Diabetes Is an Independent Predictor for Severe Osteoarthritis: Results from a longitudinal cohort study

The Bruneck cohort is a prospective population-based study on the epidemiology and pathogenesis of cardiovascular, neurologic, and musculoskeletal diseases (15–17), which has already been used to define the impact of type 2 diabetes on cardiovascular disease (18,19). This cohort is characterized by unique homogenous access to one public health care system, represented by the Bruneck Hospital, which is the only health care provider in the entire region. Thus, all medical records of each individual were collected and completely accessible in the same place, minimizing bias due to incomplete medical information.

The study population was recruited in 1990 (baseline) as a random sample (n = 1,000), stratified according to sex and age, of all inhabitants of Bruneck (125 women and 125 men in each of the fifth to eighth decades of age). A total of 927 men and women were enrolled and followed until 2010, with examinations conducted every 5 years (1990, 1995, 2000, 2005, and 2010) and participation rates >90%. Population mobility within the Bruneck area was low at 0.2% per year and migration to outside communities was a rare event (n = 4 during the 20 years of follow-up). We succeeded to trace all of these few individuals, who participated in the follow-up exam or provided access to their medical records.

Data on type 2 diabetes and arthroplasty were available in all 927 subjects, including 576 subjects who were alive in 2010 and 351 subjects who died during the 20 years of follow-up. The study protocol of the Bruneck Study was reviewed and approved by the ethics committees of Bolzano and Verona, and all study subjects gave their written informed consent.

At baseline and each 5-year follow-up, demographic and socioeconomic characteristics, comorbidities, and medication (including antidiabetic treatment, analgesic, and nonsteroidal anti-inflammatory drugs) were recorded. Assessment of comorbidities included cardiovascular disease (angina, myocardial infarction, heart failure, claudication, and stroke), cancer, chronic obstructive pulmonary disease (including asthma, bronchitis, and emphysema), arterial hypertension, chronic inflammatory disease, and neuropsychiatric disorders. Data were retrieved from 1) clinical history taking and medical record review; 2) complete physical and neurologic examination; and 3) measures such as blood pressure recording, electrocardiography, and carotid ultrasonography. Polyneuropathy was ascertained by a thorough clinical examination conducted by an experienced senior neurologist and graded as mild (sub- or preclinical disease), moderate (manifest sensory and/or motor deficits), or severe (disabling disease with severe impairment of gait).

Presence of type 2 diabetes was diagnosed according to the classic American Diabetes Association criteria (fasting glucose ≥7 mmol/L [126 mg/dL] or a clinical diagnosis of type 2 diabetes with ongoing antidiabetic treatment) (20). In a sensitivity analysis, the most recent diagnostic criteria were applied (fasting glucose ≥7 mmol/L [126 mg/dL] or a clinical diagnosis of type 2 diabetes with ongoing antidiabetic treatment or HbA_1c levels ≥6.5% or 2-h glucose level ≥11.1 mmol/L [200 mg/dL] after a standard 75-g glucose load) (21). The BMI was calculated as weight divided by height squared (kg/m²). Smoking status and alcohol consumption were recorded as detailed previously (15–17). Physical activity was recorded by composed score for work (three categories) and sports/leisure activities (0, ≤2, and >2 h per week) (15–17). Socioeconomic status was categorized on a three-category scale (low, medium, and high) based on information about occupational status and educational level of the person with the highest income in the household (15–17). Blood samples were drawn after an overnight fast and abstinence from smoking. Total cholesterol, HDL cholesterol, uric acid, creatinine, ferritin, and high-sensitivity C-reactive protein were all assessed by standard methods (15–17). LDL cholesterol was calculated with the Friedewald formula except for patients with a triglyceride level >400 mg/dL. Vascular cellular adhesion molecule (VCAM)-1 was detected by enzyme-linked immunosorbent assay (17).

Localization, date, and circumstances of total hip and knee arthroplasty occurring between 1990 and 2010 were carefully recorded using three sources of information: subject self-report, medical records of the Bruneck Hospital and general practitioners, and a standardized evaluation of all radiographs ever taken on study subjects. Arthroplasty following bone fractures (documented by radiographs) was not considered in the current analysis. All subjects with arthroplasty had radiographically documented OA of the hip or knee joint in at least two sequential radiographs (Kellgren-Lawrence score of at least one) before surgery and met the criteria for the diagnosis of OA (22,23). Two independent readers (R.L. and C.P.) unaware of the status of diabetes assessed the radiographs.

In the 2010 follow-up exam, a total of 576 individuals were still alive and had full information on arthroplasty and of these, 488 participated in fourth follow-up. The self-administered questionnaires of the Knee Injury and Osteoarthritis Outcome Score (KOOS) and the Western Ontario and McMaster University Osteoarthritis Index (WOMAC), both addressing the severity of symptoms (pain, stiffness, and joint function) associated with OA, were filled out by 347 men and women who did not differ from the overall group of 488 participants in any of the population characteristics assessed.

In order to evaluate the severity of knee OA, bilateral ultrasound (US) examination of the knee joints was performed by an independent experienced examiner blinded to clinical symptoms and presence of laboratory or radiographic abnormalities in 443 individuals (90.8%) (24–26). The US exam was performed by a single reader (C.P.) using an ESAOTE MyLab25 US system (Esaote Biomedica, Genoa, Italy) equipped with a broadband 4–13 MHz linear transducer for US examination. Both transverse and longitudinal scans of the supra-and infrapatellar region and the lateral and medial recesses were performed. Presence of synovitis (i.e., hypoechoic synovial hypertrophy and joint effusion according to the Outcome Measures in Rheumatology definition) in grayscale and power Doppler signal in the synovial membrane (settings: pulse repetition frequency 750 Hz, gain 40–50 dB) were recorded in each compartment. Presence of grayscale synovial hypertrophy and joint effusion as well as Doppler signal was recorded according to a semiquantitative scale (0 = normal; 1 = mild change; 2 = moderate change; 3 = severe change). The summed total score of inflammatory-related abnormalities (effusion, synovial hypertrophy, and power Doppler signal) was also calculated (range from 0–9 in each joint).

Data were analyzed using the statistical packages SPSS version 18.0 (SPSS) and STATA version 10.1 (StataCorp). For computation of intervention rates, person-years of follow-up for each participant were accrued from the 1990 baseline until arthroplasty, age 90 years, death, or 1 October 2010, whichever came first. Cox proportional hazard models were fitted to assess the association between baseline type 2 diabetes and arthroplasty due to severe OA during follow-up. The proportional hazard assumption was confirmed for type 2 diabetes by testing the interaction of type 2 diabetes with a function of survival time (Cox model with time-dependent covariates). When the exact date of arthroplasty was not available, we used the date of the first X ray showing the implant as a surrogate. This may be viewed as an adequate approximation in most individuals and, if at all, errs on the conservative side. The base model was unadjusted, whereas multivariable models were adjusted for established predictors of severe OA [age, sex, prior arthroplasty, BMI, and log_e-transformed level of soluble VCAM-1 (17); model 1] and, additionally, for social class, smoking, alcohol consumption, physical activity score, and levels of uric acid, creatinine, LDL cholesterol, log_e-transformed high-sensitivity C-reactive protein (hsCRP), and ferritin (model 2). Differential effects of type 2 diabetes on the probability of arthroplasty in subgroups according to age, sex, and BMI were tested by inclusion of appropriate interaction terms. Sensitivity analyses with an update of key variable (diabetes status and BMI) were fitted using Cox regression analysis with time-dependent covariates in order to control for switches in diabetes status and for weight gain during follow-up. Comparison of US measures and clinical score levels between subjects with and without type 2 diabetes (2010) was performed with the t test and χ² test (for trend in case of more than two categories) and, in a multivariable setting, with unconditional logistic regression analysis and general linear models. All reported P values were two-sided.

At baseline, 69 of the 927 participants of the Bruneck Study met the diagnostic criteria for type 2 diabetes. Mean HbA_1c at baseline was 7.2%. Oral antidiabetic drugs were used in 41 and insulin in 3 of the 69 participants. A total of 31 of the 69 patients with diabetes (44.9%) had clinical signs of a polyneuropathy at baseline (mild, 20; moderate, 11; severe, 0), and this proportion increased to 55.1% during follow-up (mild, 22; moderate, 16; severe, 0). No subjects had been affected by type 1 diabetes. Baseline characteristics of diabetic and nondiabetic subjects are summarized in Table 1.

Between 1990 and 2010, 13 arthroplasties due to severe symptomatic hip/knee OA were performed in diabetic subjects corresponding to an intervention rate (95% CI) of 17.7 (9.4–30.2) per 1,000 person-years. During the same period, 73 interventions were done in the nondiabetic group corresponding to an intervention rate (95% CI) of 5.3 (4.1–6.6) per 1,000 person-years. In unadjusted Cox models, the hazard ratio (HR) (95% CI) of joint failure for type 2 diabetes was 3.8 (2.1–6.8) (P < 0.001) (Table 2). Remarkably, type 2 diabetes remained a significant predictor of arthroplasty after controlling for a variety of other putative risk factors for OA including age and BMI (HR [95% CI], 2.1 [1.1–3.8]; P = 0.023). To account for weight gain during follow-up (e.g., related to insulin treatment), BMI was updated every 5 years during follow-up from 1990–2010 (Table 2). This sensitivity analysis yielded similar risk estimates for type 2 diabetes (HR [95% CI], 2.2 [1.2–4.1]; P = 0.013). Several further sensitivity analyses with an update of diabetes status during follow-up or application of the most recent American Diabetes association definition of diabetes (21) yielded very similar results (Table 2). Findings were also consistent when additionally adjusting the analysis for lifestyle factors including smoking and physical activity and serum parameters such as CRP (HR [95% CI], 2.1 [1.1–4.0]; P = 0.026) (Table 2). Moreover, results were similar in men and women and subgroups according age and BMI (P values for interaction: 0.421, 0.844, and 0.192). Of note, severity of OA, as assessed by the Kellgren-Lawrence score (range 0–10 points) in the last radiograph prior to arthroplasty, was similar in subjects with and without type 2 diabetes (mean ± SD: 5.5 ± 1.7 vs. 5.8 ± 1.6 points, both values resembling severe radiographic Kellgren-Lawrence stage 3 OA; P = 0.84), as was the mean age at surgery (74.1 vs. 72.6 years; P = 0.35).

In the cumulative hazard plot (Fig. 1), the lines for diabetic and nondiabetic subjects diverged over time, indicating relevance of disease duration for the manifestation of severe OA. Mean time between diagnosis of type 2 diabetes and joint failure necessitating surgical intervention was 10.7 years in our cohort. In line, type 2 diabetes disease duration was significantly higher in diabetic patients with arthroplasty than in those without arthroplasty (Fig. 1).

Subjects with impaired fasting glucose at baseline (1990, n = 86) showed a moderately elevated risk of arthroplasty due to severe OA (unadjusted HR [95% CI], 2.0 [1.1–3.7]; P = 0.030), which lost significance in the multivariable model (HR [95% CI], 1.6 [0.9–3.0]; P = 0.145).

On a cross-section (2010), patients with prevalent diabetes (n = 74) were more likely to have received replacement of knee and hip joint (implanted for alleviation of OA symptoms) than nondiabetic individuals (knee arthroplasty, intervention proportion 10.8 vs. 3.3%, P = 0.006; hip arthroplasty, intervention proportion 9.5 vs. 5.0%, P = 0.167). To further elaborate the association between type 2 diabetes and OA, a number of ancillary cross-sectional analyses were performed. Evaluation of the clinical symptoms of OA by total WOMAC score, a validated patient-related measure of signs and symptoms of OA, showed significantly more severe OA symptoms in subjects with type 2 diabetes than in controls (Table 3). Results were consistent in unadjusted (P = 0.034) and fully adjusted models (P = 0.030). Moreover, WOMAC subscales for joint pain and function were lower (i.e., more pathologic in type 2 diabetic patients than controls) (P values from multivariable analyses 0.014 and 0.004; Table 3). Finally, results were replicated when applying another validated score for the quantification of clinical disease activity in OA, the KOOS questionnaire (pain score, P value from multivariable analyses 0.001; Table 3).

We performed examinations of knee joints by US in 443 individuals of the Bruneck Study population in 2010 (90.8%), searching for signs of inflammation including synovitis and joint effusion. Subjects without US data (n = 55) were more likely to be female and diabetic but otherwise similar to the entire population. Bilateral joint effusion was found in 51.1% of subjects with type 2 diabetes but only in 30.9% of the subjects without type 2 diabetes (Table 3). In accordance, uni- and bilateral signs of synovitis in the knee joints were significantly more frequent in diabetic than nondiabetic subjects (P value from multivariable analyses 0.004; Table 3). Data were consistent when focusing on both synovitis and effusion in a single model (P value from multivariable analyses 0.003; Table 3). Finally, severity of synovitis as quantified by the synovitis score was more pronounced among patients with type 2 diabetes (Table 2).

To our best knowledge, this is the first time that it was shown that diabetes can be considered as an independent predictor of severe OA necessitating joint arthroplasty. OA is a major challenge to health in the industrialized world. 1) Its prevalence is high and further increases due to ageing of the populations. 2) Furthermore, its impact on quality of life is substantial due to an impairment of locomotor function and decrease in functional capacities. 3) Finally, its conservative treatment options are limited, and surgical replacement of damaged joints may be necessary in severe cases. Therefore, definition of risk predictors for OA is of key importance, particularly if these risk factors can be modified by either lifestyle changes or medications. In this study, we show that long-standing type 2 diabetes is independently associated with advanced OA of knee and hip joints. This finding adds to the yet short list of risk predictors for OA established in prospective evaluations (5,6,17). After controlling the analysis for age, BMI, and other potential confounders, type 2 diabetes comprised a twofold risk of severe OA necessitating arthroplasty.

The population-based design is a strength of this study (27,28), as well as its long-term and complete follow-up and the use of joint failure necessitating arthroplasty as a hard end point. A further strength of our study is that the link between type 2 diabetes and OA was consistent when using three distinct approaches of OA ascertainment as follows: 1) joint failure as determined by arthroplasty; 2) signs and symptoms of OA as quantified by the WOMAC and KOOS scores; and 3) severity of joint changes as determined by musculoskeletal US.

The moderate number of arthroplasties (<100) can be considered as a limitation of study. However, this is not surprising because arthroplasty accomplishes the most severe disease phenotype of OA, which is less prevalent. Moreover, results from our study launched in a small town, in which accessibility to health resources is very high, are not necessarily representative for other Western populations and await confirmation in independent studies. Also, we cannot exclude different decision-making for arthroplasty in diabetic versus nondiabetic subjects. Subjects with diabetes may receive more medical attention than nondiabetic individuals, leading to increased awareness of or even overtreatment of OA. Moreover, fear of deterioration of the general medical condition in diabetic subjects could, at least theoretically, lead to earlier interventions in this group. In contrast, studies showing a higher risk of complications and of revisions of arthroplasty in diabetic subjects could translate into a more conservative indication for surgical interventions resulting in an underestimation rather than overestimation of the effect size of diabetes in our study (29). Actually, in the current study, radiographic severity of OA prior to surgery (Kellgren-Lawrence score) and the age at arthroplasty both were identical in subjects with and without type 2 diabetes, suggesting no major differences in the indication for interventions in these groups.

The link between obesity and OA is well-established and particularly supported by longitudinal data showing that obesity entails a higher risk for developing severe OA (6). Traditionally, mechanical factors have been considered to explain more rapid joint degeneration in obese individuals. However, this concept has been challenged by the link between obesity and OA in non–weight-bearing joints, suggesting that metabolic changes directly enhance the risk of OA. Despite evidence from cross-sectional studies that OA is linked to the metabolic syndrome and higher blood glucose levels (10–12), no study addressed whether diabetes is independently linked to OA and in particular whether such a link is independent of body weight. Our study is the first to show that type 2 diabetes predicts severe OA independent of age, sex, and BMI and provides both longitudinal and cross-sectional evidence for an independent association between type 2 diabetes and OA.

Moreover, several lines of experimental evidence may help to explain the link between type 2 diabetes and OA: 1) chondrocytes express the GLUT/SLC2A, and high blood glucose levels shift the synthesis pattern of chondrocytes from type II collagen to reactive oxygen species, potentially mediating cartilage destruction (30–32). 2) Advanced glycosylation end products elicited by sustained hyperglycemia can stimulate chondrocyte expression of proinflammatory and prodegenerative proteins via the receptor for advanced glycosylation end products (33–35). In this context, it is worth mentioning that elevated circulating levels of interleukin-6 and tumor necrosis factor-α have recently been demonstrated in patients with knee OA (36). Moreover, our data from knee US showed that signs of inflammation are more frequent and severe in the joints of diabetic subjects than nondiabetic individuals. 3) Finally, sensory and motor polyneuropathy is a common complication of type 2 diabetes and may increase the risk of OA by muscle weakness and loss of vibratory sense (37) with subsequent inappropriate use of the affected joint due to altered mechanical stress forces. In this context, it is noteworthy that the pain subscales of the WOMAC and KOOS scores exhibit particularly pronounced associations with type 2 diabetes (Table 2). Sensory polyneuropathy obviously does not offset the high burden of pain in OA among diabetic subjects, rendering this potential link even more robust.

Our study fosters the concept that OA is part of the metabolic syndrome. This notion shifts the traditional perception of OA as a degenerative joint disease based on continuous mechanical overload to a metabolic etiopathogenesis. The link between OA and type 2 diabetes suggests that alterations in glucose metabolism directly affect joint integrity independently of body weight and creates room for hope that adequate control of glucose metabolism hampers development of OA. Our findings may thus add a novel, modifiable, and highly prevalent risk condition for the development of OA.

In summary, our data show that type 2 diabetes is a strong predictor for the development of severe OA. This finding is independent of age and BMI and suggests that longstanding diabetes per se is detrimental for knee and hip joints, leading to progressive destruction and joint failure.

This study was supported by the “Bundesministerium für Bildung und Forschung” (project ANKYLOSS), the MASTERSWITCH, ADIPOA, and BTCure projects of the European Union, the “Pustertaler Verein zur Prävention von Herz-und Hirngefässerkrankungen,” the “Gesundheitsbezirk Bruneck,” and the “Assessorat für Gesundheit und Sozialwesen,” Bolzano.

No potential conflicts of interest relevant to this article were reported.

G.S. wrote the manuscript, researched data, contributed to discussion, and reviewed and edited the manuscript. A.K., C.P., E.S., J.Z., R.L., and F.A. researched data. A.I. reviewed and edited the manuscript. F.B., M.-A.D., and J.W. contributed to discussion. S.K. researched data, contributed to discussion, and reviewed and edited the manuscript. G.S. and S.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.