Background
Individuals with a severe mental illness (SMI), such as schizophrenia, bipolar disorder or major depression, constitute a part of the population that is more prone to suffer from obesity and the related co-morbidities [
1‐
3]. The reasons for this are many-faceted and include socio-economic factors, sedentary lifestyle, diet, smoking and medication [
2‐
4]. It has been demonstrated repeatedly that the segment of the population with SMI suffers from metabolic syndrome (MetS) more frequently in comparison to the background population [
1,
5‐
8].
Metabolic syndrome is comprised of a cluster of reversible risk factors that focus the attention of the clinician to patients with an increased risk of cardiovascular disease (CVD) and type 2 diabetes mellitus (DM). Several definitions of MetS have been proposed by The World Health Organization, The International Diabetes Federation and the National Cholesterol Education Program [
9‐
11]. In 2009, the harmonized criteria of the MetS were proposed including central obesity, raised blood pressure, raised blood glucose, raised triglyceride level (TG), and/or lowered high density lipoprotein (HDL) cholesterol level, of which three have to be met for a person to be diagnosed with the MetS [
12]. Individuals suffering from the MetS have double the risk of developing CVD and a five times increased risk of developing DM in comparison to the background population [
12]. While some have argued for the dismissal of the MetS as a concept [
13,
14], all of the major entities that have defined it recommend further research into it and the factors that dispose to CVD and diabetes [
9,
11,
12].
An autopsy is widely considered the ultimate diagnostic tool and several studies have demonstrated the benefit of autopsies in diagnosing overlooked diseases [
15‐
17]. Hereditary heart diseases and the ‘Back to Sleep’ campaign in sudden infant death syndrome are just some of the examples of autopsy based diagnostics and epidemiology of the dead that have prevented morbidity and mortality [
18‐
20].
The aim of this study was to define and evaluate the post mortem (PM) criteria of the MetS, based on the Danish autopsy-based SURVIVE study [
21]. Establishing well-defined PM criteria will make it possible to link the morbidity and mortality determined at the autopsy to the MetS and thereby exploit the potential of the autopsy data in elucidating diagnostics, pathophysiology and potential treatment of the MetS. To accomplish this goal, we focused on specific autopsy related data (police reports, biochemical and toxicological results; i.e. PM information) and we tested it against registry data [diagnosis codes and prescribed medication; i.e. ante mortem (AM) information].
Results
The mean (± SD) age of the study population was 50.4 (± 15.5) years with men being significantly younger than women (men: 48.0 [± 14.2] years; women: 54.3 [± 16.6] years;
P < 0.001). We determined a median PMI (interquartile range) of 114 (84–156) hours. We were unable to determine the PMI in 72 cases. All ethnicities fell within the same reference measurement of the WC, as stated by the IDF (≥ 94 cm men; ≥ 80 cm women). The basic anthropometric and biochemical data are listed in Table
3. Adjusting for age in Table
3 did not produce significant differences in either criteria (data not shown). Investigating the effect of PMI on the biochemical measurements, we found a significant correlation with TG (
r =0.22,
P < 0.001) but no significant correlations for the other criteria (total cholesterol:
r =− 0.14,
P = 0.07. Albumin/creatinine ratio:
r =0.02,
P = 0.78. HbA1c:
r = − 0.01,
P = 0.82). From the autopsy reports, we concluded that none of the deceased with LVH or an elevated albumin/creatinine ratio suffered from cardiomyopathy or kidney disease.
Table 3
Anthropometry and biochemical results related to the metabolic syndrome
Waist circumference (cm), converted | 97.3 (95.3; 99.3) | 94.2 (91.3; 97.0) | 443 (272/171) | 0.08 |
Triglycerides (mmol/L)a | 2.90 (2.60; 3.40) | 3.00 (2.60; 3.45) | 308 (185/123) | 0.39 |
Total cholesterol (mmol/L)a | 3.00 (< 3.00; 3.20) | 3.10 (< 3.00; 3.2) | 216 (90/126) | 0.11 |
Albumin/creatinine ratiob | 16.36 (7.88; 49.91) | 22.35 (12.02; 67.59) | 248 (75/173) | 0.11 |
Glycated haemoglobin (mmol/mol)a | 32 (29; 36) | 33 (29; 37) | 364 (140/224) | 0.67 |
We compared our PM definition of the MetS with each of the proposed PM criteria, based on the PM information, with the corresponding AM definition and information. The reproducibility of our combined PM MetS definition resulted in a moderate agreement (κ = 0.51). For the criteria regarding the TG, total cholesterol, hypertension and blood glucose, the agreement ranged from slight to moderate (Table
4). Information on the WC was only available from the PM information and was thus omitted from Tables
4 and
5.
Table 4
PM MetS estimation and individual criteria compared to the AM definition and corresponding criteria
Post mortem |
MetS yes | 29 (16/13) | 27 (14/13) | 0.51 (0.56/0.45) |
MetS no | 17 (7/10) | 370 (235/135) |
Post mortem |
Triglycerides | | NA | NA | 0.19 (NA/NA) |
Total cholesterol | Criterion met | 18 (13/5) | 7 (NA) | 0.36 (0.47/0.22) |
Criterion not met | 44 (21/23) | 374 (234/140) |
Hypertension | Criterion met | 103 (55/48) | 153 (106/47) | 0.24 (0.19/0.33) |
Criterion not met | 26 (14/12) | 161 (97/64) |
Blood glucose | Criterion met | 34 (18/16) | 40 (21/19) | 0.55 (0.54/0.56) |
Criterion not met | 5 (NA) | 364 (229/135) |
Table 5
Evaluation of the PM definition of the metabolic syndrome and individual criteria
PM definition MetS | 0.63 (0.70/0.57) | 0.93 (0.94/0.91) | 9.27 (6.05; 14.19) | 2.3% (− 0.7%; 5.2%) | 0.175 |
PM cholesterol | 0.29 (0.38/0.17) | 0.98 (0.98/0.98) | 15.80 (6.89; 36.27) | − 8.4% (− 11.4%; − 5.3%) | < 0.001 |
PM hypertension | 0.80 (0.80/0.80) | 0.51 (0.48/0.58) | 1.64 (1.42; 1.89) | 34.5% (29.3%; 39.8%) | < 0.001 |
PM triglycerides | 1 (NA/NA) | 0.30 (NA/NA) | 1.42 (1.34; 1.51) | 70.0% (65.7%; 74.2%) | < 0.001 |
PM triglycerides w/o PM biochemistry | 0.50 (NA/NA) | 0.98 (NA/NA) | 31.50 (6.65; 151.21) | 1.4% (0.1%; 2.6%) | 0.077 |
PM glucose | 0.87 (0.82/0.94) | 0.90 (0.92/0.88) | 8.81 (6.41; 12.10) | 7.9% (5.0; 10.8%) | < 0.001 |
We found no significant difference when we estimated the MetS based only on the PM information, in comparison to estimating the MetS based only on the AM information (Table
5). Our PM definition resulted in high specificity (0.93) and moderate sensitivity (0.63) with a reasonable likelihood ratio (LR) (9.27) (Table
5).
The biochemical measurements of the PM TG were high when compared to the cut-off value (Table
3). Although using the PM triglyceride biochemistry results for the triglyceride criteria produced no false negative results, hence the perfect sensitivity, this yielded an immense overestimation. Omitting the PM triglyceride biochemistry results produced a far more accurate estimation, with no significant difference, between the PM and the AM information. None of the deceased had a total cholesterol level above 4.0 mmol/L in the PM biochemical analyses. Therefore, using only the PM information for the cholesterol criteria resulted in a significant underestimation of that criterion. Conversely, using only the PM information on the hypertension and glucose criteria significantly overestimated those criteria.
Discussion
To our knowledge, this is the first study that defined the criteria for the diagnosis of the MetS in a post mortem perspective.
We opted to use the AM information from the registries as the gold standard. A priori, the specificity of this information is close to one. The registry information is, however, not exhaustive regarding an individual’s health status. Only diagnoses from hospital admissions were recorded; diagnoses from the GPs of the individual subjects were not available through our registries. Conversely, in a forensic autopsy setting, access is available to the police reports, which include the police questioning of the GPs regarding a decedent’s health status. Although this information might not be exhaustive, it may be regarded as information with a high specificity. The registry information regarding prescribed medication reflects the diagnoses that a GP finds appropriate to treat and is consequently an indirect source for measuring a person’s health status. A caveat when defining the health status of individuals with SMI from registries or GPs is that the prevalence of dyslipidaemia and hypertension are frequently underestimated and undertreated [
32]. Although our proposed PM definition of the MetS was not significantly different from the AM definition, it yielded only a moderate level of agreement. While there were some gender differences in Cohen’s κ for some of the individual criteria, the gender-specific agreement in the final PM model was quite similar. Since the prevalence of the MetS in our study cohort is relatively low, the calculated Cohen’s κ might underestimate the actual agreement between the PM and the AM definition and criteria [
33]. Therefore, the sensitivity and specificity may reflect the most correct estimate of agreement. Again, no gender differences regarding sensitivity and specificity were evident. In addition, since the information originates from different sources in our setup, comparing the PM information with the AM information will generally produce some discrepancies. Consequently, one should regard the level of agreement with some reservations.
Measuring and interpreting the PM biochemical analyses posed several challenges. The total cholesterol decreases PM leaving only elevated concentrations interpretable [
34,
35]. Since none of our subjects had a total cholesterol above 4.0 mmol/L, our findings may corroborate the challenge of analysing cholesterol in PM samples. The levels of PM TG have been reported to be affected by too many variables to make it useful for interpretation [
34,
35]. In line with the study by Girard et al. [
35], the measured TG in our study increased with PMI. Moreover, the reference values of cholesterol and TG in the living are frequently based on a premise of fasting. We concluded that total cholesterol and TG were imprecise parameters and omitted them from our final MetS estimation. While direct measurement of blood glucose levels PM is not feasible [
36], several studies have proven HbA1C as a robust marker of long-term blood glucose status in PM measurements [
34,
36‐
40]. As we used an immune-assay to measure HbA1c, our results might be biased towards higher values due to diminished specificity compared to chromatographic measurements [
36]. Our PM information differed significantly from the AM information. Although this could be a result of underreported and undertreated hyperglycaemia in people with SMI [
32], it may reflect the fact that our cut off value of 38 mmol/mol does not yield an ICD10 code of diabetes in a clinical setting; contributing to our PM definition overestimating the prevalence of this criterion.
Diagnosing hypertension at autopsy can be problematic. Few markers exist and they rely on the exclusion of other diseases. In hypertension, the myocytes in the left ventricle of the heart undergo hypertrophy. Although it manifests most frequently as concentric hypertrophy, it can also present as eccentric hypertrophy. While other diseases may affect the pattern of hypertrophic response, hypertension is still a part of these conditions [
41]. When it comes to ruling out cardiomyopathies, autopsy data, specifically histology, provides one of the hallmarks of diagnosing these diseases. Therefore, we considered LVH, excluding cardiomyopathies, as an appropriate marker for hypertension.
The WC was the only criterion that we were unable to validate with AM information. However, as one of the exclusion criteria was moderate to advanced putrefaction, cases with markedly PM bloating of the abdomen were not included. In addition, the use of a conversion factor from the supine to the standing position [
24] further corrected the estimated WC.
Selection bias must be considered since the study is based on forensic autopsy material. A forensic autopsy is performed on request from the police in cases of accidents, suicides, (suspected) homicides and when the manner and/or cause of death remains unknown at the medico-legal examination. Therefore, the results are in line with most other study populations based on forensic autopsies that are descriptive for a specific study population and not necessarily representative for the general population. In such a study population, a precise setup of inclusion criteria—such as age, gender, specific diagnosis and equivalent healthy controls—may accommodate some of the selection bias. Furthermore, as the present study was based on a study population of deceased individuals with SMI, we cannot rule out the possibility of a different result if the PM model was employed on deceased mentally healthy individuals. However, with the aforementioned underestimation and undertreatment of dyslipidaemia in individuals with SMI [
32], the evaluation of the PM model based on individuals with SMI might actually underestimate the specificity of the model compared to the same evaluation based on a mentally healthy study population.
Conclusions
With this study, we proposed a valid estimate for the PM diagnosis of the MetS. While neither the PM nor the AM information are exhaustive regarding an individual’s health status, combining both the PM and the AM information might provide the best estimation. However, bearing in mind that the result is an approximation and not a conclusive answer, employing PM information alone is applicable. Although PM TG and total cholesterol measurements are unreliable and should be omitted, PM HbA1C remains a valid measurement. Employing autopsy material and our PM MetS definition, we believe that autopsy-based data may promote the research of the MetS and associated morbidities from a novel perspective.
Authors’ contributions
MRC set up the study design, analysed and interpreted the data, and drafted the manuscript. AB was a major contributor to the design of the post mortem metabolic syndrome model. MEM conducted a pilot study for the application of the post mortem metabolic syndrome model on the study population. JLT revised the manuscript critically. JR, NL, and JB revised the manuscript critically and provided continuous guidance throughout the study. All authors read and approved the final manuscript.