Frequency of consumption of home cooked meals

Exposure was derived from an item in the participant questionnaire: ‘When eating your main meal at home, how often do you usually eat home cooked meals?’ Response categories were: never or rarely; one to two times per week; three to five times per week; or more than five times per week. The first two response categories were collapsed to yield appropriate numbers for statistical analysis, as previously [24], giving a three category variable: less than three times per week, three to five times per week, and more than five times per week.

Indicators of diet quality

We assessed a range of dietary outcome variables, namely Mediterranean Diet Score (MDS) [25], Dietary Approaches to Stop Hypertension (DASH) score [26], plasma vitamin C, and fruit and vegetable intakes. Participants completed a 130-item, semi-quantitative food frequency questionnaire (FFQ) for their food intake over the previous year [27], which has been shown to yield valid and reproducible food intake assessments, and has been validated previously in dietary data collection in the European Prospective Investigation into Cancer and Nutrition (EPIC) studies [28]. The FFQ EPIC Tool for Analysis was used to convert food intake frequency to energy, nutrient and food intakes [29]. Total daily intake was provided in grams for carbohydrate, fibre, fat, saturated fat, sugar, protein, fruit, vegetables and alcohol. Total daily sodium intake was measured in milligrams, and total daily energy intake in kilojoules. Dietary intake values were winsorized at 1st and 99th percentiles, by replacing the smallest and largest percentage values in the distribution with the observations closest to them [30]. This was undertaken to account for their positively skewed distribution, and the limitations of the FFQ as a tool to collect precise data on dietary intake [31, 32]. Data on dietary supplements were not collected.

The consumption of a more DASH accordant diet is associated with positive health indicators and lowered cardio-metabolic risk [33–35]. The DASH diet assumes that beneficial impact is derived from the overall diet, rather than individual foods or nutrients playing important roles [36]. A DASH score was computed from each participant’s dietary intake using the method developed by Fung et al. [26]. This index includes eight components (one nutrient and seven food groups) based on eating guidance from the United States (US) National Heart, Lung and Blood Institute [37]. Scoring is established through quintile rankings, on the basis of relative comparisons to the rest of the sample, with men and women classified separately. Participants are allocated a score from one (lowest quintile) to five (highest quintile) for energy-adjusted intake of: low-fat dairy products; whole grains; nuts, seeds and legumes; fruit (includes fruit juice); and vegetables (excludes potatoes). In contrast, for intakes of red and processed meat; sodium; and sugar-sweetened beverages, participants are allocated a score from one (highest quintile) to five (lowest quintile). Scores are then combined to give a total DASH score, ranging from a minimum of eight to a maximum of 40 points. In this study, DASH scores were standardised using the z-score, to yield a semi-continuous measure of participants’ relative standing.

The Mediterranean diet is generally considered to be low in consumption of red meats, moderate in consumption of fish, poultry, fermented dairy products and wine, and high in consumption of fruits, legumes, cereals and olive oil [38, 39]. Concordance with the Mediterranean diet has been linked with positive health outcomes, in particular the primary prevention of cardiovascular disease [40]. A Mediterranean diet score (MDS) was calculated from each participant’s dietary intake using sex-specific tertiles, according to relative comparisons with the rest of the sample. Scores of zero, one or two were allocated for each of nine dietary components, including legumes; fruit and nuts; vegetables; ratio of monounsaturated and polyunsaturated fatty acids to saturated fatty acids; fish; meat products; dairy products; cereals; and alcohol [25]. In order to appraise quality of diet independent of quantity, dietary intakes were adjusted to a 2000 kcal/day diet using the residual method. This also aimed to help reduce measurement errors, since energy intake is partially associated with over-reporting and under-reporting of dietary intake [41]. MDS scores were then standardised using the z-score.

Plasma vitamin C (μmol/l) provides an objective biomarker of fruit and vegetable consumption [42] and fruit and vegetable intake is promoted in dietary guidelines [43, 44]. Fasting venous blood samples drawn into heparin-containing tubes and stabilised using metaphosphoric acid (10%) were measured for plasma vitamin C levels by fluorometric assay within two months, as undertaken previously [45].

Markers of cardio-metabolic health

We used body mass index (BMI), percentage body fat, haemoglobin A_1c (HbA_1c), cholesterol and hypertension as indicators of cardio-metabolic health. Elevated total cholesterol and low levels of high density lipoprotein cholesterol (HDL) are associated with increased risk of cardiovascular disease [46], and the derived ratio of total cholesterol to HDL is used in the QRISK2 model to estimate risk of cardiovascular disease over the next ten years [47]. HDL and total cholesterol were measured in mmol/l in fasting venous blood samples, and the ratio of total cholesterol to HDL calculated for analysis. In line with UK guidance, a ratio of 4.0 or greater was used to indicate higher risks to cardio-metabolic health [48].

Excess body fat and raised BMI have been associated with increased risk of various NCDs [49]. Height and weight were measured at the clinical sites by trained observers, with participants wearing light clothing and barefoot. Height was measured to the nearest 0.1 cm using a wall-mounted calibrated stadiometer (SECA 240, Birmingham, UK). Weight was measured to the nearest 0.1 kg with a calibrated electronic scale (TANITA, BC-418MA, Tokyo, Japan). BMI was derived as weight (kg) divided by height (m²). Dual-energy X-ray absorptiometry (DEXA; Lunar prodigy advanced fan beam scanner (GE Healthcare)) was used to assess body composition, and has been described in detail elsewhere [22]. A three-compartment model (fat mass, fat-free mass and bone mineral mass) was used to estimate percentage total body fat. In line with international guidance, overweight was defined as BMI 25 kg/m² and above [50] and excess percentage body fat as 25% and over for males and 38% and over for females [51].

Haemoglobin A_1c (HbA_1c) has previously been used to assess risk of developing type II diabetes [52]. Participants’ HbA_1c was measured on entry to the study from fasting venous blood samples, in either mmol/mol or as a percentage. A conversion algorithm was used to convert all measurements to mmol/mol, and in accordance with international guidance [52], a level of 42.00 mmol/mol (6.0%) or higher was used to indicate increased risk of type II diabetes.

Hypertension is associated with an elevated risk of developing cardiovascular disease [53]. Using an upper arm cuff and automated oscillometric device, three sets of diastolic and systolic blood pressure measurements were performed on each participant. The first readings were discarded and the lowest systolic and lowest diastolic readings from the last two readings were used for assessment. In adherence to UK guidance [54], readings of at least 90 mmHg diastolic and 140 mmHg systolic were considered indicative of hypertension. Participants currently taking hypotensive medication, or self-reporting a diagnosis of hypertension from a clinician, were also classified as hypertensive.

Covariates

In view of the current evidence base regarding factors influencing dietary intake [55], a self-administered questionnaire was used to collect demographic and behavioural variables including sex, age, smoking status (current/ex-smoker or never smoker), and first degree family history of relevant diseases such as type II diabetes. Participants were asked whether or not they had been employed in the past four weeks, and those answering yes were identified as currently working. Participants reporting more than 48 h working in any one week were identified as working overtime. Socioeconomic status was assessed using age at leaving full time education, which was divided into three categories: education up to age 16 years (compulsory education); over 16 and up to 18 years (post-compulsory school education); and over 18 years (higher education).

Physical activity was measured objectively using an integrated movement and heart rate sensor (Actiheart; CamNtech, Cambridge, UK) attached to the chest via two standard ECG electrodes and worn during free-living over six days [56]. A ramped treadmill protocol test was used to individually calibrate heart rate, as undertaken previously [57]. Monitoring data were cleaned for measurement issues and sensor wear time was specified as at least 48 h, although data were not necessarily spread over a full 24 period. Periods of non-wear were inferred from the combination of non-physiological heart rate and prolonged periods of inactivity, which were taken into account to minimise diurnal information bias when summarising the intensity time-series. Data were processed [58] and a branched equation framework [59] used for modelling to estimate intensity time series. These were collated over time to yield daily physical activity energy expenditure (kJ/kg per day).

Strengths and weaknesses of the study

The Fenland study is a large cohort, with detailed sociodemographic data, objective physical measurements and samples, and comprehensive dietary measures. Participants in this study were from the county of Cambridgeshire, which is representative of the wider population in England in terms of adult obesity and several lifestyle variables, such as smoking and levels of physical activity [61].

Overall diet quality was assessed using two composite diet scores, DASH and MDS. Using two composite scores provided robust evidence in support of potential associations between consuming home cooked meals more frequently and higher diet quality. These results were supported by similar associations with higher fruit and vegetable intakes, measured by both FFQ, and plasma vitamin C as a biomarker. We used consumption, rather than preparation, of home cooked meals as our exposure, which is likely to be closer on the potential causal pathway to diet and health outcomes. The use of objective measurements for determining cholesterol ratio, BMI, percentage body fat, HbA_1c level and hypertension is likely to increase the validity of these markers of cardio-metabolic status, and the confidence in conclusions drawn from resultant analyses.

This research is also subject to some limitations. The cross-sectional nature of the data means that direction of cause and effect cannot be established, although follow-up data collection in the Fenland study is currently underway, which will enable future longitudinal analysis. Participants were recruited between the ages of 29 and 64 years, and are therefore not representative of the full UK population age range. Given that food preparation practices vary with age [62], our results may not be generalizable to younger populations. We excluded participants with missing data on any of the analytic variables, and excluded participants were systematically different from the rest of the cohort in terms of certain characteristics (see Additional file 2). Furthermore, differences in cooking and food culture internationally may mean that the relationships between consuming home cooked meals, diet quality, and cardio-metabolic health, vary between countries. Therefore, our findings may not necessarily be generalizable to other populations.

The fruit and vegetable intakes and DASH and MDS dietary scores were derived from FFQ data, which although validated, may be subject to error and biases [31, 32]. The composite scores assessed diet quality relative to other participants, rather than establishing absolute values, and ranking groups may constitute a broad range. The exposure variable for consumption of home cooked meals was derived from a questionnaire item, and given the absence of consensus on home cooking terminology [63, 64], participants may have interpreted this question differently. We collected data specifically on home cooked meals eaten at home and not those eaten elsewhere, such as packed lunches taken to work or place of study. The self-reported nature of several sociodemographic and behavioural variables, such as smoking, may have led to variables being correlated with each other, with associated risk of type II analytical errors.

Although we adjusted for a number of relevant potential confounders in our analyses, residual confounding remains possible. If people who consume home cooked meals more frequently are also more likely to engage in other health promoting behaviours, this could artificially strengthen associations between increased consumption of home cooked meals and markers of cardio-metabolic health.

Interpretation of findings in the context of existing research

Our findings reflect those of others that found associations between home food preparation and cooking and higher quality diets. A recent systematic review [18] identified that potential benefits included intake from healthier food groups [19, 65, 66]; greater fruit and vegetable preference and healthy eating self-efficacy [67]; enhanced nutrient intake [7, 68]; higher Diet Quality Index-International score and intake from healthier food groups [20]; trend towards higher Healthy Eating Index score [69]; consumption of a healthful dietary pattern [70]; and improved adherence to: Healthy People 2010 dietary intake objectives [8], Balance of Good Health (now Eatwell Guide) criteria [71], and a Mediterranean diet using the KIDMED index [72]. A greater frequency of home cooked meals has also been associated with higher Healthy Eating Index scores [24]. However, the majority of this research has been cross-sectional and therefore unable to conclusively indicate direction of causation. Most studies have also employed self-reported measures, which are vulnerable to bias [73], and have used food preparation practices as an exposure, rather than the consumption of home cooked food itself.

Our results also support previous studies that identified associations between home food preparation and cooking and potential advantages to health. Greater home cooking frequency has been linked with longer lifespan [10] and more frequent consumption of meals prepared at home has been associated with reduced risk of developing type II diabetes [12]. Amongst adolescents, healthier home cooking by a caregiver was linked with lowered risk of having an overweight or obese BMI [11]. However, our findings conflict with a US study that reported more time spent on home food preparation and associated clean-up at baseline, or increased involvement over time, was linked with an adverse cardio-metabolic profile [16]. Possible reasons for this discrepancy include that the US study used time spent preparing meals, rather than meal consumption, as the exposure, and the exposure included clean-up time, which may have a differential impact on cardio-metabolic health. Since food preparation activities are strongly patterned by gender [18, 74], this may also confound observed associations with health.

Meaning of the study: possible mechanisms and implications for clinicians and policymakers

Our findings indicate that an increased frequency of consuming home cooked meals is associated cross-sectionally with markers of a healthier diet, and indicators of improved cardio-metabolic health, particularly in terms of adiposity, cholesterol and diabetes risk. Links between more frequent consumption of home cooked meals and dietary benefits could be attributable to healthier food preparation methods, increased dietary variety and/or consumption of healthier food groups. Such links may also be due to decreased intake of convenience foods, which tend to prioritise ingredients such as fat, sugar and salt to increase palatability and preservation, over those for optimising health [75].

The association between a higher frequency of consuming home cooked meals and potential benefits for health in terms of hypertension was not significant in the unadjusted model, and in terms of cholesterol was no longer significant after adjustment for sociodemographic and behavioural variables. This may be because the hypertension variable was poorly ascertained, since in addition to blood pressure measurement, participants were required to report on any previous diagnoses of hypertension, and receipt of hypotensive medication. However, we conducted a sensitivity analysis for the relationship between frequency of consuming home cooked meals and hypertension, with the inclusion and the exclusion of participants diagnosed with hypertension by a doctor and/or receiving hypotensive medication. Regardless of whether or not these participants were excluded, the relationship was not significant. Cholesterol is strongly genetically determined [76], and the impact of home cooked meal consumption may not have been sufficient to result in statistically significant changes.

The cross-sectional association between higher frequency of consuming home cooked meals and lower adiposity was robust to adjustment for sociodemographic, lifestyle, and dietary covariates, whilst the association with lower likelihood of being classified as at risk of diabetes according to HbA_1c level was borderline significant. Although the direction of causation cannot be established, this indicates that home cooking potentially confers benefits to health, beyond those mediated through dietary changes. Such benefits from eating home cooked meals might be attributable to consumption of smaller portion sizes [77]; moderated snacking behaviour [78]; more structured mealtimes and the time of day at which meals are consumed [79]. Increased social cohesion has been linked with potential health benefits [80], and it is plausible that higher social capital may be associated with more sociable eating patterns. Given the potential time and effort involved in home cooking, home cooked meals may be more likely to be shared together than meals from other sources, and a range of benefits to diet, health and wellbeing derived from shared mealtimes have been identified [81, 82].

Our results support previous research indicating putative benefits from home cooked meals, suggesting that public health promotional messages should advocate for cooking at home as a positive approach for improving diet and health. Strategies could also be considered for supporting people to learn to cook healthy meals, and to use their skills often, for example using digital technology and social media to provide shopping list generators, food preparation teaching videos, and nutritional information. Regularity is particularly important, given that our findings indicated the greatest potential advantages from consuming home cooked meals were experienced at the highest frequencies of consumption. Infrequent home cooking, such as a weekly Sunday lunch, is unlikely to be of benefit to population health, and cooking habits should be adopted as part of the daily routine. This is in accordance with research suggesting that routinized home cooking behaviour is more likely to be maintained and prioritised over time [83].

The authors would like to thank all the volunteers who participated in the Fenland study, as well as the Fenland Study Coordination, Field Epidemiology, and Data Cleaning teams.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.