Characterization of the exploratory study population (FoCus-CS)

Differences in anthropometry, lifestyle, microbial diversity, and metabolic blood markers between 11 common NCDs and non-affected subjects were assessed using the German population-based FoCus cohort (FoCus-CS, N = 1220). Characterization of this study population, including disease prevalence of the assessed NCDs, is displayed in Table 1.

Exploratory analysis of metabolic phenotype and biomedical and lifestyle factors in association with common chronic non-communicable diseases

In an exploratory analysis, we assessed the association of biomedical and lifestyle factors (26 factors/markers in total) with the prevalence of 11 common NCDs using logistic regression models, utilizing data from the cross-sectional FoCus cohort (see Table 1). Unadjusted logistic regression analysis of each NCD is provided in the Additional file 1: Table S3 and Fig. S1. Due to initial differences in NCD prevalence with age, sex, and obesity, multiple logistic regression models shown here were adjusted for these three factors. Figure 2a displays relevant factors (p < 0.1) for each NCD. Displayed are forest plots of odds ratios (OR), 95% confidence intervals (CI), and p-values given by dot color, while exact ORs, CIs, and p-values are provided in the Additional file 1: Table S4. Since here was only a trend in triglyceride alteration (OR = 1.003 [0.999–1.006], p = 0.09) detectable with coronary artery disease but no other relevant association, this disease is not displayed in the figure.

Fig. 2

Exploratory analysis of phenotype and biomedical and lifestyle factors in association with common NCD. Presented are results of the exploratory analysis of 10 NCDs and respective age-at-disease-onset in relation to selected biomedical and lifestyle factors. a Multiple logistic regression models (disease presence/absence) adjusted for age, sex (obesity), and additionally obesity status (all other NCDs). Odds ratios (OR) estimated from factors with p < 0.1 (dot color indicating p-value). Due to very high ORs, associations with waist-to-hip ratio are not displayed. b Age-at-disease-onset for cases of the assessed NCDs was analyzed in association with the introduced factors using Spearman’s rank correlation tests. Presented is a heatmap of Spearman’s rank correlation coefficients with asterisks denoting significance levels (p < 0.01 = ***, p < 0.05 = **, p < 0.1 = *). Abbreviations : NCD, non-communicable disease; MDS, Mediterranean Diet Score; HEI-EPIC, Healthy Eating Index determined from EPIC-FFQ; HOMA-IR, Homeostatic Model Assessment of Insulin Resistance; CRP, C-reactive protein; IL, interleukin; IBD, inflammatory bowel disease; BMI, Body Mass Index; WHR, waist-to-hip ratio

Metabolic phenotype

As expected, this analysis shows that an altered metabolic phenotype and blood profile are linked to obesity, T2D and arterial hypertension. Increased waist circumferences (obesity: OR = 1.28 [1.24–1.32]; T2D: OR = 1.06 [1.03–1.09]; hypertension: OR = 1.05 [1.03–1.07]) and waist-to-hip-ratio (WHR, obesity: OR = 4.22 × 107 [2.67 × 107–6.69 × 108]; T2D: OR = 594 [21.74–16,265]; hypertension: OR = 230 [28.49–1864]) are found in people affected by at least one of these diseases. ORs associated with a one-unit-increase in WHR are very high, due to the small range of observed WHRs in the sample (Median = 0.87 [0.79–0.94]). In accordance, metabolic blood markers glucose, insulin, and HOMA-IR also exhibit increased values at metabolic disease presence (see Fig. 2a). While alterations in glucose markers are limited to metabolic diseases, alterations in anthropometry are present in IBD (WHR: OR = 18.18 [0.75–438]) and asthma (waist measure: OR = 1.03 [1.01–1.05]). Of note, psoriasis is the only disease displaying an inverse association with waist circumference (OR = 0.95 [0.92–0.98]).

Biomarkers of metabolic inflammation

To evaluate the level of metabolic inflammation, we determined triglyceride serum concentrations as a marker for metabolic disturbances and CRP levels as an inflammatory marker. Triglyceride levels are altered in some of the assessed NCDs displaying mildly increased odds ratios: obesity (OR = 1.01 [1.01–1.01]) and hyperlipidemia (OR = 1.01 [1.01–1.01]). Similarly, chronic inflammatory activity is present in obesity (OR = 1.05 [1.02–1.08]), hypertension (OR = 1.04 [1.01–1.07]), chronic heart failure (OR = 1.06 [1.02–1.1]), rheumatoid arthritis (OR = 1.06 [1.03–1.09]), and IBD (OR = 1.06 [1.03–1.09]).

Diet

Dietary differences towards reduced dietary quality, increased caloric quantity, and shift in macronutrient composition were detectable in metabolic diseases obesity (MDS: OR = 0.84 [0.76–0.93]; HEI-EPIC: OR = 0.98 [0.96–1], scaled energy intake: OR = 0.86 [0.85–0.89]; carbohydrate intake: OR = 1.03 [1,1.06]; protein intake: OR = 1.21 [1.12–1.3]; fat intake: OR = 0.96 [0.93–0.99]) and T2D (scaled energy intake: OR = 0.95 [0.92–0.98). A similar observation regarding caloric quantity can be seen in arterial hypertension (scaled energy intake: OR = 0.98 [0.97–0.99]). A uniform trend of increased dietary protein intake can be seen in metabolic diseases, IBD (OR = 1.13 [1.01–1.26]), and asthma (OR = 1.12 [1–1.25]). In return, a trend of lower protein intake (OR = 0.91 [0.81–1.02]), but higher carbohydrate intake (OR = 1.05 [1.01–1.09]) is present in rheumatoid arthritis. Evaluation of the MDS indicates higher adherence to a Mediterranean-style diet in people with hyperlipidemia (OR = 1.17 [1.07–1.28]). Dietary intake does not display potent associations of disease state with heart failure, coronary artery disease, psoriasis, asthma, and bronchitis.

Physical activity and TV consumption

For the evaluation of physical activity, everyday tasks representative for moderate and sports representative for vigorous activities were available. This analysis revealed no associations with everyday activities but decreased rates of vigorous sports activity in obesity (OR = 0.9 [0.86–0.94], hypertension (OR = 0.97 [0.94–1]), IBD (OR = 0.93 [0.87–0.99]), and chronic bronchitis (OR = 0.9 [0.81–0.99]). In addition to decreased sports activity, TV consumption was increased in obesity (OR = 1.02 [1.01–1.03]) but not hypertension, IBD, or bronchitis.

Sleep duration

Aside from diet and physical activity, it is now well known that the duration and quality of sleep is another important determinant for the preservation of health. Accordingly, altered sleep duration, which was available for our analysis, was present in obesity (OR = 0.9 [0.79–1.02]) but no other disease.

Smoking

Cigarette smoking is a risk for numerous of the observed NCDs, and while rates of smokers do not display significant differences among disease presence or absence, diseased subjects present higher cigarette consumption per day, e.g., in T2D (OR = 1.05 [1.03–1.07], IBD (OR = 1.02 [1–1.04]), and chronic bronchitis (OR = 1.04 [1.01–1.07]).

Alcohol

A higher alcohol consumption is displayed in people affected by arterial hypertension (OR = 1.04 [1.01–1.07]). Yet, most of the evaluated diseases display either no or even an opposite association with alcohol intake.

Microbial diversity

A decreased gut microbial diversity has been linked to disease etiology of various diseases and accordingly can be observed in several of the evaluated NCDs, like IBD (Shannon: OR = 0.25 [0.14–0.44]; Chao1: OR = 0.96 [0.94–0.98]), rheumatoid arthritis (Shannon: OR = 0.41 [0.24–0.71]), chronic heart failure (Shannon: 0.26 [0.1–0.66]), and asthma (Shannon: OR = 0.41 [0.24–0.71]) within our study. However, neither Shannon nor Chao1 indices display significant alternations in comparison to the respective control group for obesity, T2D, hypertension, hyperlipidemia, coronary artery disease, psoriasis, and chronic bronchitis.

The second part of this investigation involves a Spearman correlation analysis, examining the introduced biomedical and lifestyle factors in connection to age-at-disease-onset within disease cases. As presented in Fig. 2b, biomedical and lifestyle factors differently associate with the disease-onset-age for each NCD. Overall, this analysis reveals that while some of the assessed factors initially have not shown differences between disease presence or absence, they still demonstrate correlations with the age-at-disease onset, emphasizing their relevance in the disease course, onset, and prevention. For instance, this can be seen for type 2 diabetes, where the gut microbial diversity (represented through the Chao1 index) was no risk factor for the disease, but a significant association with age-at-disease-onset was found (rS =  − 0.355, p = 5.86 × 10−3). Our analysis further displays that in IBD, there is an association between reduced BMI as well as WHR and earlier age-at-disease-onset (BMI: rS = 0.292, p = 1.2 × 10−2; WHR: rS = 0.305, p = 8.67 × 10−3), while no differences in these anthropometric factors can be seen with disease presence/absence.

Characterization of disease-specific cohorts

Additionally, we performed an in-depth analysis of the age-at-disease-onset for two highly prevalent NCDs related to metabolic inflammation and alterations of the gut microbiota—type 2 diabetes and inflammatory bowel disease. T2D primarily involves metabolic dysfunction, while IBD pertains more to chronic inflammation. For both diseases, alterations of the gut microbiota are well established, in contrast to the multitude of other assessed NCDs, where such alterations could not be clearly outlined. In-depth analysis was performed using two disease-specific cohorts for (A) type 2 diabetes (FoCus-T2D, N = 514) and (B) inflammatory bowel disease (IBD-KC, N = 1110). These cohorts consist of disease cases and free-of-target-disease control subjects. The IBD-KC cohort is a family-based dataset, here used as a case–control study design, where controls were recruited as relatives from index IBD cases. A comprehensive characterization of study populations, including comparison between cases and control subjects is given in Table 2; information on disease-specific factors and microbial beta diversity can be found in the Additional file 1: Tables S4A + B and Fig. S2, respectively.

Association between biomedical and lifestyle factors and age-at-disease-onset in T2D and IBD

Each biomedical lifestyle factor (independent variable) was investigated using individual Cox proportional hazards regression, adjusting for sex and BMI class (for details see “Statistical analysis”), with age-at-disease-onset (for cases) or censored age-at-examination (for controls) as dependent variables and disease status as censoring indicator. Continuous biomedical and lifestyle factors were categorized into either terciles or deciles. Presented are biomedical and lifestyle factors and disease specific markers, relevant for T2D (Fig. 3a) and IBD (Fig. 3b). The Kaplan Meier survival plots depict the probability of disease occurrence in relation to age, with each evaluated factor considered individually. In both conditions, interesting associations (p-value < 0.1 when comparing the full model with the reduced model) emerged for various lifestyle factors related to nutrition, gut microbiome, and anthropometrics. Hazard ratios and confidence intervals for those lifestyle factors can be found in the Additional file 1: Tables S6A + B. It should be noted that the data used are cross-sectional, with both disease status and biomedical and lifestyle factors collected at baseline. That is why only hazard ratios (reference group is “low” for all factors), but no absolute age-dependent prevalences can be estimated and no statements about causality can be made, interpretation will focus on correlation and association.

Fig. 3

a Kaplan-Meier survival plots for biomedical and lifestyle factors and type 2 diabetes risk. This figure displays Kaplan-Meier survival plots for biomedical and lifestyle factors (a) HEI-EPIC, (b) MDS, (c) Smoker, and (d) Triglycerides in relation to the risk of T2D, assessed based on age-at-disease-onset (diabetes cases) and censored age-at-examination (control subjects). Color-coded levels within each parameter signify various groups or levels. Asterisks and color denote significance levels (pCox< 0.01 = ***, pCox< 0.05 = **, pCox < 0.1 = *), additionally the total hazard ratio (HR) between the lowest and the highest level for each biomedical or lifestyle factor is given from the Cox regression models. Abbreviations: HEI-EPIC, Healthy Eating Index adapted to EPIC Food Frequency Questionnaire data; MDS, Mediterranean Diet Score. b Kaplan-Meier survival plots for biomedical and lifestyle factors and IBD risk. This figure displays Kaplan-Meier survival plots for biomedical and lifestyle factors (a) HEI-EPIC, (b) MDS, (c) Alcohol, (d) Smoker UC, (e) Smoker CD, (f) Calprotectin, (g) Chao1 index, (h) Shannon index, and (i) Bristol stool scale in relation to the risk of IBD, assessed based on age-at-disease-onset (IBD cases) and censored age-at-examination (control subjects). Color-coded levels within each parameter signify various groups or levels. Asterisks and color denote significance levels (pCox< 0.01 = ***, pCox< 0.05 = **, pCox < 0.1 = *), additionally the total hazard ratio (HR) between the lowest and the highest level for each biomedical or lifestyle factor is given from the Cox regression models. Abbreviations: MDS, modified Mediterranean Diet Score; UC, ulcerative Colitis; CD, Crohn’s disease; calp: calprotectin

For T2D, the association between age-at-disease-onset and BMI has been evaluated in a simple model only adjusted for the participants’ sex. This displays very low risk among under- and normal weight subjects and monotonously earlier onset from overweight to obesity III. Due to that association, further evaluation of biomedical and lifestyle factors was performed not only adjusting for sex but also for BMI class. Associations with p-values < 0.1 can be found for the diet quality assessed by the HEI-EPIC and MDS, current smoking status and triglyceride serum levels, while physical activity, sleep duration, microbial diversity, inflammatory, and glucose metabolism markers could not directly be linked to age-at-diabetes-onset. In more detail, the HEI-EPIC is a measure for an overall quality diet. People with high diet quality, indicated by the HEI-EPIC level, show a delay in T2D onset indicated by a hazard ratio of 0.64 [0.396–1.034] (low-moderate) and 0.453 [0.284–0.722] (low–high) (see Fig. 3a (a)). The MDS, however, does not show such clear association with only slightly later T2D onset in people with moderate (HR = 0.947 [0.669–1.340]), but not high MDS values (HR = 1.495 [1.025–2.180]) within our study population (see Fig. 3a (b)). In this context, it should be considered that dietary recommendations for individuals with diabetes closely resemble a Mediterranean diet [31]. Consequently, the assessment of the Mediterranean Diet Score (MDS) may be susceptible to bias, a problem of the case–control study design, where no information on dietary habits before disease onset is available. Consistently, within our study population, HEI-EPIC levels have a similar distribution in diabetes cases and controls, while a larger proportion of cases (28.93%) were assigned to the high level of MDS compared to controls (18.93%). Cigarette smoking is also associated with the probability of developing diabetes in this population. Smokers experience an approximately 4 years earlier median age-at-disease-onset (see Fig. 3a (c)). Finally, as seen in Fig. 3a (d), elevated blood triglyceride levels may also be of interest when considering earlier diabetes onset (low-moderate HR = 1.166 [0.579–2.349], low–high HR = 1.249 [0.647–2.412], p = 0.05).

For IBD, factors associated with age-at-disease-onset with p-values < 0.1 include the diet quality (HEI-EPIC and MDS), alcohol consumption, current smoking status, microbial diversity markers (Shannon and Chao1 index), as well as disease-specific markers like calprotectin and the Bristol stool scale. Again, the case–control study design does not allow estimation of age-specific prevalences, and interpretation is done with a focus on relative disease risk. Concerning diet, individuals with a high HEI-EPIC score experience a median age-at-disease-onset 8 years later than those with a low diet score (see Fig. 3b (a)). Additionally, the MDS score demonstrates an even stronger delay in disease onset by 16 years (low–high HR = 0.691 [0.567–0.841]) as illustrated in Fig. 3b (b). This positive correlation between a healthy diet and a lower likelihood of developing IBD at any age is supported by a consistent decrease in the hazard ratio with increasing diet scores. A less intuitive result pertains to the association between age-at-disease-onset and alcohol intake (Fig. 3b (c)). Individuals in this study with a higher alcohol intake show a decreased risk of developing IBD with an HR of 0.488 [0.397–0.599]. No significant associations were found for the impact of physical activity (sports, watching TV, and everyday activity), sleep, and total or scaled energy intake and age-at-disease-onset. Another interesting finding (Fig. 3b (d + e)) is that for UC, smoking correlates with a delayed age-at-disease-onset (HR = 0.467 [0.277–0.785]). However, for CD, there is no significant association with current smoking habits (HR = 0.836 [0.604–1.158]). As anticipated, participants with greater gut microbiota diversity experience a delayed age-at-disease-onset compared to those with lower diversity (Fig. 3b (g + h)). The hazard ratio for both the Chao1 and the Shannon index is consistently below one, indicating a significantly decreased risk of developing IBD associated with a higher gut diversity in this study population. As depicted in Fig. 3b (i), individuals from the analyzed families with diarrhea-like stool types have a higher IBD hazard (1 < HR < 2.9) compared to those with normal stool types. Another disease-specific marker, calprotectin (Fig. 3b (f)), indicates that individuals from this cohort with high calprotectin values are more likely to develop IBD (low–high HR = 2.21 [1.819–2.691]).

When examining the Kaplan–Meier plots for T2D (Fig. 3a) and IBD (Fig. 3b), it is noteworthy that the curves for diabetes exhibit a linear increase in disease prevalence with age. In contrast, the relationship is more accurately characterized by an exponential function for IBD. This signifies that hardly any new-onset IBD cases were observed in the older age groups in this study, a pattern not observed in diabetes.

Role of biomedical and lifestyle factors-gut microbiota interactions in diabetes and IBD

Recognizing inconsistencies in previously described connections between certain biomedical and lifestyle factors and the composition of the human gut microbiota, we aim to delve deeper into understanding potential differences between diseased (T2D and IBD) and healthy individuals regarding the impact of biomedical and lifestyle factors on the abundance of genera in the gut microbiome. We use 16S abundance data at genus level and the previously described pre-processing steps (see “Statistical analysis”) for the IBD-KC and FoCus-T2D cohorts.

Our results, presented in Tables 3 and 4, highlight significant associations (FDR-corrected p-value < 0.05) between each biomedical or lifestyle factor and genus. To gain a comprehensive understanding, we initially examined cases and controls together to identify interacting factors. Subsequently, based on the differences found in microbial diversity between T2D and IBD patients and their respective control groups (see the Additional file 1:Fig. S2), we conducted separate analyses for cases and controls, replacing the interaction term with the biomedical or lifestyle factor alone. Distinct patterns emerged in both cohorts; specific genera exhibited significant associations with biomedical and lifestyle factors for cases but not controls.

Table 3 Interaction analysis of selected biomedical and lifestyle factors with T2D on the abundance of gut bacterial generaTable 4 Interaction analysis of selected biomedical and lifestyle factors with IBD on the abundance of gut bacterial genera

The effects of these biomedical and lifestyle factors in the same genera were either smaller or even in the opposite direction in unaffected controls, as indicated by significant interactions. However, in the stratified analysis of the control groups alone, none of these were statistically significant.

In the diabetes cohort, current smokers exhibited a higher abundance of the genus Bilophila (Coefcases = 1.31). Additionally, T2D patients with a higher BMI demonstrated reduced abundances of Catabacter (Coefcases =  − 0.65). Furthermore, higher values of both diet scores were associated with an increased abundance of the genus Mitsuokella in the group of cases (HEI: Coefcases = 1.15, MDS: Coefcases = 1.21). This underscores the importance of diet and its impactful influence on the microbiome.

For the IBD-KC cohort, the following significant associations were found: IBD patients who are current smokers or exhibited a propensity for diarrhea-like stool types showed decreased abundances of Oscillibacter (Coefcases: Smoker status =  − 0.21; Bristol stool scale =  − 0.25). Additionally, female IBD patients show a higher abundance of Oscillibacter (Coefcases = 0.17). Those with a higher frequency of diarrhea-like stool types additionally demonstrated reduced abundances of Clostridium IV (Coefcases =  − 0.35) and Subdoligranulum (Coefcases =  − 0.21). In return, current smokers further exhibited lower abundances of Intestimonas (Coefcases =  − 0.37) and Catabacter (Coefcases =  − 0.44). Lastly, higher calprotectin rates in IBD patients were linked to an increased abundance of Flavonifractor (Coefcases = 0.18) and Lactobacillus (Coefcases = 0.44).