Introduction

Obesity is one of ten risk factors for cardiovascular disease, and the incidence of obesity has increased to epidemic proportions1,2. Obesity increases the risk of other non-communicable diseases, such as heart disease, type 2 diabetes, fatty liver, high cholesterol, osteoarthritis, sleep apnoea and cancer3,4,5,6. More than 4 million people died in 2017 because of overweight or obesity7. Therefore, the problem of obesity cannot be ignored. More than one-third of the prevalent cases have a digestive aetiology8, and the number of associated deaths tends to increase between the ages of 45 and 799. Helicobacter pylori is the pathogen that causes most peptic ulcers and gastric cancer10. Intestinal diseases can trigger water and electrolyte imbalances and disorders of acid‒base balance11. Non-alcoholic fatty liver is an essential cause of liver damage, which can lead to cirrhosis and liver cancer, lower the body’s immunity and damage the digestive system5.

Obesity is closely associated with digestive disorders, not only causing an increased burden of associated digestive disorders, such as gastrointestinal and hepatic diseases12, but also contributing to the poor prognosis and low survival rates of patients with obesity-associated cancers13. Previous studies have mostly used body mass index (BMI) or waist circumference (WC) to define obesity; BMI is the most conventional indicator of obesity, and it can reflect subcutaneous fat but not visceral fat14. Waist circumference is an important indicator of obesity, and it reflects subcutaneous and visceral fat14. Several studies15,16 have demonstrated that WC can better represent the parameters of abdominal obesity and can be a better predictor of health risks associated with obesity. Lack of studies combining BMI and WC to investigate the obesity-mortality relationship in digestive disease population. A single measure of obesity may have limitations in exploring the effects of obesity on health and mortality. It is important to identify the negative effects of one or more specific types of obesity. Therefore, this study aims to investigate the effect of obesity on mortality in people with digestive diseases by considering both BMI and WC.

Materials and methods

Population and study design

Our study is a longitudinal analysis based on the UK (United Kingdom) Biobank (Application Title: Integration of clinical data and genomic data to construct diagnosis and prognosis system for digestive diseases and related complications, Application ID: 84347). The UK Biobank is the world’s most detailed, long-term prospective health study. Between 2006 and 2010, 22 centres in Scotland, England, and Wales invited people aged 40–69 years who were living in the UK to visit their nearest assessment centre by mail inquiry and telephone. Trained professionals collected baseline information, body measurements, and biospecimens. Of the 9.23 million people who were invited to join the UK Biobank, 503,317 (5.45%) consented and were recruited17. The UK Biobank assessment process had five components: written consent; touch screen questionnaires; face‒to-face interviews; measurements; and blood, urine, and saliva sample collection. All the participants approved this UK Biobank study and provided written informed consent to participate in the study. Further details of these measurements, study design, and data collection are available in the UK Biobank online protocol and study protocol (http://www.ukbiobank.ac.uk). We obtained data from 502,411 participants from the UK Biobank when our application was approved.

Participants are included in the study if they have been diagnosed with any digestive disease, defined by International Classification of Diseases, Tenth Revision (ICD-10) codes K00 to K93. The detailed process for screening the study subjects is shown in Fig. 1.

Fig. 1
figure 1

Flow chart of population selection and analysis for the study.

Full size image

Outcomes

The final follow-up status was divided into death and censored groups. The time to death, time lost to follow-up, and time of latest data collection (at the time of analysis, mortality data as of February 2022 were collected) were defined as the end time. The period from recruitment to the endpoint was defined as the survival time.

Ascertainment of covariates

Body mass index (BMI) was calculated as weight (kg)/height (m)2. Weight was measured via a Tanita BC-418 MA body composition analyser, which is accurate to 0.1 kg, and height was measured with a Seca 202 height measure. The participants were required to remove their shoes and heavy clothing while the measurements were taken. Waist circumference (at the level of the umbilicus) was be measured with a Wessex nonstretchable sprung tape measure and entered manually by staff. Trained staff carried out these measurements18.

We categorized the population into four groups based on the use of BMI19 and WC20 to define the type of obesity, namely “General obesity” (BMI ≥ 30 kg/m2 & WC < 102 cm in males and WC < 88 cm in females), “Abdominal obesity” (BMI < 30 kg/m2 & WC ≥ 102 cm in males or WC ≥ 88 cm in females), “Combined obesity” (BMI ≥ 30 kg/m2 & WC ≥ 102 cm in males or WC ≥ 88 cm in females) and “Non-obese”. A sleep duration of 7–8 h was considered good21. Otherwise, sleep was considered poor. Behaviours of watching TV or using a computer were recorded as screen time, which we divided into “<2 hours” and “≥2 hours”.

The Townsend Deprivation Index (TDI) 22, which combines information on housing, employment, and car availability, was calculated on the basis of census data and postcode before participant recruitment. The index was used mainly to measure socio-economic status, with higher values indicating greater deprivation. In addition, we selected sex, age, income, smoking, alcohol, sleep, screen time, and the International Physical Activity Questionnaire (IPAQ) score23 as covariates for the study.

Statistical analysis

In the primary analysis, all 254,445 participants were included in the study. Responses of “Preferred not to answer”, “uncertain/unknown”, or “invalid” were recoded as missing or null. Missing values were interpolated via multivariate imputation via the chained equations method with the R software “mice” package (with 5 imputed datasets, 10 iterations, and the random forest method). The population was divided into death and censored groups according to the final follow-up status. Continuous variables with a normal distribution are presented as the mean ± standard deviation (\(\bar{\text{x}}\) ± sd) and Student’s t test was used for comparisons between groups; continuous variables with a nonnormal distribution are presented as the median and quartiles [M (Q1, Q3)] and the Mann‒Whitney U test was used for comparisons between groups; categorical variables are presented as frequencies and percentages, and the χ2 test was used for the analysis of differences in distribution.

We used Kaplan‒Meier curves to estimate survival. The Cox proportional risk model and marginal structural Cox regression model were used to assess covariates that may affect mortality in people with digestive diseases. We estimate inverse probability weights by adjusting for the confounding effects of sex, age, TDI, income, smoking, alcohol consumption, sleep, screen time, and IPAQ on obesity. This weight is used to fit a marginal structural Cox regression model. In addition, we construct four different Cox regression models and two different marginal structural Cox regression models by adjusting for different variables, which were used to observe whether the effect of obesity on mortality from digestive diseases is robust.

The model’s differentiation is represented by Harrell’s consistency index (C-index), which reflects the ability of the model to distinguish between outcome events correctly. When the C-index is close to 1, the model is considered to have intense discrimination. Calibration curves were used to measure the agreement between the predicted and observed probabilities of digestive system disease mortality. The 45° diagonal line indicates that the predicted probability equals the actual probability. The closer the calibration plot is to the ideal straight line, the better the calibration of the prediction model is.

For a sensitivity analysis of the preliminary study, we repeated the analyses, excluding cases with missing values, to compare whether there was a significant change in the primary outcome. In this study, R software version 4.1.2 was used for all the data analyses.

Results

Characteristics and mortality

A total of 254,445 participants were included in the analysis, with a mean age of 57.8 ± 7.8 years and a median follow-up time of 154 months. Of these, 227,111 (89.3%) participants were censored, and 27,334 (10.7%) participants died. The characteristics of the participants with different outcomes are presented in Table 1. Demographically, the death group had higher proportions of individuals with combined obesity, individuals who were males, individuals who were most deprived, and individuals with low income. In terms of behaviour, the death group had higher proportions of smokers (both previous and current smokers) and those who had previously consumed alcohol. In addition, the death group had lower proportions of individuals with good sleep, low screen time, and high physical activity than did the censored group (Table 1).

Table 1 Characteristics of patients with digestive system diseases.
Full size table

Survival analysis

Figure 2 shows that the combined and abdominal obesity survival function decreased more than the other two obesity types according to the conventional Kaplan‒Meier curve. However, in the Kaplan‒Meier curve based on inverse probability weights, the survival function for combined obesity alone was noticeably lower than that of the other three groups.

The results of univariate Cox regression revealed that all variables possibly influence mortality in people with digestive diseases (Table 2). In addition, we constructed several models to explore the effect of obesity on mortality by adjusting for different variables (Table 3, Models 2–4). It could be seen that abdominal obesity and combined obesity are risk factors for mortality in the population with digestive disorders. No effect of general obesity on mortality was observed.

Table 2 Univariate cox regression results in people with digestive system diseases.
Full size table
Table 3 Outcome of the effect of obesity on mortality in people with digestive system diseases.
Full size table
Fig. 2
figure 2

Survival curves.

Full size image

We constructed a marginal structural Cox model by inverse probability weighting (Table 3, Model 5). The model found that only combined obesity is a risk factor for mortality in people with digestive diseases [HR = 1.190, 95% CI= (1.156, 1.225)] and that abdominal obesity [HR = 1.020, 95% CI= (0.984, 1.057)] was not a statistically significant predictor. We further adjusted for other covariates on the basis of inverse probability weighting (Table 3, Model 6). The results showed that abdominal obesity [HR = 1.102, 95% CI= (1.063, 1.142)] and combined obesity [HR = 1.236, 95% CI= (1.201, 1.272)] were predictors of mortality in people with digestive disease and that the combined obesity HR was greater. This result is similar to that of Model 4.

Models 4 and 6, which include all the significant covariates, also provide additional information (Fig. 3). The results of these two models are consistent. In Model 6, male sex [HR = 1.638, 95% CI= (1.596, 1.618)] and increasing age [HR = 1.084, 95% CI= (1.082, 1.087)] were risk factors for mortality in people with digestive diseases. Higher levels of deprivation increased the risk of death compared to the lowest deprivation. Income is an independent predictor of mortality in the digestive disease population. The risk of death decreases with increasing income. Both prior [HR = 1.259, 95% CI= (1.224, 1.295)] and current [HR = 12.191, 95% CI= (2.113, 2.271)] smoking habits are risk factors for mortality in people with digestive diseases compared with never smokers. Previous alcohol use [HR = 1.289, 95% CI= (1.198, 1.388)] is a mortality risk factor, but current alcohol use [HR = 0.873, 95% CI= (0.824, 0.925)] is a protective factor compared with never having consumed alcohol. In addition, good sleep [HR = 0.885, 95% CI= (0.863,0.907)], no more than 2 h of screen time [HR = 0.913, 95% CI= (0.887,0.938)], moderate [HR = 0.811, 95% CI= (0.785,0.838)] and high [HR = 0.730, 95% CI= (0.707,0.755)] physical activity are all protective factors for mortality in the population with digestive diseases and reduce the risk of death in people with digestive diseases.

Fig. 3
figure 3

Forest plot for regression analysis. HR hazard ratio, CI confidence interval, TDI Townsend deprivation index, IPAQ International Physical Activity Questionnaire.

Full size image

A total of 171,715 participants (mean age = 57.4 ± 7.9 years) were included in the sensitivity analysis after participants with missing values were removed. Of these, 154,471 (90.0%) participants were censored, and 17,244 (10.0%) participants died. Figure S1 shows the content of the predicted survival probabilities via the nomogram. The sensitivity analysis results are consistent with the main analysis results, which indicates that our results are robust. (Supplementary files Tables S1, S2, Figs. S2, S3).

Validation

We evaluated the models’ discrimination ability via the C-index (Table 3). Except for the univariate regression analysis models (Models 1 and 5), the C-indices were all close to 0.7. The C indices for Models 4 and 6 were 0.712 [95% CI = (0.709,0.715)] and 0.711 [95% CI = (0.708,0.714)], respectively, indicating that the multivariate regression models could distinguish well between death and censoring in the population with digestive disorders. The calibration plots for 3-, 5- and 10-year mortality (Fig. 4) were near the 45° ideal line. These findings suggest that the predicted mortality for digestive system diseases is mainly consistent with the observed mortality.

Fig. 4
figure 4

Calibration curves for regression analysis.

Full size image

Discussion

This is an analysis of a large prospective cohort study based on the UK biobank. We used the marginal structural Cox model with inverse probability weighting to investigate the role of different types of obesity on death due to digestive diseases. The following are several key findings that were found in people with digestive diseases:

  1. (1)

    Abdominal obesity and combined obesity are important factors for increased mortality.

  2. (2)

    Higher social status, i.e., higher income and lower levels of deprivation, reduces the risk of death.

  3. (3)

    Physical activity and less screen time are independent, modifiable determinants of digestive disease outcomes.

Obesity and digestive system diseases mortality

Most studies have shown that obesity decreases the life expectancy of individuals24,25. Our study further revealed that abdominal obesity and combined obesity are independent predictors of mortality in people with digestive diseases. Notably, unlike some previous studies26, our study does not show the results of the obesity paradox.

Obesity could have an impact on the survival of patients with many cancers of the digestive tract. It has been suggested that waist circumference is a better predictive risk factor for mortality and morbidity after colorectal surgery than BMI is27. A European Prospective Investigation into Cancer and Nutrition (EPIC) cohort study28 found that general obesity and abdominal obesity significantly increased the risk of all-cause mortality in people with colorectal cancer (CRC). This result is partially consistent with our study, which only found that combined obesity affects death to a greater extent than abdominal obesity, but this does not completely negate the predictive value of BMI for outcomes such as morbidity and mortality in some digestive diseases29. BMI-based obesity does not assess the distribution of fat mass, nor does it distinguish whether fat mass or muscle mass is responsible for obesity30,31; this means that people who are generally obese do not necessarily have a poor health status32, which may explain the lack of statistical significance of the effects of general obesity. Obesity is heterogeneous and not all types of obesity are equally pathogenic33. BMI and WC reflect the characteristics and degree of obesity of individuals in different ways. Therefore, we believe that it is valuable and meaningful to consider the type of obesity by integrating BMI and WC. Research has also suggested that a higher waist‒to-hip ratio and waist circumference are associated with an increased risk of death from pancreatic cancer 34. Obesity may be independently associated with the risk of hepatocellular cancer-related mortality35. Obesity not only increases mortality in digestive cancers but also worsens the prognosis of patients with other digestive diseases. Kim et al. reported that obesity increased the risk of liver disease-related death36. An American central adiposity cohort study concluded that increased WC was associated with increased liver disease mortality37. These findings further demonstrated the harmful effects of abdominal obesity on digestive disorders.

Obesity can affect the body in several ways38. For example, increasing abdominal pressure can increase the risk of gastroesophageal reflux disease (GERD) 39. It can also cause cancer 40 and inflammatory responses41 through the release of protumor factors and proinflammatory cytokines from visceral fat. The metabolic mechanisms may differ in populations with different obesity characteristics42,43. These differences in metabolic mechanisms may be associated with the impaired distribution of adipose tissue and the promotion of adipocyte differentiation, leading to an increased risk of death44. These findings could explain to some extent the increased mortality from obesity.

Others and digestive system diseases mortality

In the multivariate model, we were also able to identify relationships between covariates and mortality.

Sex and socioeconomic status characteristics

Among people with digestive diseases, males have a greater risk of death than females do. A Norwegian study reported similar sex differences45, with different curves for a body shape index (ABSI, a measure of obesity) and all-cause mortality between males and females. Males typically have more visceral adipose tissue than females do 46, although BMI and even WC values may not differ. This may account for the difference in the association between obesity and mortality from digestive diseases in females and males.

The TDI is calculated by combining various factors, including employment and housing, and to some extent, indicates people’s behavioural patterns47. We used TDI and income to represent socioeconomic status and found that high levels of deprivation and low income are risk factors for increased mortality in people with digestive diseases. The negative impact of low socioeconomic status on health and mortality is well established48,49, and our results provide new support. The association of low socioeconomic status with higher mortality is complex. Low socioeconomic status may cause changes in one or more risk factors that lead to poor health status. Poor health status further reduces the labour supply and income50, thus leading to a vicious cycle.

Lifestyle habits and physical activity

Consistent with previous studies51,52,53, our study suggested that smoking, poor sleep, high screen time, and low physical activity are detrimental to health and survival. Our study found that previous alcohol consumption was a protective factor for mortality in people with digestive diseases, whereas current alcohol consumption was a risk factor. These results are not the first findings.

A cardiovascular disease study noted that the association between alcohol consumption and cardiovascular disease incidence is a “J” curve54, suggesting that the effect of alcohol consumption on health and survival is not a one-way dose‒response relationship. Our previous studies also illustrated this point53. Individuals with a history of prior alcohol consumption may abstain from alcohol because of poor physical condition55,56, and the poor physical condition often predicts a low survival time, which in turn shows that previous alcohol consumption is a risk factor for death. In addition, genetic factors and drinking patterns may also influence the metabolism of alcohol. We should therefore be cautious about interpreting the effects of alcohol on survival.

Limitations

However, our study also has several limitations that should be considered when these results are interpreted.

  1. (1)

    There are sampling errors, recall biases, reporting biases, and nonresponse biases in the data from the sample and questionnaires;

  2. (2)

    The survey was conducted in the UK and may not be generalizable to the world population;

  3. (3)

    The categorisation of current/former/never smokers may make it difficult to capture the confounding effects of smoking; and.

  4. (4)

    Analysing digestive diseases as a whole may overlook some potentially meaningful associations, and it is valuable to consider disease stage or severity for more in-depth analyses.

Conclusion

Overall, combined obesity, as determined by BMI and WC, is an important factor that influences mortality in the digestive disease population, with a greater role than abdominal obesity alone. Our study also revealed several demographic and behavioural factors that may be associated with mortality in the digestive disease population. These modifiable risk factors for mortality can provide guidance to patients with digestive diseases to avoid poor lifestyles and prolong survival time. Decreased quality of life and reduced survival time are not effects of one factor alone, but rather, factors may interact; therefore, comprehensive, rational, and robust interventions are important for reducing mortality.