Introduction

Posttraumatic stress disorder (PTSD) [1], a severe sequela of traumatic experiences, characterized by symptoms such as flashbacks; nightmares; avoidance; a sense of “numbness” and emotional blunting; anhedonia and acute burst of fear or panic, frames a significant part of the population. Lifetime prevalence of PTSD is estimated to be 3.9%, with a variation of 2.1% in lower to low middle income countries and 5.0% in high-income countries [2]. PTSD, in turn, is associated with poorer physical health and health related quality of life [3] and lower life expectancy [4].

Although several systematic reviews and meta-analyses during the past years indicate increased risk for developing cardiometabolic diseases, e.g., cardiovascular diseases (CVD), cardiovascular risk factors such as type 2 diabetes mellitus (T2DM), and metabolic syndrome (MetS) in people with PTSD [5,6,7], a causal relationship has been difficult to establish [8]. The pathogenesis of cardiometabolic diseases in people with PTSD is complex, consisting of biological, psychosocial and behavioural factors (e.g., eating behavior, physical inactivity and smoking) [9]. For example, traumatic stress might impact eating behavior [10], that over time can cause adverse health consequences. Koenen, et al. [8] suggest at least five different causal structures explaining the statistical association between PTSD and cardiometabolic diseases; 1) a confounding bias explaining the relationship (e.g., trauma exposure and socioeconomic status (SES)); 2) a reverse causal association, i.e., myocardial infarction can lead to PTSD; 3) a one-way link between PTSD and cardiometabolic disease, i.e., PTSD causes cardiometabolic diseases; 4) a mediating factor explaining the association, e.g. smoking; or 5) a two-way association, i.e., scenario 2 and 3. However, although the majority of previous studies have found an association between PTSD and an increased risk of cardiometabolic diseases, Crum-Cianflone, et al. [11] found that PTSD symptoms in United States military personnel was not associated with coronary heart diseases after adjustment for depression and anxiety.

People with PTSD are often diagnosed with concurrent alcohol use disorder (AUD) [12], depressive disorders [13] and anxiety [14]. Estimated prevalence for co-occurring AUD range between 13% in Denmark to 55% in USA [14,15,16], while for co-occurring major depressive disorder and anxiety, the prevalence is estimated to be 52% and 49%, respectively [13, 14]. Both anxiety, AUD, and depressive disorders are in turn associated with increased risk of cardiometabolic diseases [17,18,19]. However, research on the effect of comorbid AUD on the risk of developing cardiometabolic diseases in people with PTSD is so far limited, and the majority of studied populations regarding impacts of comorbid depression on the association between PTSD and cardiometabolic diseases consist of war veterans and military service personnel. In addition, since the effect of depression and anxiety on the cardiovascular system seems comparable, Batelaan, et al. [20] suggests adjusting for anxiety when examining the effect of depression on the cardiovascular system.

As people with PTSD are a particularly vulnerable group for developing cardiometabolic diseases, which may negatively impact quality of life and life expectancy, knowledge about how health and disease develops over time, including possible risk factors for development of serious diseases, is of importance for developing measures to prevent diseases and early death. Confounding factors such as gender, increased age and SES appear to impact on health outcome [5, 21,22,23], and should thus be considered when studying health and diseases in a population. Moreover, mental disorders are associated with lower SES [24]. Large population-based retrospective cohort studies, investigating the occurrence of cardiometabolic diseases, including the effect of SES, comorbid anxiety, comorbid AUD and comorbid depression, in people with PTSD, are limited and there is a need for further exploration. The present study thus aims to answer the following research questions; 1) what is the risk of developing cardiometabolic diseases including T2DM over time in PTSD patients? and 2) to what extent do SES, comorbid anxiety, comorbid AUD and comorbid depression attenuate associations between PTSD and risk of developing cardiometabolic diseases?

Methods

Study design and population

This is a register-based cohort study combining sociodemographic information from Statistics Norway and information on physical diseases and mental disorders obtained from the Norwegian Patient Registry (NPR). NPR holds data on all registered diagnoses obtained during contacts with specialist health care services. The unique number assigned to each person enables tracking the same patient over time from hospital to hospital, and analysing the data without concern of duplicates of cardiometabolic events. All diagnoses are based on the background of the International Classification of Diseases 10th Revision (ICD-10) codes [1]. As illustrated in Fig. 1, the total identified study population consisted of all adults 18 years or older who were legal residents in Norway starting from January 1, 2008 to December 31, 2016 (N = 4 652 365). PTSD patients (N = 8 997) were identified during the preceding 2 years (2008–2009), and then followed for cardiometabolic comorbidity for 6 years, from January 1, 2010, through December 31, 2016. People who were registered as deceased (N = 363 783) during the study period (2008–2016) were excluded from analysis, and a two years washout period (2008–2009) was performed to exclude all individuals with cardiometabolic diseases (N = 792 with PTSD and N = 238 572 without PTSD).

Fig. 1
figure 1

Study population flow chart

Full size image

Exposure, explanatory and outcome variables

All diagnoses were received during outpatient and inpatient contacts with specialist healthcare between 2008 and 2016. We defined cardiovascular diseases, endocrine and metabolic diseases according to the ICD-10 codes [1]. The following dichotomous variables in Table 1 representing ICD-10 diagnostic categories were coded to specific disorders.

Table 1 ICD-10 codes and year of diagnosis for primary exposure variables, explanatory variables and outcome variables
Full size table

Covariates

Age and gender were used as covariate variables. The age variable (per 1. January 2008) was used as a continuous variable. Gender was coded 0 for males and 1 for females. We included being recipient of social benefits as an index for SES. According to the EU-scale, household poverty is defined as less than 60% of the annual median disposable equivalised household income after adjusted for the size of household members. A dummy variable (0 and 1) was constructed to indicate whether participants were defined as living in a household with poverty in each year. A sum score for all years was calculated with 0 being the lowest score (no years in poverty) and 9 being the highest sum score (in poverty during all years in the study). A higher sum score indicates more years with household poverty between 2008 and 2016.

Statistical analysis

Before the analysis, a directed acyclic graph (DAG) [25] was created to identify and illustrate unobserved and adjusted variables, and causal pathways (Online Supplementary Fig. 1). The Cox proportional regression models were applied to estimate the risks of cardiometabolic diseases (event outcomes) among PTSD patients (an independent risk factor). Hazard ratios (HRs) with 99% confidence intervals (99% CIs) were reported, with calendar year as the underlying time axis. A stepwise regression was applied: model 1 presents age and gender adjusted HR estimates; HR estimates in model 2 adjusted for age, gender and recipient of social welfare (SES indicator); HR estimates in model 3 adjusted for age, gender, SES indicator and comorbid anxiety; HR estimates in model 4 adjusted for age, gender, SES indicator, comorbid anxiety and comorbid AUD; HR estimates in model 5 adjusted for age, gender, SES indicator, comorbid anxiety and comorbid depression; and HR estimates in model 6 adjusted for age, gender, SES indicator, comorbid anxiety, comorbid depression and comorbid AUD. Estimates were judged as statistically significant when p-values ≤ 0.01. The analyses were performed using Stata SE/16.

The proportional hazard assumptions were tested for each model on the basis of Schoenfeld residuals. Since some of the covariates did not often meet the proportional hazards assumption after fitting each model (“estat phtest tests”), interaction terms between time and time-variant covariates (age and SES), including specified stratification for gender in all models, were added.

To evaluate our findings against potential confounding factors, a sensitivity analysis was performed. The E-values for the association in the model adjusted for comorbid mental disorders found in our Cox regression analysis were calculated. The E-value represent the minimum strength of association between outcome and unmeasured confounding factors that would be required to explain away the association between the outcome variable and explanatory variables presented in the Cox regression. An E-value that is small relative to the HR would indicate that even weak confounding effects could explain the association, while a relatively large E-value suggests that the association is unlikely to be explained by confounding factors [26] (Online Supplement Table 1).

Ethics

All study procedures were approved by the Norwegian Regional Committee for Medical and Health Research Ethics (ref: 17/26919–5).

Results

Description of the study population

A descriptive summary of the study population's characteristics is presented in Table 2. Among the total of 7 852 adults (> 18 years) registered with the diagnosis PTSD in the period 2008–2009, a higher number of women (N = 5 372, 68.4%) than men (N = 2 480, 31.6%). For the population with PTSD patients, the mean age was 37.8 years (SD = 11.5), and the mean for receiving social welfare (ranging from 0 to 9) 0.6 (SD = 1.2). For the population without PTSD, the mean age was 44.7 years (SD = 15.9), and the mean for receiving social welfare was 0.1 (SD = 0.4).

Table 2 Descriptive summary of the study population
Full size table

During 2008–2009, the proportion registered with comorbid mental disorders was higher among PTSD patients compared with the general population without PTSD. Compared with the population without PTSD, a smaller proportion of the population with PTSD patients was registered with hypertensive diseases, ischaemic heart diseases and cerebrovascular diseases during the period 2010–2016. A larger proportion of the population with PTSD patients was registered with pulmonary heart diseases, diabetes mellitus, obesity and metabolic disorders than the population without PTSD during the period 2010–2016.

Risks of cardiometabolic diseases among PTSD patients and the effect of SES and comorbid mental disorders

Table 3 presents results (HR with 99% CI) from stepwise Cox regression models, showing HR for cardiometabolic diseases in PTSD patients adjusted for age and gender in model 1; age, gender and SES indicator in model 2; age, gender, SES indicator and comorbid anxiety in model 3; age, gender, SES indicator, comorbid anxiety and comorbid AUD in model 4; age, gender, SES indicator, comorbid anxiety and comorbid depression in model 5; and age, gender, SES indicator, comorbid anxiety, comorbid depression and comorbid AUD in model 6.

Table 3 Cox regression models showing adjusted HRs of cardiometabolic comorbidities among PTSD patients compared to the population without PTSD
Full size table

Age and gender adjusted estimates in model 1 presents significantly higher HRs of all cardiometabolic diseases among PTSD patients compared to the population without PTSD, ranging from 3.5 (99% CI 3.1–3.9) for hypertensive diseases to 6.5 for obesity (99% CI 5.7–7.5). Although adjusting for SES and comorbid mental disorders in model 2–6 resulted in reductions of HR estimates of all cardiometabolic diseases, PTSD patients still had statistically significant higher HRs in all cardiometabolic diseases than the population without PTSD. E-values that were calculated for model 6 indicated relatively robust estimates of HRs (Online Supplement Table 1).

Compared to model 1, HR estimates in model 2 decreased for all cardiometabolic diseases; 11.4% for hypertensive diseases; 14.3% for ischaemic heart diseases, 17.4% for pulmonary heart diseases; 16.7% for cerebrovascular heart diseases; 18.0% for diabetes mellitus; 23.1% for obesity, and 19.4% for metabolic disorders.

Compared to model 2, HR estimates in model 3 decreased for all cardiometabolic diseases; 16.1% for hypertensive diseases; 19.4% for ischaemic heart diseases; 23.7% for pulmonary heart diseases; 20.0% for cerebrovascular diseases; 19.5% for diabetes mellitus; 28.0% for obesity, and 26.0% for metabolic disorders.

Compared to model 3, HR estimates in model 4 decreased for all cardiometabolic diseases; 7.7% for hypertensive diseases; 6.9% for ischaemic heart diseases; 6.9% for pulmonary heart diseases; 10.7% for cerebrovascular heart diseases; 6.1% for diabetes mellitus; 5.6% for obesity, and 16.2% for metabolic disorders.

The largest reduction in HR was observed in model 5 and 6. Compared to model 3, HR estimates in these models decreased for all cardiometabolic diseases; 30.8% for hypertensive diseases; 31.0% for ischaemic heart diseases; 34.5% for pulmonary heart diseases; 35.7% for cerebrovascular heart diseases; 33.3% for diabetes mellitus; respectively 41.7% and 44.4%, for obesity, and respectively 37.8% and 40.5% for metabolic disorders. Compared to model 5, HR estimates in model 6 only decreased for obesity and metabolic disorders, 4.8% and 4.3% respectively.

Discussion

The main finding in this large retrospective registry-based population study is that PTSD patients had increased risk of developing cardiometabolic diseases, compared to the general population without PTSD. Adjusted for SES and comorbid mental disorders, the association was attenuated, strongest by comorbid depression.

Our results confirm previous findings of associations between PTSD and increased risk of developing cardiometabolic diseases [5,6,7]. The theory of allostatic load [27], which refers to the cumulative burden of chronic stress and stressful life events, can contribute to the understanding of the association between PTSD and cardiometabolic diseases. Allostatic load is associated with increased risk of several cardiometabolic diseases, and early traumatic life experiences have been associated with high levels of allostatic load in adulthood [28]. If environmental challenges exceed the individual's ability to cope, there is an allostatic overload, which might involve consequences such as low SES, poor sleep, physical inactivity, smoking, alcohol consumption and unhealthy eating [28], which in turn can lead to poor health outcomes. For example, in a recent qualitative study [10], trauma exposed inpatients described how traumatic stress impacted their eating behaviors, and for some, consumption of food and snacks with high amounts of sugar, salt and saturated fat.

Furthermore, traumatic experiences can cause several physiological and somatic responses, which in turn can influence the developing of cardiometabolic diseases. Stress affects secretion of stress hormones and thereby peripheral organs, especially the cardiovascular and immune systems, i.e. via neurotransmitters and hormones in the hypothalamus–pituitary–adrenal axis (HPA axis) and the sympathetico-adrenal-medullary system (SAM system) [29]. Chronic stress appears to affect metabolic processes through oxidative stress in fat cells and the cells' mitochondria. This might lead to increased visceral fat deposits [30]. Increased inflammatory cytokine levels in PTSD patients has also been found [31].

The attenuating effect of SES on the association between PTSD and the risk of developing cardiometabolic diseases supports previous reports showing that the risk of developing cardiometabolic diseases such as CVD and T2DM increases with lower SES [21]. Long-term social assistance recipients can limit life opportunities and income trajectories [32], and lead to poor health outcomes and social health inequities [23], which should be taken into account by the healthcare providers and policy makers.

Furthermore, results from previous research suggests comparable effects of depression and anxiety on the risk of cardiovascular diseases [20]. Surprisingly, our results documented a larger attenuating effect of comorbid depression than comorbid anxiety on the risk of all cardiometabolic diseases in PTSD patients. However, according to the ICD-10, PTSD is categorized as an anxiety disorder. Since symptoms of anxiety disorders and PTSD can overlap, this may explain why depression showed a greater effect than anxiety in this current study.

The attenuating effect of comorbid AUD on the association between PTSD and the risk of developing cardiometabolic diseases has been less studied, although the consequences of long-term alcohol use on the cardiovascular system, e.g. hypertension, cardiac arrhythmia, cardiomyopathy and heart failure are well established [33]. However, PTSD patients with comorbid AUD have also been shown to have poor access and use of healthcare services [34, 35], which might explain the increased risk of cardiometabolic diseases.

Our findings of an attenuating effect of comorbid depression are in line with previous studies. Earlier research has documented attenuating effects of comorbid depression in several cardiometabolic diseases among people with PTSD [36,37,38]. Furthermore, biological mechanisms in depression, including dysregulation of the HPA-axis function [39], and increased levels of pro-inflammatory cytokines [40], might explain the increased risk of developing cardiometabolic diseases. Genetic mechanisms are possible underlying factors for the development of both depression and cardiovascular diseases [41].

Additionally, the attenuating effect of comorbid mental disorders might both be related to confounding behavioral factors such as physical inactivity, poor dietary habits and smoking [42,43,44,45,46], while in turn can increase the risk of cardiometabolic diseases including T2DM [47].

Strengths and limitations

There are several strengths in this large population-based cohort-study. Firstly, the study includes data about all registered cases of cardiometabolic diseases across two large groups over six years, which thus brings adequate statistical power to detect differences. Secondly, the coverage of healthcare services and the quality of health records in Norway is considered to be high [48], and clinical set diagnosis in the NPR are found to be accurate and consistent compared to research-based diagnosis of severe mental disorders [49]. The NPR data used in this study are therefore considered to be highly reliable. Thirdly, clinically set diagnosis from specialist care may, or may not, be reliable, but systematic differences between the populations are nevertheless unlikely, making the relative estimates valid.

There are, however, some challenges in the present study that are important to note. The e-values calculated for model seven should be interpreted with caution because they do not necessarily reflect reality. Further, the registries do not have information about the type of traumatic exposure, psychotherapy, medication use, volume of health care utilization (i.e. number of clinic encounters per month or year), or health behavior (e.g., eating behavior, nicotine dependence and physical activity level) that may constitute confounding factors; information about severity of AUD (e.g., volume and frequency of consumption); information about the onset and duration of problematic alcohol consumption; or information about the time where diagnosis of PTSD and comorbid mental disorders were set. There is a possibility to adjust for all F-codes (00–99) in the ICD-10 as an indicator for volume of healthcare utilization, however due to concerns for overadjustment, we did not. Further, a worryingly large proportion of trauma-exposed people with PTSD symptoms do not seek help and are more likely to use alcohol as self-medication [50]. In addition, misdiagnosis and under diagnosis of PTSD is not uncommon [51]. Finally, use of registry-data from NPR might promote a risk of misclassification.

Implications and further research

The results of this present study highlight some important implications that could be relevant for clinical practice and policy development practice. As this study documented higher risk of cardiometabolic diseases among PTSD patients compared to the population without PTSD, this suggests a need for somatic health in PTSD patients to be given high priority by healthcare professionals, and that screening for somatic health should be considered as a routine part of the follow-up of PTSD patients to expose cardiometabolic risk factors and diseases and be able to start early preventive interventions. Furthermore, healthcare professionals should be aware of the additional burden these comorbid mental disorders may represent for PTSD patients' cardiometabolic health, and consider offering concomitant treatment of the disorders. In addition, with a political goal of levelling out social inequalities in health, the authorities should be attentive to, and take seriously, the increased risk that low SES may entail for cardiometabolic health in vulnerable groups such as PTSD patients. The complexity of the association between PTSD and cardiometabolic risk factors and diseases warrants future research to explore the effects of possible confounding factors such as health behavior (e.g., diet, physical activity, alcohol use and smoking).

Conclusion

The current study shows that PTSD was associated with increased risk of developing cardiometabolic diseases, and that SES and comorbid mental disorders attenuated the risk. In other words, low SES and comorbid mental disorders in PTSD patients were associated with increased risk of developing cardiometabolic diseases. This implies that social inequalities in health among vulnerable populations like PTSD patients should be targeted.