Gender differences in bodyweight change following COVID-19 lockdown measures in the Netherlands: a prospective longitudinal study

Abstract

Background

The current study aimed to prospectively assess bodyweight change following the implementation of lockdown measures to contain the COVID-19 pandemic in the Netherlands and to explore the potentially moderating role of gender in this association.

Design

Using Dutch DNB Household Survey panel data collected between 1993 and 2020, we estimated fixed-effects regression models of bodyweight change. Models were stratified by gender and formal tests of gender differences in coefficient estimates were performed.

Participants

4365 women and 4583 men aged 18–65 were included in the study. The total number of observations was 41 330.

Outcome measures

The outcome of interest was self-reported bodyweight in kilograms. Additional analyses were performed using body mass index (self-reported weight in kilograms divided by self-reported height in metres squared) as the outcome.

Results

The implementation of Dutch lockdown measures in 2020 was associated with bodyweight gain of approximately 800 g in working-age women compared with the 3 prior years. Bodyweight gain in 2020 relative to the prior years was significantly stronger for women than for men (F(4, 8947)=3.9, p<0.01). No evidence of bodyweight gain in working-age men was found.

Conclusion

Results indicate that bodyweight gain following COVID-19 lockdown measures in the Netherlands was more pronounced among women than among men. Although necessary to contain the COVID-19 pandemic, lockdown measures may contribute to a different public health challenge in the rising prevalence of overweight and obesity.

Strengths and limitations of this study.

• In contrast to most prior work, the current study uses a prospective approach and data from a random national sample to assess bodyweight change following the implementation of lockdown measures to contain the COVID-19 pandemic among Dutch men and women.

• The current study acknowledges that the bodyweight implications of lockdown measures may differ between men and women by estimating models stratified by gender and performing formal tests of gender differences in coefficient estimates.

• A limitation of the current study is that the measure of bodyweight is self-reported.

Introduction

On 11 March 2020, the WHO declared the novel coronavirus SARS-CoV-2, also known as COVID-19, a pandemic. Almost immediately, countries started implementing the so-called lockdown measures, such as closures of schools and gastronomy and urgent calls to work from home whenever possible, in an effort to slow down the spread of the disease. Although necessary to contain the pandemic, these measures also affected people’s daily activities in a way that may compromise health, for instance through adverse lifestyle changes.

Shortly after the introduction of the first lockdown measures, scholars already specifically expressed concerns that such measures may result in bodyweight gain.^{1 2} Although initial evidence suggests that these concerns are justified (for reviews, see Bennett et al³ and Khan et al⁴), results of the work hitherto conducted should be interpreted with caution for multiple reasons. First, most earlier studies were cross-sectional or retrospective,^{3 4} which makes estimates of bodyweight change prone to recall bias.⁵ Second, existing studies have drawn almost exclusively on non-probability samples,^{3 4} and consequently results cannot be generalised.^{6 7} Third, only few studies have explored potential gender differences in the bodyweight implications of the measures to contain the COVID-19 pandemic. This is unfortunate because, as described in further detail later, such differences may be expected given the central role of stress in the presumed mechanism linking measures to contain the COVID-19 pandemic to changes in bodyweight.

The current study assesses the impact of the measures to contain the COVID-19 pandemic on the bodyweight of working-age women and men in the Netherlands. It extends existing work on the links between the implementation of lockdown measures and bodyweight change (1) by adopting a prospective approach, (2) by drawing on data from a random national sample and (3) by acknowledging that the bodyweight implications of lockdown measures may differ between women and men.

Background and hypotheses

In March 2020 the Dutch government announced the first general measures to contain the COVID-19 pandemic. Mid-March, the Netherlands went into a so-called intelligent lockdown that included closure of gastronomy, schools and sports clubs, and a travel ban. The government also made an urgent appeal to work from home whenever possible. These measures were extended through April. During the months of May and June, the previously taken lockdown measures were relaxed somewhat and the testing policy got expanded. Over the course of July and August, the number of new COVID-19 cases started rising again, but no new measures were taken yet. At the end of September, however, the Dutch government decided that, because of the rising number of infections, additional measures were needed again and a second lockdown was announced in October 2020. Measures of the first intelligent lockdown were reintroduced. On top of this, non-essential shops had to close. The second lockdown was extended until early 2021. During this period, wearing masks in public places became mandatory and a curfew was introduced. Measures remained in place until February 2021, when the implemented lockdown measures started gradually being relaxed. In response to rising infection rates and the emergence of the SARS-CoV-2 Omicron variant, restrictive measures were reimplemented from November 2021 onwards, and a complete lockdown, including closures of non-essential shops, was announced mid-December. It is important to note that rules and urgent guidelines regarding hygiene, keeping distance, group formation and working from home as much as possible had continuously remained in place since March 2020 and were emphasised time and again by the government.

The implemented measures had considerable implications for the personal lives of adults in the Netherlands, particularly among those of working age. In a survey commissioned by the Dutch Ministry of Health, Welfare and Sport collected among Dutch adults in the spring of 2020, the majority of working-age respondents reported substantial changes to their personal situation in the wake of the measures implemented to contain the COVID-19 pandemic.⁸ One in three working-age respondents reported having started to work from home and one in nine working-age respondents reported that the measures to contain the pandemic precluded them from performing their job altogether. One-sixth of the working-age respondents moreover reported taking care of children as schools and nurseries were closed. In contrast to respondents of working age, 89% of respondents aged 65 years and older reported that little had changed in their daily lives.⁸

The changes in the daily lives of the working-age population may be expected to have detrimental lifestyle implications. In the spring of 2020, Bhutani and Cooper¹ already speculated ‘that increases in stress, anxiety, and boredom on a daily basis during the pandemic may be contributing to higher energy intake, sleep disturbances, and less exercise’ (p1576), which could ultimately result in bodyweight gain. Similarly, Mattioli et al² warned that stress resulting from quarantine and isolation measures may lead to unhealthy dietary choices and reduced physical activity. Consistent with this reasoning, research suggests that Dutch adults were more likely to be physically inactive in the spring of 2020 than in a typical spring.⁸ In an online survey collected among a nationally representative sample in April 2020, 22% of respondents moreover reported an increased consumption of snacks and sweets since lockdown measures were in place and 14% reported eating more frequently overall.⁹ These findings from the Netherlands are in line with the results of studies conducted in other contexts,¹⁰ for example, in the USA,^{11 12} Canada,¹³ UK,¹⁴ Germany,^{15 16} France,¹⁷ Denmark,¹⁸ Spain,¹⁹ China²⁰ and Japan.²¹ We therefore hypothesise that bodyweight increased among working-age women and men following the implementation of COVID-19 lockdown measures in the Netherlands.

Given that stress arguably plays an important role in linking lockdown measures to poorer dietary choices and decreased physical activity,^{1 2} gender differences in bodyweight gain following the implementation of measures to contain the pandemic may be expected. This is because these measures may elicit a stronger stress response among women than among men.^{11 22} Moreover, stress has been found to be more strongly associated with suboptimal dietary choices and bodyweight gain among women than among men.^23–25 We therefore hypothesise that bodyweight change following the implementation of COVID-19 lockdown measures in the Netherlands was more pronounced for women than for men.

Data and methods

Sample

We draw on anonymised public release data from the DNB Household Survey^26–28 (see www.dhsdata.nl), a panel survey collected annually among a random national sample of Dutch households by CentERdata at Tilburg University in the Netherlands. Data have been collected online since 1993. A basic computer and an internet connection were provided to sampled households without a computer or internet access. Currently, 28 waves of data are available.

We restricted the sample to observations of men and women of working age (18–65) who provided valid information on all variables of interest (ie, bodyweight, age, partner status, primary activity status) in at least two waves. These inclusion criteria resulted in an analytical sample of 19 468 observations nested in 4365 women and 21 862 observations nested in 4583 men.

Patient and public involvement

The DNB Household Survey is collected among the general population of the Netherlands. The panel members provided consent via a multistage agreement including the initial recruitment as well as the activation of an account (after login only) in the panel environment. Since the introduction of the General Data Protection Regulation in 2018, panel members who already participated and newly recruited panel members have been asked to give an explicit informed consent via a web form to (continue) taking part in research projects in the panel, among which the DNB Household Survey. Only respondents who complied could continue to participate in the panels.²⁸

Panel respondents have the possibility to comment on the questionnaire online, or they may call the free helpdesk with any comments. This helpdesk is open during regular office hours. In case of (technical) problems related to the completion of the questionnaires or the need for further assistance, a member of CentERdata will visit the household on appointment to help them solve the issue at hand.²⁷

Measures

Our outcome of interest is self-reported bodyweight in kilograms. Consistent with earlier studies,^{29 30} we considered values below 25 kg implausibly low and excluded observations with such values (n=122) from our sample.

The main explanatory variable, that is, whether or not lockdown measures were in place, was derived from the year of data collection. We compare observations from 2020 (the lockdown year) with observations from, respectively, 2019 (prelockdown year), 2018 (2 years before lockdown), 2017 (3 years before lockdown) and 1993–2016 (4+ years before lockdown). Consistent with prior years, the 2020 data collection took place between week 15 and week 41, when lockdown measures were in place.

Controls included in the models were age (continuous, centred on 45), age squared, primary activity (in paid employment, unemployed, student, home maker, retired, disabled, other) and presence of a partner in the household (yes, no). A brief overview of sample characteristics is provided in table 1. Descriptive statistics stratified by period of data collection (2020, 2019, 2018, 2017, and 2016 and prior) are presented in online supplemental appendix A.

Table 1.

Sample characteristics for women and men

	Women		Men
	Mean/%	n	Mean/%	n
Mean bodyweight in kilograms (SD)	72.0 (15.1)		84.1 (13.6)
Year of observation
2020 (COVID-19 year)	3.9	768	3.4	746
2019 (pre-COVID-19 year)	4.3	845	3.7	808
2018 (2 years pre-COVID-19)	3.7	714	3.2	709
2017 (3 years pre-COVID-19)	4.1	805	3.5	769
2016 and prior	83.9	16 336	86.1	18 830
Mean age* (SD)	44.4 (12.1)		46.3 (11.7)
Lives with partner	77.3	15 040	79.9	17 472
Primary activity status
In paid employment	54.8	10 666	77.3	16 906
Unemployed	2.7	516	2.7	600
Student	7.3	1412	3.0	653
Home maker	21.2	4119	1.2	255
Retired	4.0	774	8.1	1778
Disabled	4.4	864	4.3	933
Other	5.7	1117	3.4	737
Number of observations	19 468		21 862
Number of persons	4365		4583

Data are from the DNB Household Survey 1993–2020.

*Based on values before centring.

Statistical analysis

We performed fixed-effects regression analyses of intraindividual bodyweight change,³¹ in which within-person means over time are subtracted from scores in each observation for both outcome and explanatory variables. Consequently, all time-invariant characteristics, regardless of whether observed, are accounted for and omitted variable bias issues are limited to time-varying factors. We regressed within-person bodyweight change on the year of observation and adjusted for the aforementioned controls.

Models were stratified by gender. In order to assess whether coefficient estimates significantly differed between women and men, we additionally estimated a pooled model with interaction terms to allow the slopes of all explanatory variables to vary as a function of gender.³² All models were estimated with robust standard errors to account for the nested nature of the data.

Results

The results of our fixed-effects analyses are presented in table 2. As hypothesised, the fixed-effects model adjusted for age, age squared, partner status and primary activity indicated that women’s bodyweight increased significantly in the COVID-19 lockdown year of 2020 relative to 2019, 2018, 2017, and the period 2016 and earlier. The estimated magnitude of the adjusted bodyweight increase in women in 2020 relative to the 3 preceding years was approximately 800 g.

Table 2.

Results of fixed-effects analyses predicting bodyweight change in women and men (coefficient estimates with 95% CI)

	Women		Men		Women vs men
	b	95% CI	b	95% CI	Δb	95% CI
Year
2020 (COVID-19 year)	Ref		Ref		Ref
2019 (pre-COVID-19 year)	−0.804**	−1.322 to −0.285	0.043	−0.432 to 0.518	−0.847*	−1.550 to −0.144
2018 (2 years pre-COVID-19)	−0.799*	−1.504 to −0.095	−0.028	−0.489 to 0.432	−0.771†	−1.612 to 0.071
2017 (3 years pre-COVID-19)	−0.816*	−1.485 to −0.148	0.250	−0.258 to 0.758	−1.067*	−1.906 to −0.227
2016 and prior	−1.133**	−1.962 to −0.303	0.883**	0.274 to 1.492	−2.016***	−3.044 to −0.987
Time-variant controls
Agea	0.255***	0.197 to 0.312	0.363***	0.313 to 0.412	−0.108**	−0.184 to −0.32
Agea (squared)	−0.003*	−0.006 to −0.001	−0.007***	−0.009 to −0.004	0.003†	−0.000 to 0.007
Lives with partner	0.944†	−0.061 to 1.949	0.489	−0.487 to 1.465	0.455	−0.956 to 1.856
Primary activity status
In paid employment	Ref		Ref		Ref
Unemployed	0.064	−0.710 to 0.838	−0.182	−0.860 to 0.495	0.246	−0.782 to 1.274
Student	0.247	−0.338 to 0.833	−0.368	−1.339 to 0.602	0.616	−0.518 to 1.749
Home maker	−0.073	−0.596 to 0.450	−0.713	−2.154 to 0.729	0.640	−0.893 to 2.173
Retired	0.022	−0.656 to 0.700	−0.704†	−1.486 to 0.078	0.727	−0.308 to 1.761
Disabled	−0.218	−1.302 to 0.867	−0.390	−1.283 to 0.502	0.173	−1.231 to 1.577
Other	−0.038	−0.718 to 0.643	−0.119	−0.762 to 0.525	0.081	−0.856 to 1.017
Number of observations	19 468		21 862
Number of persons	4365		4583

Data are from the DNB Household Survey 1993–2020.

^aCentred on age 45.

*P<0.05, **P<0.01, ***P<0.001, †P<0.1.

Ref, reference.

For men, no significant differences between the year 2020 and the 3 preceding years were found. Interestingly, however, the year 2020 was associated with a significant bodyweight decrease relative to the period 2016 and earlier, but this decrease could not be attributed to the lockdown measures of 2020 because it already manifested itself in 2019, 2018 and 2017. The analyses thus did not provide support for the hypothesised bodyweight weight gain in men following the implementation of the Dutch lockdown measures.

The finding of a significant bodyweight gain in women but not in men is insufficient to conclude that the bodyweight gain in women was significantly stronger than in men.³³ We therefore performed formal tests of differences between the coefficient estimates in the model for women and those in the model for men. As shown in the final columns of table 2, the results indicated that bodyweight change in 2020 relative to, respectively, 2019, 2017, and the period 2016 and earlier was significantly stronger for women than for men. The gender difference in the estimated bodyweight change between 2018 and 2020 was marginally significant (p=0.07). The combined gender differences in the estimates of the year effects were statistically significant (F(4, 8947)=3.9, p<0.01). These results are consistent with our hypothesis that bodyweight gain following the implementation of the Dutch lockdown measures was more pronounced among women than among men.

Discussion

The current study extends prior work on the bodyweight implications of measures to contain the COVID-19 pandemic by adopting a prospective approach, by drawing on data from a random national household sample and by acknowledging that the bodyweight implications of lockdown may differ between men and women. Our analyses indicate that the Dutch lockdown measures were associated with bodyweight gain of approximately 800 g in working-age women. The effects of lockdown measures on bodyweight in working-age men were significantly less pronounced. In fact, no statistically significant evidence that the measures were associated with bodyweight gain in men was found.

We presented the results of analyses of weight change in kilograms because the interpretation of this outcome measure is highly intuitive. It could be argued, however, that a similar bodyweight gain in kilograms is more meaningful for shorter persons than for their taller counterparts. We therefore also estimated models with body mass index (self-reported weight in kilograms divided by self-reported height in metres squared) as the outcome measure. The results of these analyses (see online supplemental appendix B) were substantively similar to the results of the analyses of bodyweight in kilograms presented in table 2.

The results presented here are consistent with our hypotheses built on prior work showing that both the stress response to the pandemic^{11 22} and the association between stress and bodyweight gain^{23 24} were stronger in women than in men. However, given the absence of a stress measure in the data used, we cannot be conclusive that stress indeed plays a central role in the mechanism underlying bodyweight increases following the COVID-19 lockdown measures in the Netherlands. Such bodyweight gains could arguably also be related to losing the exercise associated with a physically demanding job, which may be ruled out via lockdown. However, we estimated additional models in which we dichotomised primary activity status (in paid employment vs not in paid employment) and allowed the lockdown effects to vary as a function of whether one was employed or not. The results of these analyses did not provide evidence that bodyweight gain was more pronounced for persons who were in paid employment than for their counterparts who were not (see online supplemental appendix C). Moreover, in 2020 approximately 20% of male workers in the Netherlands had a physically demanding job versus approximately 15% of female workers.³⁴ If bodyweight gain following the lockdown measures implemented in the Netherlands were attributable to loss of the exercise that comes with having physically demanding jobs, one might therefore have expected more pronounced effects for men than for women. Yet the opposite pattern was found in our analyses.

An important limitation of the current study is that our measure of bodyweight was self-reported. Self-reports of bodyweight are, on average, lower than measured bodyweight.^{29 35} However, given that the extent to which people under-report their weight tends to be stable over time, within-person changes in self-reported bodyweight, such as analysed here, have been found to only have minor discrepancies with changes in measured bodyweight.³⁶

Overweight and obesity currently account for almost 4% of the total burden of diseases in the Netherlands.³⁷ Given that the prevalence of overweight and obesity is projected to increase in the next decades,³⁷ this percentage may be expected to rise even further. Our results suggest that the Dutch measures to contain the COVID-19 pandemic may aggravate this trend. Research has shown that short-term weight gain, for instance during the holiday season, often tends to be retained and that it is a major contributor to long-term excess bodyweight.³⁸ Working-age adults typically gain bodyweight with every additional year of age. For instance, Peeters et al³⁹ reported an average annual bodyweight increase of 0.34 kg among Australian adults, and Orpana et al⁴⁰ found that Canadian men gained 0.74 kg and women 0.57 kg over 2 years. These estimates are approximately similar to our estimates of annual bodyweight gain in Dutch working-age women (0.26 kg at age 45, 95% CI 0.20 to 0.31; see table 2) and men (0.36 kg at age 45, 95% CI 0.31 to 0.41; see table 2). It is worth noting that the additional estimated bodyweight gain in women associated with the COVID-19 lockdown measures was three times larger than the estimated bodyweight gain associated with a 1-year age increase. Differently put, our model suggests that women’s increase in bodyweight between 2019 and 2020 was approximately equivalent to what they in non-COVID-19 times would have gained over 4 years rather than 1 year. Restrictive measures furthermore remained in place after the period analysed here. Future studies could extend the current study by analysing data from upcoming DNB Household Survey data waves to test whether concerns about the persistence of bodyweight gain related to the implementation of lockdown measures are justified.⁴¹

Data availability statement

Data are available in a public, open access repository. Data are available for scholarly reuse upon registration via www.dhsdata.nl.

Ethics statements

Patient consent for publication

Not required.

Ethics approval

This study involves human participants and was assessed and approved by the Research Ethics Review Committee of the Erasmus School of Health Policy & Management at Erasmus University Rotterdam (reference: 21-030).