Introduction

Type 2 diabetes mellitus (T2DM), characterized by hyperglycemia and dyslipidemia, is a metabolic disorder caused by imbalance among β-cell function, physical inactivity, obesity, chronic fuel surfeit status, and insulin resistance [1]. T2DM may lead to a substantial health burden on people and direct and indirect burdens on the healthcare system and society [2].

The prevalence of diabetes has increased substantially in Asian countries in recent years [3]. This increase has been attributed to the increase in central obesity, physical inactivity, the aging population, and the prevalence of Western diets in Asian populations [3–5]. Although the metabolic manifestations in both types of diabetes are similar, T2DM is more heterogeneous than type 1 diabetes, and involves additional aspects, including genetic factors, epigenetic effects, early-life environment, lifestyle, nutrition imbalance, socioeconomic status, a deregulated neurohumoral network, and insulin resistance [6–9]. A previous study documented that adipocyte dysfunction may increase the concentrations of inflammatory cytokines and aggravate insulin resistance in muscle [10]. However, the underlying mechanism and complex interaction between T2DM and inflammation, particularly the chronic ones (e.g., adipocyte inflammation, periodontitis, and hepatitis C infection), remain unclear [10–12].

Chronic osteomyelitis (COM), a well-known chronic infection–inflammation status that is resistant to treatment and prone to relapse, is a common complication of T2DM [13]. After conducting a literature review regarding the relationship between T2DM and COM, we observed that most previous studies have focused on the risk of COM among patients with T2DM [13–16], whereas few have addressed the possibility that COM might be a risk factor for T2DM. Using a large cohort of 23 million patients identified from the Taiwan National Health Insurance Database (NHIRD), we assessed the risk of newly developing T2DM among patients with COM.

Materials and methods

Data sources

We used reimbursement claims data from the Taiwan National Health Insurance (NHI) program launched in March 1995. The dataset is managed by the National Health Research Institutes (NHRI) and the details of the NHI program have been described elsewhere [17]. We used a subset of the NHIRD containing healthcare data, including files on inpatient claims and a registry of beneficiaries. The International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) was used to identify diseases. The high accuracy and validity of the T2DM and COM diagnosis used in this study based on the ICD-9 codes applied in the NHIRD were previously confirmed [18–20]. Taiwan launched the NHI program in 1995, which is operated by a single buyer, the government. All insurance claims are scrutinized by medical reimbursement specialists and through peer review. The diagnoses of T2DM and COM were based on the ICD-9 codes determined by clinical physicians. Therefore, the diagnoses and codes for T2DM and COM should be accurate and reliable.

Study patients

The COM cohort included patients with newly diagnosed COM (ICD-9-CM code 730.1) from 1997 to 2010. For each COM patient, four people without a medical history of COM and diabetes who were frequency-matched according to sex and age (each 5-year span) in the same period were randomly selected as the comparison cohort. The index date for the COM cohort was defined as the date of diagnosis. The index date for the comparison cohort was defined as the middle date of the same index month used for their matched COM patients. Patients with a medical history of diabetes who were diagnosed before the index date, younger than 20 years old, or were missing information on age or sex were excluded.

Outcome measures

We followed up the COM cohort and the comparison cohort to compare the incidences of diabetes (ICD-9-CM code 250) until the end of 2010 or until the patient was censored because of death or withdrawal from the insurance program. Sociodemographic characteristics including sex, age, and income of the two cohorts were assessed. The baseline comorbidities included hypertension (ICD-9-CM codes 401–405), hyperlipidemia (ICD-9-CM code 272), and chronic kidney disease (CKD) (ICD-9-CM code 585).

Statistical analysis

Differences in sociodemographic factors and comorbidities between the COM and the comparison cohorts were examined using the Chi-square test for category variables and the t-test for continuous variables. The incidence–density rates according to demographic status and comorbidity were calculated for both cohorts. The incidence rate ratio (IRR) with a 95 % confidence interval (CI) was calculated using the Poisson regression model. Multivariate Cox proportional hazard models, adjusted for potential confounding factors, were used to assess the risk of developing diabetes associated with COM. To assess the difference in the diabetes-free rates between the two cohorts, the Kaplan–Meier analysis and log-rank test were applied. The statistical significance level was set at a two-tailed probability value of p < 0.05. All analyses were performed using SAS software (SAS System for Windows, version 9.1) and the Kaplan–Meier survival curve was plotted using R software (R Foundation for Statistical Computing, Vienna, Austria).

Ethics and consent

Personal information was de-identified before the release of the NHIRD data; therefore, this study was exempt from approval by the Institutional Review Board.

Results

This study consisted of 20,641 patients in the COM cohort and 82,564 people in the comparison cohort. Both cohorts exhibited similar sex and age distributions. Comorbidities including hypertension, hyperlipidemia, and CKD were more prevalent in the COM cohort than in the comparison cohort (p < 0.0001) (Table 1).

Table 1 Demographic characteristics and comorbidities in the chronic osteomyelitis (COM) cohort and the comparison cohort
Full size table

The COM cohort had a higher incidence rate of diabetes than the comparison cohort (29.1 vs. 18.2 per 10,000 person-years; IRR = 160, 95 % CI = 1.43–1.86), with an adjusted hazard ratio (aHR) of 1.64 (95 % CI = 1.44–1.87) (Table 2). The COM-to-reference IRR was higher in men (IRR = 1.67, 95 % CI = 1.58–1.77) than in women (IRR = 1.45, 95 % CI = 1.33–1.58). The age-specific IRR increased with age and was the highest in the subgroup aged 20–34 years (IRR = 5.90, 95 % CI = 5.26–6.62), with an aHR of 4.74 (95 % CI = 2.37–9.49). The aHR of diabetes was the highest in the subgroup with a monthly income of over 22,800 New Taiwan dollars (NT$) (aHR = 3.37, 95 % CI = 2.04–5.55).

Table 2 Overall and specific incidence and hazard ratio of diabetes
Full size table

The incidence rate of diabetes was higher in the subgroups with hypertension, hyperlipidemia, or CKD compared with that of the non-comorbid counterparts. The COM-to-reference IRR was 1.58 (95 % CI = 1.06–1.32) and 1.18 (95 % CI = 1.50–1.67) for the groups without and with hypertension, respectively. The significant COM-to-reference IRR was also observed in the non-hyperlipidemia and non-CKD groups. Moreover, the aHRs indicated that COM was associated with an increased risk of diabetes in patients without hypertension (aHR = 1.70, 95 % CI = 1.47–1.96), hyperlipidemia (aHR = 1.66, CI = 1.45–1.89), or CKD (aHR = 1.65, 95 % CI = 1.45–1.87) (Table 3). The Kaplan–Meier survival analysis revealed that the diabetes-free rate was 1.18 % lower in the COM cohort than in the comparison cohort (log-rank p < 0.0001) (Fig. 1).

Table 3 Incidence and hazard ratio of diabetes by the presence of comorbidities
Full size table
Fig. 1
figure 1

The Kaplan–Meier curves of the diabetes mellitus-free rates in patients with and without chronic osteomyelitis

Full size image

Discussion

This study yielded several intriguing results. First, this study is the first population-based cohort study to report a significantly increased T2DM risk associated with COM. Second, the association between COM and the increased risk of T2DM was stronger in the wealthier subgroup and the younger subgroup.

The biological mechanisms of the relationship between COM and the risk of T2DM remain unclear. Previous studies have demonstrated that increasing levels of proinflammatory cytokines that block downstream insulin signaling and interrupt insulin action, such as TNF-α, interleukin-β, and interleukin-6, contribute to preclinical stages of T2DM [21, 22]. In COM patients, these proinflammatory cytokines were observed to increase markedly in the bone compartment [23, 24]. Therefore, we suspected that localized proinflammatory cytokines might exert system effects on whole-body insulin resistance. In addition to the proposed effect of proinflammatory cytokines, other possible explanations exist for the association between COM and the risk of T2DM. First, patients with COM might make frequent medical visits and have routine check-ups, including check-ups on biochemical profiles and blood glucose levels, which may increase the likelihood of detecting T2DM. Second, patients with COM are typically immobile because of pain or difficulty in engaging in physical activity [25]. Because physical activity prevents the occurrence of T2DM [26], the immobile status of COM patients might increase the risks of obesity, metabolic syndrome, and T2DM. Furthermore, the genetic predisposition of these two diseases to COM is linked with IL-1α, IL-4, and IL-6 polymorphism [27]. A Finnish diabetes prevention study reported that promoter polymorphisms of TNF-α and IL-6 could predict the risk of T2DM [28]. The association between COM and the risk of T2DM might also result from the genetic predisposition to both COM and T2DM of our study cohorts. Further studies on genetic analysis and the role of physical activity are recommended in order to elucidate the association.

The present results indicated that, in the subgroup with higher income, COM patients had a higher risk of developing T2DM compared with that of the non-COM counterparts, which is intriguing because T2DM has been reported to be more prevalent among those with a lower socioeconomic status [29–31]. Socioeconomic disparities are associated with insulin resistance and altered glucose metabolism [32, 33]. One possible explanation for our finding is that the effects of insulin resistance produced by COM are more easily observed in wealthier people who are assumed to have fewer alterations in glucose metabolism. Future studies are warranted for further investigation of this discrepancy.

In addition, we observed that, among young people aged less than 55 years, COM patients had a higher risk of T2DM compared with that of the non-COM people, which is remarkable because T2DM is a well-known age-related disease [34]. Chronic low-grade inflammation is a converging process linking both normal aging and age-related diseases, including T2DM [35–37]. However, the present study reported that younger patients had a significantly higher COM-to-reference hazard ratio of diabetes mellitus than the elderly subgroup. The underlying mechanism requires further investigation. We suggest that aging might mask the effects of the parallel, pathogenic, inflammation-related process of COM in the development of T2DM.

This study has several strengths. It is the first population-based study in which the risk of T2DM was compared between patients with and without COM. Patients with COM and the age- and sex-matched controls were identified from a dataset of 23 million enrollees in a national insurance program comprising over 98 % of the entire population in Taiwan. Insurance claims of hospitalization and consecutive care of COM ensure the diagnosis of COM. The Taiwan NHI program has a strict auditing system to prevent fraudulent healthcare claims, thereby ensuring the reliability of COM diagnosis based on insurance claims. This large population-based database containing comprehensive electronic medical records provided complete information on the incidence of T2DM and associated comorbidities, including hypertension, hyperlipidemia, and CKD.

However, several limitations should be addressed. We did not obtain precise information on patients’ body weight, body mass index, and fat distribution. Information on physical activity was also unavailable. Additionally, we could not rule out the possibility that the immobilization exhibited by a proportion of COM patents might predispose COM patients to a higher risk of T2DM. Consequently, the T2DM risk assessed in this study might be overestimated. Although our data indicated that comorbidities were more prevalent among patients with COM compared with the comparison cohort, the results suggested a significant association between COM and T2DM after we controlled for comorbidities and sociodemographic characteristics. Future studies are required in order to investigate the underlying mechanisms and confirm the causal relationship between COM and T2DM.

In conclusion, the present study reported a significantly increased risk of developing T2DM in COM patients. We also observed an increased risk of T2DM in young and wealthy COM patients. The findings of this study emphasize the importance of detecting T2DM in COM patients because medical delays and negligence may reduce the quality of medical care and increase healthcare costs. Therefore, in addition to providing adequate antimicrobial treatment and surgical intervention for COM patients, physicians should be aware of the possibility of developing T2DM in these patients, particularly in those who are young and economically advantaged.