Heterogeneity of the patients in the studies (for example, with regard to whether they were receiving insulin), differences in measurement techniques and regions of interest, the use of an inappropriate group of persons for comparison, and the absence of control for several possible confounders may explain the dissimilar results. Additionally, several previous studies were restricted to patients with previously diagnosed non-insulin-dependent diabetes mellitus. To address the effect of non-insulin-dependent diabetes mellitus on bone density, studies should include persons with untreated non-insulin-dependent diabetes mellitus; studies that exclude them may address the effect of antidiabetic therapy rather than the effect of non-insulin-dependent diabetes mellitus. In addition, we know of no study that has linked the effect of non-insulin-dependent diabetes mellitus on bone mineral density to the occurrence of fractures.

We investigated the associations among non-insulin-dependent diabetes mellitus, bone mineral density, and fractures in 5931 unselected men and women, aged 55 years or more, who were examined in the population-based Rotterdam Study.

Methods

Top Methods Results Discussion Author & Article Info References

Study Population

The Rotterdam Study is a prospective follow-up study of persons aged 55 years or more; its intent is to investigate the incidence of and the risk factors for chronic disabling diseases. Its rationale and design have been described previously [13]. All 10 275 inhabitants of a district in Rotterdam who were aged 55 years or more were invited to participate in the study, which consisted of an initial home interview by a trained research assistant and a series of medical examinations made during two visits to the research center. The study was approved by the Medical Ethics Committee of the Erasmus University, and written informed consent was obtained from all participants. For the present study, institutionalized persons (n = 1114) were ineligible for examination at the research center. Interview data were collected for 77% of the remaining eligible persons; the overall response rate for the Rotterdam Study was 78%. Bone mineral density was measured in 2481 men and 3450 women; scan data were unavailable for 563 persons. Most of these persons took part in a pilot study (n = 338) in which no scans were obtained; maintenance procedures accounted for the remaining missing scan data.

Measurements

During the home interview, a medical history was obtained using a questionnaire administered by an interviewer; medication use was verified by examining pills. To estimate the frequency of fractures, participants were asked about the number and types of fractures they had had during the preceding 5 years.

All participants at the research center, except for those using antidiabetic medication, were given a 37.5% oral glucose solution (75 g of glucose) while in a nonfasting state. Venous glucose levels were then measured before and after 2 hours; non-insulin-dependent diabetes mellitus was diagnosed if at least one measurement was 11.1 mmol/L or more or if the participant used antidiabetic medication.

Bone mineral density was measured at the lumbar spine and proximal femur using a Lunar DPX-L densitometer (Lunar Corp., Madison, Wisconsin). Standard positioning was used for anterior-posterior scans of the lumbar spine and proximal femur. The right proximal femur was scanned unless the participant had a history of prosthesis implantation, in which case the left side was scanned. The in vivo coefficient of variation for bone mineral density, calculated from 12 participants who were each scanned twice on the same day, was 3.2% at the femoral neck, 3.1% at the Ward triangle, 2.5% at the greater trochanter, and 0.9% at the lumbar spine. Quality assurance, including calibration, was done routinely every morning using the standard provided by the manufacturer.

When height and weight were measured, participants wore indoor clothing and no shoes. Body mass index (weight in kg/height in m²) was calculated as a measure of obesity. The waist-to-hip ratio was used as an indicator of central obesity. Waist circumference was measured midway between the lower rib margin and the iliac crest; hip circumference was measured at the point of maximum circumference over the buttocks with the tape held horizontally [14].

Osteoarthritis was scored using standard radiographs of the hips in a random subsample of 1136 men and 1697 women, according to the method proposed by Kellgren and colleagues [15]. Participants were classified as having osteoarthritis if they had a Kellgren score of at least 2 at the right hip or if they had a hip prosthesis due to osteoarthritis.

For current and past smokers, the number of pack-years was calculated as the average number of packs of cigarettes smoked per day multiplied by the total years of smoking. Falling was recorded as never, less than once per month, less than once per week but more than once per month, and more than once per week. In the analysis, however, falling was regarded as a dichotomous variable. Impairment in activities of daily living was assessed using a questionnaire adapted from the Stanford Health Assessment Questionnaire [16]. Because impaired function of the lower extremities in particular will influence bone mineral density, a disability index was calculated for the lower limbs by averaging scores for arising, walking, bending, and getting in and out of a car. Each of these four categories was scored on a scale ranging from 0 to 3, where 0 indicates no impairment and 3 indicates inability to do the activity. Disability was then categorized as absent (0), mild (>0 to 1), or severe (>1).

Data Analysis

Men and women were analyzed separately. Patients newly diagnosed with non-insulin-dependent diabetes mellitus and those who were already being treated for it were analyzed in combination and separately. Furthermore, analyses were done with and without patients receiving insulin. A chi-square test was done for differences in proportions. A Student t-test was used for continuous variables. Multiple linear regression analysis was used to evaluate the association between non-insulin-dependent diabetes mellitus and bone mineral density and to control for possible confounders. Effect modification by age was studied by including interaction terms for age and the presence of non-insulin-dependent diabetes mellitus in the model.

To examine selective nonresponse to the research-center visit of participants with diabetes, we calculated the age-adjusted prevalence of antidiabetic medication use in participants who were interviewed only, and compared this prevalence with that in participants who had both the interview and the examination at the research center. There appeared to be no selective nonresponse in women; the age-adjusted prevalence of current antidiabetic medication use was similar for those with and without a visit to the center (5.0% compared with 4.9%; P = 0.9). In men, however, this prevalence was significantly higher in participants who did not visit the center (4.2% compared with 10.0%; P = 0.0001).

Logistic regression was used to study the association between non-insulin-dependent diabetes mellitus and a fracture (regardless of cause) in the preceding 5 years. Vertebral fractures were excluded from the analyses because participants were often unaware of them. All other self-reported fractures were included in the analyses. The association between diabetes and fracture frequency was expressed as an odds ratio with a 95% CI. Two-sided P values were calculated for all analyses. Analyses of variance and covariance were used to estimate multiple adjusted mean values.

Results

Top Methods Results Discussion Author & Article Info References

Among the 2481 men and 3450 women with bone mineral density measurements, 243 men and 335 women were classified as having non-insulin-dependent diabetes mellitus. Ninety-four men and 140 women were already being treated with antidiabetic medication. Table 1 shows the characteristics of the men and women with and without non-insulin-dependent diabetes mellitus. As expected, persons who had this condition were older than those who did not have it. The body mass index of women with non-insulin-dependent diabetes mellitus was substantially higher than that of women without diabetes; this was not true for men. The group of participants with non-insulin-dependent diabetes mellitus included more current users of loop diuretics. Furthermore, the group of women with non-insulin-dependent diabetes mellitus included a higher percentage of current thiazide users. Serum creatinine levels were substantially higher in men with diabetes than in men without diabetes; this was not true for women. Both men and women with non-insulin-dependent diabetes mellitus reported a higher frequency of impairment in function of the lower extremities. Although the difference was not significant, men with non-insulin-dependent diabetes mellitus tended to fall more often than men without diabetes.

Bone mineral density in men with non-insulin-dependent diabetes mellitus was higher at all four sites than it was in men without this condition; this was true both for newly diagnosed patients and for those already being treated. For instance, bone mineral density at the femoral neck was 0.873 g/cm² for those without diabetes, 0.898 g/cm² for newly diagnosed persons, and 0.908 g/cm² for those persons already being treated. The bone mineral density measured at the Ward triangle was the same for women receiving oral antidiabetic treatment and women without diabetes, but women receiving treatment did have higher values at the three other sites. Women with newly diagnosed non-insulin-dependent diabetes mellitus had higher values at all four sites. The bone mineral density of patients treated with insulin did not differ substantially from that of controls (femoral neck bone mineral density: 0.878 g/cm² in men and 0.796 g/cm² in women), but this group comprised only 25 men and 31 women.

Figure 1 shows the difference in the percentage of persons with at least one reported nonvertebral fracture in the preceding 5 years between those with and those without non-insulin-dependent diabetes mellitus. Twenty-two men and 38 women with non-insulin-dependent diabetes mellitus and 222 men and 493 women without non-insulin-dependent diabetes mellitus reported having had at least one fracture in the preceding 5 years. For men, ankle and foot fractures were reported most frequently (33%), followed by fractures of the wrist and forearm (32%). For women, wrist and forearm fractures were reported most frequently (60%), followed by ankle and foot fractures (27%). After adjustment for age and body mass index, women with non-insulin-dependent diabetes mellitus had a lower frequency of fractures (odds ratio, 0.63; CI, 0.44 to 0.90). Results were similar for patients with newly diagnosed non-insulin-dependent diabetes mellitus and for those already being treated (odds ratio, 0.62; CI, 0.35 to 1.10). Adding bone mineral density to the model did not substantially influence the odds ratio. Although the percentage of men who reported a nonvertebral fracture was slightly lower for those with than for those without non-insulin-dependent diabetes mellitus, after adjustment, no significant difference remained (odds ratio, 0.96; CI, 0.60 to 1.52).

View larger version (12K): [in this window] [in a new window]   Figure 1. Difference in the percentage of patients reporting at least one nonvertebral fracture in the preceding 5 years between those with and without non-insulin-dependent diabetes mellitus, and the 95% CI for the difference for each sex.

Discussion

Top Methods Results Discussion Author & Article Info References

In this large epidemiologic study, we were unable to do a fasting oral glucose tolerance test. Therefore, in accordance with the guidelines of the World Health Organization Study Group on diabetes mellitus for epidemiologic studies, we used the 2-hours value [17]. The prevalence values obtained for non-insulin-dependent diabetes mellitus are consistent with what is known about prevalence in the Netherlands [18]. Nevertheless, it remains possible that some patients were misclassified, most likely as false negatives. If this occurred, however, it would have diluted the association between non-insulin-dependent diabetes mellitus and bone mineral density rather than inducing a difference.

We found a higher percentage of persons known to have diabetes among the men who did not visit the research center than among those who did. It is possible that patients with diabetes who did not respond had poorer health and were less active than responders and therefore had a lower bone mineral density. However, it seems unlikely that this has biased the results; bias would have occurred only if inactive patients with diabetes were more likely than inactive patients without diabetes to be nonresponders. More important is the fact that selective nonresponse is unlikely in patients newly diagnosed with diabetes, and we saw similar results when we restricted the analyses to this group. For the same reason, it is unlikely that our observations can be explained by increased mortality in patients with osteoporosis and diabetes.

Aortic calcifications and osteophytes are known to cause a spuriously elevated bone mineral density measurement at the lumbar spine. Aortic calcifications in particular may be present more often in patients with diabetes than in those without it. However, this would not explain the elevated bone mineral density at the proximal femur.

The data on fractures in the preceding 5 years were reported by the participants. Self-report has been shown to be an accurate and reliable method for calculating fracture rates, especially for major osteoporotic fractures, including those of the hip, wrist, and humerus [19, 20]. Recall over time periods longer than 1 year may lead to a degree of misclassification, but this misclassification is unlikely to be systematically different for persons with and without non-insulin-dependent diabetes mellitus. Furthermore, we found frequencies of fractures similar to those seen in previous studies in which the fractures were verified using radiographs [21]. The lower frequency of fractures in women with non-insulin-dependent diabetes mellitus, seen in our study, accords with the findings of Heath and colleagues [22].

Our results differ from those of several previous studies. For instance, Levin and colleagues [7] found a lower bone mineral density in patients with non-insulin-dependent diabetes mellitus. However, they may have included patients with insulin-dependent diabetes mellitus, given the age distribution and the proportion of persons receiving insulin. Furthermore, their study was not population based, and they were unable to estimate the effect of non-insulin-dependent diabetes on bone mineral density in undiagnosed persons. Other investigators had findings that agree with our own [4, 5]. In particular, Barrett-Connor and colleagues [2] did a similar but smaller population-based study of elderly persons and recently reported finding a higher bone mass in women with non-insulin-dependent diabetes mellitus, but not in men. They suggested that this apparent difference may be explained by the greater androgenicity reported in women with hyperglycemic and hyperinsulinemic conditions. However, the similar elevations of bone mineral density that we observed in both men and women with non-insulin-dependent diabetes mellitus makes this hypothesis less likely.

Rather, our findings support an anabolic effect of insulin on bone tissue, as suggested by Weinstock and colleagues [11]. Non-insulin-dependent diabetes mellitus is preceded by a period of insulin resistance that causes hyperinsulinemia before the onset of diabetes [23]. It has been suggested that this insulin resistance is restricted to the effect of insulin on glucose transport [24]. Hyperinsulinemia will still lead to stimulation of the unopposed mitogenic and anabolic actions of insulin. The already elevated bone mineral density found in persons with newly diagnosed diabetes indicates that this mechanism is operating. Hyperplasia of the theca folliculi and the vascular endothelial proliferation that develop in insulin-resistant patients are other examples of the stimulating effect of hyperinsulinemia [24, 25]. Insulin may act either directly on bone or by binding to the receptor of insulin-like growth factor I [26]. Some [25, 27], but not all [28, 29], investigators have found a positive association between circulating insulin levels and bone mineral density. Hyperinsulinemia may also lead to an increase in bone mineral density through its negative effect on sex-hormone-binding globulin [30-33]. Van Hemert and colleagues [34] found a strong negative correlation between sex-hormone-binding globulin levels and bone mineral density in postmenopausal women. Lower levels of sex-hormone-binding globulin may lead to higher free serum estradiol and testosterone levels. Subsequently, the higher free sex-hormone levels could protect patients from the age-related bone loss that occurs in both men and women [35, 36]. Nevertheless, we can only speculate on the mechanism underlying the elevated bone mineral density in patients with non-insulin-dependent diabetes mellitus.

Caution is needed when trying to explain the lower fracture frequency in women with non-insulin-dependent diabetes mellitus, particularly because—at least in newly diagnosed patients with diabetes—fractures occurred before non-insulin-dependent diabetes mellitus was apparent. Bone mineral density is an important determinant of risk for fracture [37], and the higher bone mineral density in women with non-insulin-dependent diabetes mellitus may cause a lower reported fracture frequency. Nevertheless, the fact that the "protective effect" remained after correction was made for bone mineral density is at least curious. Alternatively, because women with non-insulin-dependent diabetes mellitus are substantially more obese than are women with normal glucose tolerance, one could attribute this "protective effect" to an energy-absorbing effect of their fat tissue [38]. However, the effect also persisted after adjustment was made for body mass index. Whether the absence of this protection in men is due to selection or to the fact that diabetic complications like neuropathy and retinopathy lead to more frequent falling and therefore to an increase in fracture frequency needs to be elucidated.

We present evidence of increased bone mineral density in patients with non-insulin-dependent diabetes mellitus. In women, non-insulin-dependent diabetes mellitus appears to be accompanied by a lower frequency of nonvertebral fractures.