habits were recorded as the number of drinks, one drink was de ned as drinking 100 ml wine or liquor, or 1 glass/bottle/can of beer.

COVID-19 induced stress, depression and anxiety
The question used to measure COVID-19 induced stress was designed based on previous research(10): How concerned are you about yourself, or family members/friends being infected by COVID-19? To which the possible answers were: None; Mild; Moderate; Major; Severe. Two psychological indexes were also used, Becks' Depression Inventory, second edition (BDI-II)(11) and the State-Trait Anxiety Inventory (STAI) (Form Y-1). (12) Statistical analysis The primary outcome (dependent variable) of the present study was weight change over the lockdown, the independent variables were COVID-19 induced stress, depression, anxiety, physical activity, and dietary habits. Paired t-test and Wilcoxon signed-rank test were used for assessment in mean difference between baseline and follow-up of continuous and ordinal variables respectively. The independent relationships of COVID-19 induced stress, depression, anxiety, physical activity, and dietary habits to weight changes within the lockdown, using multivariable linear or binary logistic regression models and accounting for within individual repeated measures were assessed accordingly. Potential nonlinear effects of decreases versus increases in each variable were evaluated by modeling changes in indicator categories, with "no change" as the reference. Multivariable models were used to adjust for sex, age, baseline weight, physical activity, dietary habits, and psychological status simultaneously. To minimize confounding from geolocation and baseline overweight/obesity, university-and BMI-strati ed multivariate linear regression were used. Additionally, results were analyzed as relative weight changes (percent), and weight change was a binary variable (weight-gain versus weight loss). Analyses were carried out with the use of SAS software, version 9.4 (SAS Institute), a two-tailed alpha level of 0.05 was considered signi cant.

Results
Demographics and Weight-gain Demographics of 12,889 participants aged 19 ± 1 are presented in Table 1. Female subjects were proportionally a larger percentage of the population, which is due to subjects being selected from medical schools largely populated by females. The average weight-gain across the universities was 2.6 (95% CI, 2.0 to 3.2) kg for males and 2.1 [1.9 to 2.4] kg for females, while across sex was 4.2% of baseline body weight. There was a signi cant increase in overweight and obese individuals as a percentage of the population by 4.5% and 2.7% and 4.8% and 3.4% in males and females respectively. According to their self-reports, none of the participants of the study were infected with COVID-19. More details are shown in Figure 1 A-E and supplementary Table 1.

Changes in physical activity
Exercise volumes per week before lockdown were 14.0 ± 12.3 and 8.9 ± 8.3 MET-hr/wk in males and females respectively. There was a mild decrease in males (mean [95% CI], -0.6 [-1.1 to -0.1]) and increase in females (2.5 [2.3, 2.7]). Sedentary times before lockdown were 6.9 ± 3.3 and 7.3 ± 3.4 hr/day in males and females respectively. There were signi cant increases in sedentary time both in males (3.9 [3.7 to 4.1]) and females (3.6 [3.5 to 3.7]). More details are shown in Figure 1F and G, and Supplementary

COVID-19 induced stress, depression and anxiety
More than one-third of all participants (34.6% of males and 42.4% of females) suffered COVID-19 induced stress in different degrees, and 6.9% of males and 7.4% of females responded to having severe stress. There was a signi cant difference in severe COVID-19 induced stress between males and females (P<0.001). The depression score obtained from BDI-II were 5 ± 8 (mean ± SD) and 6 ± 8 in males and females respectively. The anxiety score using STAI was 39 ± 10 both in males and females. More details are shown in Supplementary Table 2.
Relationships between changes in physical activity and weight change Sedentary time was independently related to weight change in males (0.53 [0.49 to 0.57] kg) and females (0.46 [0.45 to 0.48] kg). Findings were comparable, in direction and magnitude, for males and females, after strati cation for BMI < 24 and ≥ 24 kg/m 2 , and two universities (Table 3, Figure B and C, Supplementary Table 3 Asian population cut-off < 24 2 , two universities, and in analyses when evaluating weight change as an absolute and relative weight change, and as a binary variable (weight again versus weight loss) the results remained consistent (Table 3, Figure 2 A-C, Supplementary Table 3-5). However, the weight of the subjects whose stress is between mild-major had no statistical signi cance. Depression score was also independently associated with weight change in males (0.02 [0.01 to 0.05] kg) and females (0.02 [0.01 to 0.03] kg). Across quartiles, participants with a greater increase in depression score gained 0.14 kg more ( Figure 2B). There is a signi cant relationship between anxiety and weight change in females (0.01 [0.01 to 0.02] kg), but not in males (0.01 [-0.01 to 0.03] kg).

Additional Analyses
All results were similar when evaluated as relative (percent) weight changes rather than absolute weight changes, and when weight change as a binary variable (weight-gain versus weight loss) (Table 3, Figure 2 A-C, Supplementary Table 3 -5). Correlations of COVID-19 induced stress to DBI-II depression score, STAI anxiety score, physical activity and dietary habits were generally small (r<0.05).

Discussion
This is the rst multivariate analysis aimed at exploring the burden of imposed lockdowns practices on weight-gain. Longitudinal studies have shown that the average yearly weight-gain in general populations is approximately 0.45 kg a year. (4). This study has found that over a 4-month lockdown period there was signi cant weight-gain in the cohort, an average increase of 2.2 [2.2 to 2.3] kg, and an increase in overweight and obese individuals as a percentage of the population by 4.8% and 3.3% respectively across sexes in youth (P<0.001). Sharp increases in weight have been associated with an increase in blood pressure.(13) Even more moderate increases over a longer period of time are associated with chronic diseases, (14) all of which are exacerbated when weight is gained during adolescence and early adulthood to have later life repercussions. (15,16) In this study, sedentary time, rather than exercise was signi cantly associated with weight-gain during the lockdown period, as were other factors, including change in dietary habits, severe COVID-19 induced stress, and depression.
Energy balance and therefore weight-gain is heavily mediated through physical activity. Increased levels of PA, whether as exercise or free-living PA will accordingly increase energy expenditure, any decrease in such will likewise decrease energy expenditure. Furthermore, in the present paper, self-reported exercise volume decreased modestly in males and increased in females ( Figure 1F). The univariate analysis demonstrated that exercise volume was modestly associated with weight change in females alone (-0.007 [-0.014 to -0.000] kg), possibly due to the low exercise volume both before (14.0 and 8.8 MET-hr/wk in males and females) and during lockdown (13.4 and 11.3 MET-hr/wk) far below the ACSM-recommended exercise volume (16.7 METhr/wk), (9) and change in exercise volume across sexes being small. Various studies have looked into free-time and PA. While there is some disagreement within population cohorts, in university students it appears that exercise habits are largely mediated by time constraints, (22) where students perceive limited time as a primary reason for not exercising. Also, adherence to exercise is often extrinsic, particularly in youth, where exercisers do so because of physical appearance, weight-gain and stress management adding plausibility that exercise may have been increased as management for perceived weight-gain and/or stress and due to more available time.
Weight-gain also comes through increased intake of calorie-rich food, large portions, and dependent on food composition (23) leading to weight-gain. Several studies have alluded to the signi cant changes in dietary habits during lockdown. In the UK as many as 48% of research participants reported increased food intake (24) and increased detrimental dietary practices, echoed throughout western nations and the middle-east (25,26). Although this study did not report on dietary composition, it did investigate dietary habits using four parameters, namely: snacking, breakfast, and lunch frequency and alcohol intake. When adjusted for age and baseline weight, snacking frequency per day was positively related to weight change in females alone (0.46 [0.21 to 0.71] kg); while alcohol drinks per week was positively associated with weight change in males (0.03 [0.2 to 0.8] kg), but negatively associated in females (-0.07 [-0.12 to -0.03] kg). Moreover, breakfast frequency per week was positively related to the weight-gain only in females and lunch breakfast frequency per week in males only. However, after multivariable adjustment, no signi cant associations remained between weight change and dietary habits.
Evidence has been accumulating that lockdown conditions have also increased the prevalence of depression, anxiety, and stress across multiple populations. (27) The present study found that severe stress over COVID-19 was associated with a 0.49 [0.13 to 0.82] kg and 0.88 [0.15 to 1.61] kg gained weight in females and males respectively. Stressful life events, such as university/school-related stress are known to be associated with weight-gain. (28) Additionally, instances of high to extreme levels of stress can in uence eating behaviors, a factor which is more evident with higher severity of stress. (29) Stress is thought to affect PA also, a comprehensive review found the majority of 55 longitudinal studies supported an association between stress and lower levels of PA. (30) However, in our study, the correlation between stress and changes in PA or dietary habits was quite low (r<0.05). (31) Furthermore, social isolation and con nement are possible routes towards depression.(32) BDI-II depression score was associated with increased body weight in males (0.02 [0.01 to 0.05] kg) and females (0.02 [0.01 to 0.03] kg). This is in agreement with a consortium of previous studies, whereby depression appears to have a signi cant effect on short-term and long-term weight-gain. (33,34) Possibly due to a lack of motivation towards activities that require physical effort such as exercise and increased sedentary time, or the adoption of less healthy eating patterns.or. (33,35) Anxiety was a minimal contributor to weight-gain in females, but not males (Table 2), and appears to be more commonly present in females, which supports previous studies that during the pandemic, state anxiety appears to affect females more than males. (36) Limitations There are a number of limitations to this study. The methods of collecting weight in the two time periods differed, with the rst being under supervision of professionals, the second being self administered, due to the constraints of lockdown. While instructions were given to participants regarding how and when to weigh themselves and we had veri ed this with the weight the self reportedly gained over the period, measurement errors as well as over/under estimation may occur to some degree. Due to limitations in research, we have no data representative of yearly or semesterly weight gain in Chinese university students to compare our results with. Psychological status of depression and anxiety was collected at the second time-point only therefore causality may not be established from these.
Furthermore, just as with any medical intervention or organic nding, the results here represent the average public effect, and variations within individuals exist. The present cohort study only followed up the 4-month lockdown that resulted from the COVID-19 pandemic, and largely comprised Han, universityeducated Chinese youth, which potentially limits the generalizability of the ndings. However, our ndings were consistent among two universities, sex, and individuals with BMI <24 and ≥ 24 kg/m 2 , with multivariable adjustment of age, baseline weight, change in physical activity, dietary habits, psychological status, smoking status. All results remained persistent in additional analyses when evaluated weight change as relative (percent) weight changes rather than absolute weight changes, and when weight change as a binary variable. It seems plausible that the biologic effects of many factors would be qualitatively similar in other populations.

Conclusion
The COVID -19 lockdown appeared to have a more pronounced effect on sedentary time than on exercise. In this study, weight gain was accordingly predicted by increased sedentary time rather than decreased exercise. Furthermore, severe COVID-19 induced stress, depression were mutually independently associated with weight-gain during a 4-month lockdown period, but causality cannot be established. These ndings suggest that lockdowns promote sedentary time, and that strategies to improve lifestyle habits and psychological health may be warranted to mitigate weight-gain.

Declarations Con icts of interests
The authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest.
Authors contributions SXL, YSD and JWR conceived the study. SXL and YSD were involved in securing funding for the study.
NJZ, BYY, WLZ, YCD, JL, HL and QXL coordinated the study conduct and data collection. YSD and SYX did the study analyses, supervised by SXL. YSD and JWR wrote the article, with assistance from RJT, TPO, BYY, QXL, HL and SXL. All authors approved the nal version of the manuscript. SXL had full access to all the data in the study and had nal responsibility for the decision to submit for publication.