This was a longitudinal, preregistered (https://osf.io/gwfdb/) online survey using Qualtrics platform (www.qualtrics.com) following up a survey on nutritional quality and food choice motives before and during the first lockdown in France conducted in spring 2020; results have been published elsewhere (10). In the 2020 survey, we retrospectively recorded at the same time dietary data and food choice motives for 938 individuals during the month before the first lockdown (from 17 February to 16 March 2020) and during the first month of lockdown (from 17 March to 16 April 2020). A number of 938 individuals who took part in the 2020 survey were emailed exactly 1 year later, that is on 2 May, 2021. They were asked to answer the same set of questionnaires as in 2020 regarding the 1-month period from the 1 to 30 April 2021: a food frequency questionnaire (FFQ), a food choice motives questionnaire, and a questionnaire on organic and local food consumption. We thus recorded longitudinal data on three 1-month periods per individual: before the first lockdown, during the first lockdown, and 1 year after the first lockdown.

Data were collected from 2 to 15 May to 2021. The participants were all registered in the Chemosens Platform's PanelSens database declared to the relevant authority (Commission Nationale Informatique et Libertés; CNIL; n°1148039). Eligible participants took part in the 2020 survey and had not moved since then. All participants were contacted by email and provided consent for their participation after being informed that the purpose of the study was to investigate food choices 1 year after the first lockdown. They received a 10 € Amazon voucher in return for their participation. As in the 2020 survey, three attention check questions (e.g., “How many times have you visited the planet Mars?”) were included in various parts of the questionnaire. The study was approved by the relevant ethical evaluation committee for research (reference: Comité d'Evaluation Ethique de l'Inserm n°20-683bis, delivered on 30 March 2021). All the questionnaires (in French) are available on the OSF page of the project (https://osf.io/gwfdb/).

Participants reported their diet for the previous month using a validated FFQ that includes 109 foods, 12 non-alcoholic drinks, and 4 alcoholic drinks with frequency assessed by a 6-item scale from “never” to “several times a day” (23). Food groups included in the FFQ were as follows: bread and cereals (4 items), raw vegetables (4 items), cooked vegetables (13 items), starchy foods (10 items), meat and eggs (15 items), fish (9 items), mixed dishes (12 items), dairy products (10 items), fruit (11 items), biscuits and cakes (14 items), and sauces (7 items). Usual portion sizes were estimated with the photographs for different food types on a 5-level scale, derived from the SU.VI.MAX portion book (24) or using average portion size. The consumption frequency of each item was transformed into daily frequency, and daily intake was calculated by multiplying the daily frequency by the estimated portion size. Individual nutrient intakes were calculated by multiplying the daily intake of each food item by its nutritional values from the SU.VI.MAX nutrient composition database (25).

Adherence to the French dietary recommendations before, during, and 1 year after the first lockdown was evaluated using the simplified PNNS-GS2 (sPNNS-GS2), an index previously designed by Chaltiel et al. (26) to reflect adherence to the main 2017 French dietary recommendations. The sPNNS-GS2 builds on the distinction between malus components (less healthy food groups which consumption should be limited, carrying a negative score, i.e., red meat, processed meat, sugary foods, sweet-tasting beverages, alcoholic beverages, and salt) and bonus components (healthier food groups carrying a positive score, i.e., fruits and vegetables, nuts, legumes, whole-grain food, milk and dairy products, fish, and seafood). The sPNNS-GS2 was computed for each participant with slight modifications to the calculation (range: 17–11.5). The sPNNS-GS2 originally included bonus points for added fat below 16% of energy intake (26). The FFQ did not make it possible to calculate the percentage of energy intake accounted for added fat, especially added oils, and this component was not included in the score calculation, as in Marty et al. (10). We also calculated short-term Δ nutritional quality and mid-term Δ nutritional quality, as the difference in sPNNS-GS2 between during and before the first lockdown and between 1 year later and before the first lockdown, respectively. Δ nutritional quality > 0 indicated an increase in nutritional quality over time.

Greenhouse gas emissions (GHGE) in kg CO2eq/kg of the individual diets were calculated before, during, and 1 year after the first lockdown as an indicator of the environmental impact of the diet (27). GHGE were derived from the French food environmental impact database Agribalyse 3.0 drawn up by the French Agency for Ecological Transition that includes GHGE values for 2,480 common food items based on Life Cycle Analyses of food products (28). The items of the FFQ were associated with all the corresponding food items from Agribalyse 3.0. GHGE of each item of the FFQ were calculated as the average GHGE of individual foods from Agribalyse 3.0 associated with each item. GHGE of participants' daily diets were calculated by multiplying the daily intake of each food item by its associated GHGE per kg. We also calculated short-term Δ GHGE and mid-term Δ GHGE, as the difference in GHGE of diets between during and before the first lockdown and between 1 year later and before the first lockdown, respectively. Δ GHGE > 0 indicated an increase in GHGE over time.

Participants answered questions about their consumption frequency of organic and local food products for 12 food categories on a 3-point scale: 2 = most of the time, 1 = occasionally, and 0 = never. The 12 categories were as follows: fruit, vegetables, dairy products, meat and fish, eggs, grains, bread, oil, ready-to-eat meals, biscuits, tea and coffee, and wine and beers. This questionnaire was adapted from a previously published version that assessed organic food consumption only and included six additional food or non-food product categories (29); we excluded the six additional categories because they were not part of the FFQ. Organic consumption scores and local consumption scores were calculated before, during, and 1 year after the first lockdown as the mean of participants' responses across the 12 food categories.

Food choice motives were assessed using a French version of the Food Choice Questionnaire developed in English by Steptoe et al. (30) and adapted by Cottet et al. (31). The French version included 24 items and nine subscales: health (3 items), convenience (3 items), sensory appeal (3 items), natural content (3 items), ethical concern (2 items), weight control (3 items), mood (3 items), familiarity (2 items), and price (2 items). For each subscale, scores before, during, and 1 year after the first lockdown were computed by averaging ratings for individual items. The scores ranged from 1 to 4: 1 = not at all important; 2 = a little important; 3 = moderately important; 4 = very important. Short-term Δ motives and mid-term Δ motives were calculated as the difference of the score for each of the nine subscales between during and before the first lockdown and between 1 year later and before the first lockdown, respectively. Δ motives > 0 indicated an increased importance of the motives over time.

Participants were asked for their gender, employment status, highest educational qualification, dietary restriction (none, vegetarian, vegan, gluten-free, sugar-free, lactose-free, allergies), dieting status (yes or no), and weight and height at the time of completion of the initial 2020 online survey. Self-reported body mass index (BMI) was calculated in kg/m². In 2021, they were asked about their perceived change in eating habits (yes a lot, yes moderately, yes a bit, and no) and in time spent cooking (7-point scale from increased a lot to decreased a lot) during the pandemic. COVID-19-related questions (being or having been ill) were also included for descriptive purpose.

We followed an analytic plan that was preregistered prior to 2021 data collection (https://osf.io/gwfdb/). Any deviations from the preregistered analytic plan are described in Supplementary Table S1. Only participants who completed the 2020 and 2021 surveys were included in the analyses. Participants who failed at least one attention check were excluded. We analyzed data from participants who reported plausible energy intake before, during, and 1 year after the first lockdown, that is, ≥500 and ≤ 3,500 kcal/ day for women, and ≥800 and ≤ 4,000 kcal/day for men (32, 33). We compared participants that were included and excluded from the present analyses based on sociodemographic measures and dietary outcomes (2020 data).

Cronbach's α were calculated for the organic and local food consumption questionnaire (before, during, and after 1 year, respectively): organic score (0.91; 0.90; 0.89) and local score (0.87; 0.88; 0.85). Cronbach's α were also calculated for each food choice motive (before, during, and after 1 year, respectively): health (0.70; 0.71; 0.74), convenience (0.89; 0.85; 0.88), sensory appeal (0.65; 0.66; 0.64), natural content (0.87; 0.86; 0.85), ethical concern (0.64; 0.63; 0.69), weight control (0.81; 0.84; 0.83), mood (0.64; 0.64; 0.66), familiarity (0.63; 0.62; 0.76), and price (0.58; 0.63; 0.66).

For primary analyses examining the effect of the COVID-19 outbreak on the nutritional quality and environmental impact of diet, linear mixed models were used to test the effect of time (categorical: before, during, and after 1 year of the first lockdown) on sPNNS-GS2 and GHGE, with random effect of participant to account for correlation between repeated measures. For secondary analyses examining the effect of the COVID-19 outbreak on organic and local food consumption, linear mixed models were used to test the effect of time on organic and local consumption scores, with random effect of participant to account for correlation between repeated measures. For secondary analyses examining the effect of the COVID-19 outbreak on food choice motives, linear mixed models were used to test the effect of time on the nine subscales of the Food Choice Questionnaire, with random effect of participant to account for correlation between repeated measures. All linear mixed models were replicated controlling for age, gender, educational level, and declared initial BMI. Pairwise adjusted mean comparisons were conducted for all raw and adjusted models.

Exploratory analyses aimed to examine the influence of short- and mid-term changes in food choice motives on short- and mid-term changes in the nutritional quality and environmental impact of diet, respectively. Four multiple linear regressions testing the effect of Δ motives on Δ nutritional quality and Δ GHGE at both short- and mid-term were run. They were also replicated controlling for age, gender, educational level, and declared initial BMI.

All statistical analyses were performed using SAS version 9.4 (SAS Institute, Inc. 2013 SAS^® 9.4. Cary, NC). The level of significance was set at p < 0.05 for preregistered primary, secondary, and exploratory analyses.

A number of 938 participants who took part to the 2020 survey were contacted by email and 594 consented to participate. We excluded 18 participants who did not complete the study, 17 who failed an attention check, 34 who reported implausible energy intake, and one who reported an incorrect identification number resulting in a sample size of 524 participants with correct responses before, during, and 1 year after the first lockdown. Table 1 presents participants' characteristics. The participants in the 2021 study were slightly older than the participants in the 2020 study who did not complete the 2021 survey (Supplementary Table S2) but did not differ regarding gender, employment status, highest educational qualification, initial BMI, and dietary outcomes before the first lockdown (energy intake, sPNNS-GS2, and GHGE). In 2021, 72% of the participants declared that their eating habits have changed since the beginning of the pandemic and 96% reported an increase in time spent cooking. Moreover, 86 participants (17%) declared that they have been affected by COVID-19 since the beginning of the pandemic.

A significant effect of time was found on nutritional quality and GHGE of diet and on organic and local food consumption (Table 2). Total energy intake and GHGE per day increased on the short-term then remained constant. We observed a short-term decrease but no mid-term changes in nutritional quality and GHGE per 2,000 kcal of diet whereas the increase in organic and local food consumption was mid-term only. A significant effect of time was also found on all food choice motives, except for sensory appeal motives that remained stable over time (Table 2). Regarding the evolution of food choice motives over time, we observed a short-term increase in health, natural content, ethical concern, and weight control motives but no mid-term increase and even an overall decrease compared to before the first lockdown for ethical concern motives. On the contrary, the decrease in convenience, familiarity, and price motives was both short-term and mid-term as was the increase in mood motives. The results were the same in both raw and adjusted models.

As exploratory analyses, we examined the associations between short- and mid-term changes in food choice motives and short- and mid-term changes in nutritional quality and environmental impact (GHGE/2,000 kcal) of diet (Table 3). Short- and mid-term increases in weight control motives were associated with short- and mid-term increases in nutritional quality, respectively, and also with short- and mid-term increases in GHGE/2,000 kcal. On the contrary, short-term increase in price motives was associated with short-term decrease in nutritional quality, and with short-term decrease in GHGE/2,000 kcal. Mid-term increase in natural content motives was associated with mid-term increase in nutritional quality only. Short-term increase in mood motives was associated with short-term increase in GHGE/2,000 kcal. On the contrary, mid-term increase in ethical concern motives was associated with mid-term decrease in GHGE/2,000 kcal.