OBJECTIVE—Hypoglycemic treatments could modulate the risk for fractures in many ways. Most studies have not explored the effect on the incidence of bone fractures of individual oral hypoglycemic agents, rather all oral treatments as a whole. The aim of this case-control study, nested within a retrospective cohort, is the assessment of the risk for bone fractures associated with exposure to insulin or different oral hypoglycemic agents.
RESEARCH DESIGN AND METHODS—A case-control study nested within a cohort of 1,945 diabetic outpatients with a follow-up of 4.1 ± 2.3 years was performed, comparing 83 case subjects of bone fractures and 249 control subjects matched for age, sex, duration of diabetes, BMI, A1C, comorbidity, smoking, and alcohol abuse. Exposure to hypoglycemic drugs during the 10 years preceding the event (or matching index date) was assessed.
RESULTS—In a model including treatment with insulin secretagogues metformin and insulin for at least 36 months during the previous 10 years, no significant association was observed between bone fractures and medications. In an alternative model considering treatments at the time of fracture, insulin treatment was significantly associated with bone fractures in men (OR 3.20 [95% CI 1.32–7.74]) but not in women (1.41 [0.73–2.73]).
CONCLUSIONS—Insulin-sensitizing treatment with metformin is not associated with a higher incidence of bone fractures, suggesting that the negative effect of thiazolidinediones is due to a specific action on bone metabolism rather a reduction of insulinemia. Conversely, current treatment with insulin increases the risk of fractures; at the same time, exposure to this agent in the longer term does not appear to affect bone frailty.
Type 2 diabetes is associated with an increased risk for bone fractures (1–10) with no reduction of bone density (4,6,7,10–13); several factors, including frequency of falls, diabetes complications, and comorbidities, could contribute to the higher risk of fractures (5,7,10,14).
Hypoglycemic treatments could modulate the risk for fractures in many ways. A higher incidence of fractures has been reported in insulin-treated patients in comparison with non–insulin-treated type 2 diabetic individuals (1,2), although some studies disagree (15). This could be due to a higher risk of hypoglycemic episodes and falls (5,6). However, a higher prevalence of diabetes complications and comorbid conditions could also contribute to the apparent increase of the incidence of fractures in insulin-treated patients; in fact, most available studies did not provide adjustments for comorbidities (1,4,6,15).
Recent evidence from an epidemiological study (16) and an exploratory clinical trial (17) suggests that the use of the insulin-sensitizing agents thiazolidinediones is associated with a decrease in bone density in postmenopausal women; similar results have been reported by an observational retrospective study exploring variations of bone density in older men with type 2 diabetes (18). This could explain the increased incidence of bone fractures in patients treated with rosiglitazone in a large randomized trial (19). Manufacturers of pioglitazone have also sent a letter to health professionals to inform them of the risk of fracture associated with this drug. Thiazolidinediones, which act as stimulators of the nuclear receptor peroxisome proliferator–activated receptor (PPAR)-γ, could reduce bone density through the inhibition of osteoblast differentiation and activity; in fact, PPAR-γ activation induces the differentiation of multipotent mesenchimal stem cells into adipocytes, rather than osteoblasts, and increases osteoblast apoptosis (20,21). On the other hand, the insulin-sensitizing effect of thiazolidinediones reduces circulating insulin levels and therefore the insulin anabolic effect on the bone (22). If this second hypothesis should be true, other insulin-sensitizing agents, such as metformin, could also be associated with an increased risk of bone fractures. Most available epidemiological studies have not explored the effect of individual oral hypoglycemic agents on the incidence of bone fractures but rather considered all oral treatments for diabetes only as a whole (2–4,6). The aim of this case-control study, nested within a retrospective cohort, is the assessment of the risk for bone fractures associated with exposure to insulin or different oral hypoglycemic agents.
RESEARCH DESIGN AND METHODS
Data collection
The study was performed on a consecutive series of 1,945 diabetic outpatients living within the region of Tuscany and attending, for the first time, the Diabetes Outpatient Clinics of the University of Florence between 1 January 1998 and 31 December 2004. Demographic and clinical data, including history of hypoglycemic medication, self-reported smoking habits, and alcohol intake were collected. Alcohol consumption of more than two drinks per day was used as the cutoff to define alcohol abuse. At first visit, following a standard procedure of the clinic, all patients underwent a physical examination, including measurement of weight, height, and blood pressure, following World Health Organization recommendations (23,24). A fasting blood sample was used for determining A1C (HPLC, upper normal limit 6.2%; Menarini-Diagnostici, Florence, Italy) and creatinine (with an automated method) (Aeroset; Abbott Laboratories). Comorbidity was assessed through the calculation of Charlson's comorbidity score, which includes diabetes and its complications, cardiovascular disease, chronic skin ulcers, renal insufficiency, liver diseases, chronic obstructive pulmonary disease, malignancies, arthritis/arthrosis, HIV infections (25).
Exposure to hypoglycemic drugs for 10 years before the event in case subjects and before the matching index dates in control subjects was assessed. Drug exposure was obtained from clinical records of the outpatient clinic. These records contain self-reported history of hypoglycemic treatment before the first contact with the clinic and all drug prescriptions during follow-up. If the last available visit occurred over 3 months before the event (or the matching index date), a telephone contact with the patients or their relatives was attempted to collect further information on subsequent drug use. If no such information was obtained, the patient was assumed to have continued the last available hypoglycemic therapy.
Identification of case and control subjects
Newly diagnosed cases of bone fracture from enrolment up to 31 December 2005 were identified either through diagnosis at hospital admissions or causes of death, as ICD-9 codes 800–829. Information on hospital admission was obtained through the regional hospital discharge system, which contains ICD codes of current diagnoses. Case finding was therefore performed on 1,945 patients (1,102 and 843 women and men, respectively) aged 63.9 ± 12.8 years with a mean duration of diabetes of 10.7 ± 10.5 years.
Incident cases of bone fracture (n = 83) were compared with control cases selected from the same cohort with a ratio of 1:3. For each case, the first three following patients within the same series, of the same sex, age (±2 years), duration of diabetes (±2 years), A1C (±1%), BMI (±2 kg/m2), Charlson's comorbidity score (±1 point), smoking status (current/former/never smoker), and alcohol abuse (yes/no) were taken as control cases.
Statistical analysis
Unpaired Student's t test and Mann-Whitney U test were used to compare continuous variables whenever appropriate. The χ2 test was used for between-group comparisons of categorical variables, computing odds ratios (ORs) with 95% CIs. Conditional logistic regression was used for multivariate analysis in order to adjust for concomitant hypoglycemic treatments. All analyses were carried out with SPSS 12.0.1 statistical package, and a P < 0.05 was considered statistically significant.
RESULTS—
Mean duration of follow-up in the reference cohort was 4.1 ± 2.3 years, during which 83 new cases of bone fractures were identified, with an incidence of 1.03 per 100 persons-years. Of the observed fractures, 46 (55%), 13 (16%), and 9 (11%) were of lower limbs, upper limbs, and spine. The five most frequent fracture sites were femur (n = 31), vertebral column (n = 9), humerus (n = 7), pelvis (n = 7), and skull (n = 4).
The characteristics of case and control subjects, at the time of their first visit to the outpatient clinic, are summarized in Table 1. No significant difference was observed between the two groups.
Follow-up obtained from clinical records was incomplete in 8 case and 16 control subjects. Information about treatment was collected through a telephone contact in most subjects. In three (3.6%) case subjects and eight (3.2.%) control subjects, hypoglycemic treatment after the last contact with the clinic could not be assessed; those patients were assumed to have continued the last available hypoglycemic therapy. Assessment of drug exposure was obtained through self-report and drug prescription for 1,135 and 1,654 patient-years, respectively; for the further 521 patient-years occurring before the diagnosis of diabetes (in patients with a duration of diabetes <10 years at the index date), exposure to hypoglycemic drugs was assumed to be none.
When considering hypoglycemic treatments during the previous 10 years, a lower proportion of case subjects had been exposed to insulin secretagogues for over 36 months in comparison with control subjects (Table 2). Eighteen case (21.7%) and 63 control (25.3%) subjects did not receive any hypoglycemic treatment during the 10 years before the index date.
At the time of fracture, or the corresponding index date in control subjects, treatment with insulin was significantly more frequent in case subjects. No other significant differences were observed between case and control subjects (Table 3).
Considering that a large proportion of patients received more than one hypoglycemic treatment at the same time, a multivariate analysis was performed in order to assess the effect of each treatment after adjusting for concomitant hypoglycemic medications. In a model including treatment with insulin secretagogues, metformin, and insulin for at least 36 months during the previous 10 years, no significant association was observed between bone fractures and medications (Fig. 1). Similar results were obtained when the two sexes were analyzed separately (data not shown).
In an alternative model considering treatments at the time of fracture or at the corresponding index date, insulin treatment was significantly associated with bone fractures (Fig. 1). Considering the two sexes separately, this association was evident in men (adjusted OR 3.20 [95% CI 1.32–7.74]; P = 0.010) but not in women (1.41 [0.73–2.73]).
CONCLUSIONS—
Treatment with the insulin-sensitizing drugs thiazolidinediones has been shown to reduce bone density, increasing the risk for fractures in type 2 diabetic patients (16,17,19,26). The present study does not add any new data on the incidence of fractures associated with thiazolidinedione treatment; in fact, due to the delay in marketing of these drugs, which weren't approved in our country until September 2002, and with many limitations to the prescription, very few patients in the present sample received thiazolidinediones; therefore separate analyses for those agents were not possible.
The effect of thiazolidinediones on bone fractures could be due to either a specific inhibition of osteogenesis (20,21) or to a reduction of the anabolic action of insulin on the bone (22). This latter hypothesis is not supported by the results of the present study. In fact, no association was observed between treatment with the insulin-sensitizing drug metformin and incident bone fractures in type 2 diabetic patients. A protective effect of metformin with respect to bone fractures, suggested by a previous study (15), was not confirmed by our data. However, considering that the OR associated with metformin is similar to that previously reported in a larger investigation (15), the lack of a statistically significant effect of metformin in the present study could be due to an insufficient sample size. Furthermore, it should be considered that treatment with metformin has several relevant contraindications, such as renal insufficiency, severe liver disease, and heart failure. As a consequence, patients receiving metformin could show a lower incidence of bone fractures as a result of a lower comorbidity (5–7,10,14). The design of the present study, with a careful match between case and control subjects, overcomes such limitation.
Treatment with insulin secretagogues for at least 36 months during the previous 10 years was less frequent in case than control subjects, suggesting a possible protective effect of those drugs, which is concordant with previous findings of a large population-based study (15). However, this difference did not retain statistical significance after adjusting for concomitant hypoglycemic medications. It is conceivable that the apparent protective effect of secretagogues could be due to a lower proportion of insulin-treated patients among those receiving those oral drugs.
Most available studies report a higher incidence of bone fractures in insulin-treated patients, in comparison with non–insulin-treated type 2 diabetic individuals (1,2). In our study, the difference between case and control subjects in the proportion of patients receiving long-term insulin treatment did not reach statistical significance, even after adjusting for concomitant hypoglycemic medications. Insulin-treated patients usually show a longer duration of diabetes and a higher prevalence of diabetes complications and comorbid conditions; it is possible that some studies, which did not provide adjustments for such confounders (1,2), could have overestimated the negative impact of insulin treatment (5–7,10,14).
At the same time, treatment with insulin at the index date showed a significant association with bone fractures, maintained after adjusting for concomitant hypoglycemic medications. These results are consistent with the hypothesis that insulin could increase the risk of fractures through falls due to hypoglycemic episodes (5,6), without negative effects on bone metabolism (22). Interestingly, the effect of current insulin treatment on bone fractures was evident in men, who show a lower incidence of spontaneous fractures due to osteoporosis, and not in women.
Some limitations of the present study should be recognized. The diagnosis of bone fractures was obtained through coded diagnoses at hospital discharge; as a consequence, some events receiving no hospital-based care could have remained unnoticed. The distinction between spontaneous, traumatic, and indeterminate fractures, although available, was not considered because of its uncertain reliability. Furthermore, no data on bone density and on the frequency of falls were collected, so that considerations about mechanisms underlying associations between bone fractures and medications remain speculative. Another relevant limitation is represented by the fact that treatments affecting bone metabolism (such as biphosphonates, vitamin D, and calcium supplementation) were not considered. However, the hypothesis of a different use of such treatments in patients receiving different hypoglycemic medications appears to be remote. Finally, the possibility of misclassification of treatments is possible. In fact, it is possible that some patients underreported medication during previous years, while drug prescription could overestimate actual drug intake, due to lack of compliance.
In conclusion, insulin-sensitizing treatment with metformin is not associated with a higher incidence of bone fractures, suggesting that the negative effect of thiazolidinediones is due to a specific action on bone metabolism, rather than to reduction of insulinemia. Conversely, current treatment with insulin increases the risk of fractures; at the same time, exposure to this agent in the longer term does not appear to affect bone frailty. Bone fractures deserve to be considered among treatment outcomes for the choice of hypoglycemic medication, particularly in older patients with type 2 diabetes.
Adjusted ORs, with 95% CI, for bone fractures of exposure to different hypoglycemic drugs in a logistic regression model. Left panel: Exposure for at least 36 months (mean ± SD 67.6 ± 22.3, 66.0 ± 21.5, and 60.2 ± 18.3 months for secretagogues, metformin, and insulin, respectively). Right panel: Exposure at index date. Data are presented on a logarithmic scale. *P < 0.05.
Adjusted ORs, with 95% CI, for bone fractures of exposure to different hypoglycemic drugs in a logistic regression model. Left panel: Exposure for at least 36 months (mean ± SD 67.6 ± 22.3, 66.0 ± 21.5, and 60.2 ± 18.3 months for secretagogues, metformin, and insulin, respectively). Right panel: Exposure at index date. Data are presented on a logarithmic scale. *P < 0.05.
Baseline characteristics of the sample enrolled
. | Case subjects . | Control subjects . |
---|---|---|
n (% women) | 83 (63.8) | 249 (63.8) |
Age (years) | 69.7 ± 11.1 | 68.2 ± 10.5 |
Duration of diabetes (years) | 12.8 ± 11.6 | 12.4 ± 10.1 |
A1C (%) | 8.0 ± 1.9 | 7.8 ± 1.6 |
BMI (kg/m2) | 27.6 ± 4.2 | 27.8 ± 4.1 |
Charlson's comorbidity score | 2.0 (1.0–3.0) | 2.0 (1.0–3.0) |
Systolic blood pressure (mmHg) | 142.7 ± 19.0 | 146.0 ± 19.0 |
Diastolic blood pressure (mmHg) | 79.1 ± 10.0 | 80.4 ± 10.6 |
Alcohol abuse (%) | 13.3 | 12.9 |
Current smokers (%) | 21.7 | 22.1 |
. | Case subjects . | Control subjects . |
---|---|---|
n (% women) | 83 (63.8) | 249 (63.8) |
Age (years) | 69.7 ± 11.1 | 68.2 ± 10.5 |
Duration of diabetes (years) | 12.8 ± 11.6 | 12.4 ± 10.1 |
A1C (%) | 8.0 ± 1.9 | 7.8 ± 1.6 |
BMI (kg/m2) | 27.6 ± 4.2 | 27.8 ± 4.1 |
Charlson's comorbidity score | 2.0 (1.0–3.0) | 2.0 (1.0–3.0) |
Systolic blood pressure (mmHg) | 142.7 ± 19.0 | 146.0 ± 19.0 |
Diastolic blood pressure (mmHg) | 79.1 ± 10.0 | 80.4 ± 10.6 |
Alcohol abuse (%) | 13.3 | 12.9 |
Current smokers (%) | 21.7 | 22.1 |
Data are means ± SD or OR (95% CI) unless otherwise indicated. P value is NS for all.
Exposure of at least 36 months to hypoglycemic treatments in case subjects and matched control subjects
. | Case subjects . | Control subjects . | OR (95% CI) . | P . |
---|---|---|---|---|
Secretagogues (%) | 24 (28.9) | 102 (41.0) | 0.59 (0.34–1.00) | 0.050 |
Glibenclamide | 17 (20.5) | 66 (26.5) | 0.71 (0.39–1.30) | NS |
Gliclazide | 7 (8.4) | 29 (11.6) | 0.70 (0.29–1.70) | NS |
Glimepiride | 0 (—) | 0 (—) | — | — |
Repaglinide | 2 (2.4) | 1 (0.4) | 3.02 (0.19–48.90) | NS |
Other sulphonylureas | 2 (2.4) | 4 (1.6) | 1.51 (0.27–8.40) | NS |
Insulin sensitizers (%) | 19 (22.9) | 83 (33.3) | 0.59 (0.32–1.07) | NS |
Metformin | 19 (22.9) | 82 (32.9) | 0.60 (0.34–1.08) | NS |
Fenformin | 0 (—) | 1 (0.4) | — | — |
Glitazones | 0 (—) | 0 (—) | — | — |
Acarbose (%) | 0 (—) | 1 (0.4) | — | — |
Insulin (%) | 25 (30.1) | 62 (24.9) | 1.30 (0.75–2.25) | NS |
Combined therapy (%)* | 16 (19.3) | 68 (27.3) | 0.64 (0.34–1.17) | NS |
. | Case subjects . | Control subjects . | OR (95% CI) . | P . |
---|---|---|---|---|
Secretagogues (%) | 24 (28.9) | 102 (41.0) | 0.59 (0.34–1.00) | 0.050 |
Glibenclamide | 17 (20.5) | 66 (26.5) | 0.71 (0.39–1.30) | NS |
Gliclazide | 7 (8.4) | 29 (11.6) | 0.70 (0.29–1.70) | NS |
Glimepiride | 0 (—) | 0 (—) | — | — |
Repaglinide | 2 (2.4) | 1 (0.4) | 3.02 (0.19–48.90) | NS |
Other sulphonylureas | 2 (2.4) | 4 (1.6) | 1.51 (0.27–8.40) | NS |
Insulin sensitizers (%) | 19 (22.9) | 83 (33.3) | 0.59 (0.32–1.07) | NS |
Metformin | 19 (22.9) | 82 (32.9) | 0.60 (0.34–1.08) | NS |
Fenformin | 0 (—) | 1 (0.4) | — | — |
Glitazones | 0 (—) | 0 (—) | — | — |
Acarbose (%) | 0 (—) | 1 (0.4) | — | — |
Insulin (%) | 25 (30.1) | 62 (24.9) | 1.30 (0.75–2.25) | NS |
Combined therapy (%)* | 16 (19.3) | 68 (27.3) | 0.64 (0.34–1.17) | NS |
Data are n (%) unless otherwise indicated.
Insulin secretagogues + insulin sensitizers.
Hypoglycemic treatments in case subjects and matched control subjects at the index time
. | Case subjects . | Control subjects . | OR (95% CI) . | P . |
---|---|---|---|---|
Secretagogues (%) | 38 (45.8) | 139 (55.8) | 0.67 (0.41–1.10) | 0.10 |
Glibenclamide | 18 (21.7) | 73 (29.3) | 0.67 (0.37–1.20) | NS |
Gliclazide | 7 (8.4) | 39 (15.7) | 0.50 (0.21–1.16) | 0.10 |
Glimepiride | 6 (7.2) | 14 (5.6) | 1.31 (0.49–3.52) | NS |
Repaglinide | 5 (6.0) | 13 (5.2) | 1.16 (0.40–3.37) | NS |
Other sulphonylureas | 3 (3.6) | 1 (0.4) | 9.30 (0.95–90.70) | 0.09 |
Insulin sensitizers (%) | 41 (49.4) | 132 (53.0) | 0.89 (0.55–1.49) | NS |
Metformin | 40 (48.2) | 130 (52.2) | 0.85 (0.52–1.40) | NS |
Fenformin | 1 (1.2) | 2 (0.8) | 1.51 (0.13–16.83) | NS |
Glitazones | 2 (2.4) | 0 (—) | — | — |
Acarbose (%) | 1 (1.2) | 4 (1.6) | 0.75 (0.08–6.80) | NS |
Insulin (%) | 43 (51.8) | 91 (36.5) | 1.87 (1.13–3.08) | 0.014 |
Combined therapy (%)* | 27 (32.5) | 96 (38.6) | 0.77 (0.45–1.30) | NS |
. | Case subjects . | Control subjects . | OR (95% CI) . | P . |
---|---|---|---|---|
Secretagogues (%) | 38 (45.8) | 139 (55.8) | 0.67 (0.41–1.10) | 0.10 |
Glibenclamide | 18 (21.7) | 73 (29.3) | 0.67 (0.37–1.20) | NS |
Gliclazide | 7 (8.4) | 39 (15.7) | 0.50 (0.21–1.16) | 0.10 |
Glimepiride | 6 (7.2) | 14 (5.6) | 1.31 (0.49–3.52) | NS |
Repaglinide | 5 (6.0) | 13 (5.2) | 1.16 (0.40–3.37) | NS |
Other sulphonylureas | 3 (3.6) | 1 (0.4) | 9.30 (0.95–90.70) | 0.09 |
Insulin sensitizers (%) | 41 (49.4) | 132 (53.0) | 0.89 (0.55–1.49) | NS |
Metformin | 40 (48.2) | 130 (52.2) | 0.85 (0.52–1.40) | NS |
Fenformin | 1 (1.2) | 2 (0.8) | 1.51 (0.13–16.83) | NS |
Glitazones | 2 (2.4) | 0 (—) | — | — |
Acarbose (%) | 1 (1.2) | 4 (1.6) | 0.75 (0.08–6.80) | NS |
Insulin (%) | 43 (51.8) | 91 (36.5) | 1.87 (1.13–3.08) | 0.014 |
Combined therapy (%)* | 27 (32.5) | 96 (38.6) | 0.77 (0.45–1.30) | NS |
Data are n (%) unless otherwise indicated.
Insulin secretagogues + insulin sensitizers.
References
Published ahead of print at http//:care.diabetesjournals.org on 16 November 2007. DOI: 10.2337/dc07-1736.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C Section 1734 solely to indicate this fact.