Diet as a hot topic in psychiatry: a population‐scale study of nutritional intake and inflammatory potential in severe mental illness

People with severe mental illnesses (SMIs) – including schizophrenia, major depressive disorder (MDD) and bipolar disorder – have excessive caloric intake, a low‐quality diet, and poor nutritional status compared to the general population1, 2. Poor diet increases the risk of diabetes and cardiovascular mortality in this population3. Furthermore, excessive consumption of high‐fat and high‐sugar foods can increase systemic inflammation4. Indeed, all classes of SMI show heightened levels of peripheral inflammatory markers, which is linked to worse prognosis in these conditions. However, there currently is an absence of large‐scale studies comparing the nutritional intake and inflammatory profile of the diets of individuals with SMIs.

To address this, we used detailed dietary intake data from the baseline phase (2007‐2010) of the UK Biobank study5 to examine differences in nutritional intake and diet‐associated inflammation between people with SMIs and the general population. Full details on the UK Biobank, including approval from the National Health Service (NHS) Research Ethics Committee, are available elsewhere5. We used patient hospital records to identify individuals with a ICD‐10 diagnostic history of recurrent depressive disorder, bipolar disorder (type I or II) or schizophrenia. Additionally, participants' answers to questions from the Structured Clinical Interview for DSM‐IV Axis I Disorders (SCID‐I) and Patient Health Questionnaire (PHQ), administered at the UK Biobank baseline, were used to identify additional individuals with MDD or bipolar disorder6. Participants who fell into multiple psychiatric categories were assigned hierarchically to only one, in this order: schizophrenia, bipolar disorder, MDD. Healthy controls were derived from all UK Biobank participants who had no indication of any previous or present psychotic or mood disorder. Individuals with neurological conditions known to affect memory recall were excluded from all groups.

Diet was assessed using a validated computerized questionnaire, the Oxford WebQ7. This brief, self‐administered 24‐hour recall measure queries previous day intake for >200 food items. To capture intra‐individual (day‐to‐day) variation in food intake, the WebQ was administered on up to five separate occasions over a 16‐month period. Missing individual data points were imputed as the mean average across all available time points from that individual. The first assessment was administered on‐site at the UK Biobank assessment centre; all subsequent administrations were completed online.

A total of 69,843 eligible UK Biobank subjects (mean age 56.5 years, 46.4% male) provided sufficient data for analyses. Of these, 14,619 individuals had MDD, 952 had bipolar disorder, 262 had schizophrenia, and 54,010 were SMI‐free. Multivariable linear regression was used to examine differences in total daily energy intake and each macronutrient between the SMI and control samples. Analyses were adjusted for gender and age.

The greatest differences in diet were observed for the schizophrenia sample. Age‐ and gender‐adjusted comparisons to control subjects showed highly elevated intakes in that sample (all p<0.001) for total energy (+553.4 kilojoules (kj)/day, SE=143.8), carbohydrates (+25.4 g/day, SE=4.86), sugar (+16.0 g/day, SE=2.98), total fat (+6.04 g/day, SE=1.77), saturated fat (+3.76 g/day, SE=0.76) and protein (+5.24 g/day, SE=1.51), with no difference in dietary fibre (p=0.78).

Individuals with bipolar disorder similarly showed significantly (all p≤0.01) higher intake of total energy (+298.5 kj/day, SE=75.9), carbohydrates (+11.4 g/day, SE=2.57), protein (+1.97 g/day, SE=0.80), sugar (+9.63 g/day, SE=1.57), total fat (+2.40 g/day SE=0.93) and saturated fat (+1.29 g/day, SE=0.40) compared to controls, with no difference in fibre (p=0.32).

The MDD sample also showed significantly greater (all p<0.001) age‐ and gender‐adjusted intake in comparison to controls, for total energy (+189.4 kj/day, SE=21.88), carbohydrates (+5.15 g/day, SE=0.74), sugar (+3.11 g/day, SE=0.45), total fat (+2.19 g/day, SE=0.27), saturated fat (+0.96 g/day, SE=0.12) and protein (+1.12 g/day, SE=0.23), along with a small difference in dietary fibre intake (+0.15 g/day, SE=0.06, p=0.01).

Sensitivity analyses were conducted to adjust for ethnicity, body mass index (BMI), education and social deprivation. This did not substantially alter the overall findings.

Beyond examining raw macronutrient intakes, we also explored the inflammatory potential of the diet using the dietary inflammatory index (DII®)4, which has produced consistent positive associations in over ten studies using inflammatory markers including C‐reactive protein, interleukin‐6 and tumor necrosis factor‐α as construct validators8.

A total of 68,879 participants had provided sufficient dietary intake data across the 18 macro/micronutrient intake parameters relevant for DII® calculation. Multiple linear regression was used to examine if SMI was positively associated with DII scores, adjusting for age, gender, ethnicity, BMI, social deprivation, education and total energy intake. DII scores were significantly elevated in subjects with schizophrenia (B=0.220, SE=0.084, p=0.009) and MDD (B=0.031, SE=0.013, p=0.014), but not with bipolar disorder (p=0.27), compared to controls.

Overall, this population‐scale analysis of nutritional intake confirms that people with SMIs have higher intakes of obesogenic nutrients and more inflammatory diets than the general population. Whereas dietary interventions for SMIs often focus exclusively on over‐consumption of obesogenic, pro‐inflammatory foods, this study shows that further consideration should be given to increasing consumption of nutrient‐dense foods that are known to reduce systemic inflammation4, 8.

In terms of both total caloric intake and excess obesogenic nutrients, the worst dietary patterns were observed among people with schizophrenia. This is a notable finding, as these individuals also have significantly higher rates of metabolic disorders and greater premature mortality than individuals with other classes of SMI3, 9, indicating that diet could be a key factor influencing these outcomes.

Indeed, dynamic weight change algorithms predict that each 100 kj of excess energy intake per day will eventually lead to at least 1 kg increase in body weight10. Thus, the 553 kj (132 calories) per day excess observed in the schizophrenia sample suggests that dietary differences alone can account for 5‐6 kg of the increased body weight observed in this population. Not only does excess caloric, carbohydrate and fat intake increase inflammation, but the concomitant increase in adipose tissue also enhances chronic, systemic inflammation.

The degree to which the heightened systemic inflammation observed in SMIs is attributable to dietary factors needs to be clarified. Sufficiently sized cohort studies, using detailed dietary and psychiatric data alongside biomarkers of inflammation, can provide new insights into of the role of diet in SMIs. Future work should also aim to establish the extent to which heightened dietary inflammation in SMIs independently contributes to the poor physical, psychological and neurocognitive outcomes observed in these populations, which represent a significant public health challenge.

This research was conducted using the UK Biobank Resource (application no. 22125). J. Firth is supported by a Blackmores Institute fellowship and Medical Research Council (MRC) doctoral training grant (P117413F07). B. Stubbs is in part funded by the UK National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London. J. Sarris is supported by a research fellowship from the National Health and Medical Research Council (NHMRC), Australia (APP1125000). M. Berk is supported by an NHMRC senior principal research fellowship (1059660). The views expressed are those of the authors and not necessarily those of the above entities.