Osteoporos Int (2007) 18:427–444 DOI 10.1007/s00198-006-0253-4
ORIGINAL ARTICLE
Discrepancies in bone mineral density and fracture risk in patients with type 1 and type 2 diabetes—a meta-analysis P. Vestergaard
Received: 11 July 2006 / Accepted: 6 October 2006 / Published online: 27 October 2006 # International Osteoporosis Foundation and National Osteoporosis Foundation 2006
Abstract Introduction and hypothesis Diabetes affects bone metabolism. The hypothesis was that type 1 (T1D) and type 2 (T2D) affects BMD and fracture risk differently. Material and methods Pubmed, Embase, and Web of Science were searched using the terms “diabetes”, “fracture”, and “bone mineral”. Results Hip fracture risk was increased in T1D (RR=6.94, 95% CI: 3.25–14.78, five studies) and T2D (1.38, 95% CI: 1.25–1.53, eight studies) compared to subjects without diabetes. The increase in relative hip fracture risk was significantly higher in T1D than in T2D. BMD Z-score was decreased in the spine (mean±SEM −0.22±0.01) and hip (−0.37±0.16) in T1D and increased in the spine (0.41± 0.01) and hip (0.27±0.01) in T2D. A meta-regression showed that body mass index (BMI) was a major determinant for BMD in both the spine and hip. Glycated haemoglobin (HbA1C) was not linked to BMD. The increase in fracture risk was higher and BMD lower in patients with complications to diabetes. Conclusions Hip fracture risk is increased in both T1D and T2D, whereas BMD is increased in T2D and decreased in T1D. A common factor such as complications may explain the increase in fracture risk, whereas BMI may ameliorate the increase in fracture risk in T2D. Keywords Bone mineral density . Diabetes . Fracture . Meta-analysis
P. Vestergaard (*) The Osteoporosis Clinic, Aarhus Amtssygehus, Aarhus University Hospital, Tage Hansens Gade 2, 8000 Aarhus C, Denmark e-mail:
[email protected]
Impaired glucose metabolism has a number of detrimental effects on bone metabolism, which have significant consequences for patients with diabetes in terms of decreased bone mineral density [1, 2] and increased risk of fractures [3]. The pathophysiological mechanisms can be divided into mechanisms that decrease bone mineral density (BMD) or weaken bone structure, and those which increase the likelihood of falls and other traumas. Among the mechanisms that decrease BMD are 1) an increased urine calcium excretion that leads to a negative calcium balance [1] (The increased urine calcium excretion is linked to hyperglycaemia [1]. Antidiabetic treatment reverses the urine calcium loss [4]), 2) a functional hypoparathyroidism [5] (The increased urine calcium loss should theoretically lead to secondary hyperparathyroidism, but this is not seen in patients with diabetes), 3) alterations in vitamin D metabolism, which is particularly prominent in patients with nephropathy [6], and 4) perhaps insulin itself and insulin like growth factors [7]. Insulin is an anabolic hormone, and patients with type 1 diabetes (T1D) have an absolute deficiency in insulin, whereas patients with type 2 diabetes (T2D) may have excess insulin levels. Mechanisms that lead to a decreased bone biomechanical competence besides decreases in BMD are alterations in glycosylation of collagen [8–14] brought about by hyperglycaemia in the same was as increased glycosylation of haemoglobin, which is expressed as HbA1C. These advanced glycation end products (AGE) and their receptors (RAGE) play an important role in bone metabolism and bone strength [8–14]. Other studies have also raised concern that diabetes through the impaired bone metabolism may delay fracture healing [15]. The complications of diabetes may also contribute to fracture risk, in particular in patients with renal failure and those undergoing renal transplantation [16, 17]. Neuropathy may perhaps also be involved through decreased mobility and increased
428
bone mineral loss [18]. Angiopathy may also play a role as microangiopathy may alter blood flow to the bones through altered endothelial function although this is controversial [19], and macroangiopathy may also decrease blood supply to parts of the skeleton [20, 21]. However, besides the alterations related to BMD and bone strength, diabetes may be associated with an increased risk of falls [22] and other traumas [23]. This may be related to fits during hypoglycaemia [24, 25] or falls e.g., related to decreased eyesight following retinopathy or decreased postural balance related to neuropathy. Patients with T2D often have a higher body weight and body mass index (BMI) than patients with T1D. When they fall they sustain a greater traumatic load due to obesity, but on the other hand may the soft-tissue energy absorption upon impact be greater in obese patients with T2D than in thin T1D patients. This may lead to differences in fracture pattern. Prior studies have shown discrepant results for type 1 and type 2 diabetes with a decreased bone mineral density (BMD) in patients with T1D [26] and a normal to increased BMD in patients with T2D [27]. Prior studies in patients with T1D have shown an increase in fracture risk [3], whereas discrepant results have been reported for T2D [3, 27]. It is, thus, unclear if the increased BMD actually to some degree is protective against fractures in T2D or whether the increased BMD is just the result of the generally higher body mass index (BMI) in patients with T2D, the patients despite this having an increased risk of fractures stemming from the disturbances in bone metabolism with decreased bone strength. If the hyperglycaemia alone was responsible for the alterations in skeletal metabolism, the changes in BMD and fracture risk should theoretically be equal in T1D and T2D. However, if other factors such as insulin levels and BMI were also responsible, differences might be anticipated. Furthermore, if complications to diabetes were involved this could also contribute to the differences as patients with T2D may have complications at the diagnosis of their diabetes, whereas this is not the case in patients with T21. It is thus of interest to assess the degree of BMD changes and changes in fracture risk with particular respect to differences between T1D and T2D. We addressed the following research questions in this meta-analysis: 1) What were the changes in spine and hip BMD Z-scores in patients with T1D and T2D? 2) What were the changes in fracture risk in patients with T1D and T2D? 3) Did the changes in BMD explain the changes in fracture risk in T1D and T2D? 4) What were the effects of glycaemic control and complications on the risk of fracture and the changes in BMD in T1D and T2D?
Osteoporos Int (2007) 18:427–444
Material and methods A systematic search of Pubmed (1951–December 16, 2005), Embase (1974–December 16, 2005), and ISI Web of Science (1945–December 16, 2005) was performed. The search was divided into two. The first search used the terms “diabetes” and “fracture” and produced 1,499 papers, and among these 16 were included in the analysis. The second search used the terms “diabetes” and “bone mineral” and produced 889 papers, of which 65 were included in the analysis. Due to overlap of one paper between the two search strategies, the combined number of papers was 80. Inclusion criteria were epidemiological studies (cross sectional, cohort or case control) that provided data on fracture risk or BMD in patients with T1D or T2D compared to a control group without diabetes. Data for BMD data were expressed as Z-scores. Data could either be expressed as Zscores in patients with diabetes in relationship to a reference population or as a direct difference between patients and a control group. Only published results were used. Papers in all languages and abstracts were eligible. Diabetes was categorised as T1D or T2D depending on the information provided by the authors. Subgroups of patients (say patients who had to have T2D because they were diagnosed with diabetes and were not treated with insulin but rather with oral antidiabetics and/or diet) were also eligible. The studies were excluded if no clear definition of diabetes was given [28]. The studies were also excluded provided that the entire group of patients with diabetes were selected based on presence of complications known to be associated with fracture risk say renal failure. Studies with both groups of patients with and without complications to diabetes were eligible. If more than one study group presented data from the same study population [27, 29–31], the one with the longest duration of observation and the more complete reporting of data was chosen. BMD data had to be measured on modern scanners using DXA technique [32]. Only scans of the lumbar spine (L1–L4 or L2–L4) or hip (femoral neck or total hip) were eligible. No internationally accepted criteria for evaluating quality of epidemiological studies exist, as is the case for randomised controlled trials (e.g., the Jadad scale [33]). In the actual study ad hoc classification was attempted based on reporting of criteria for diagnosing diabetes, CV for BMD measurements, consecutive selection of cases, and random selection of controls. All risk estimates (relative risk, odds ratio, hazard ration, prevalence ratio etc.) were considered estimates of relative risk. The estimates had to be crude estimates of fracture risk in patients with diabetes compared to non-diabetics. If large age or gender differences between patients and controls were present, age- and gender adjusted risk estimates were used. Estimates adjusted for say BMI, smoking, BMD etc. were
Osteoporos Int (2007) 18:427–444
429
Table 1 Relative risk estimates of fractures in patients with diabetes N controls
Age BMI (years) (kg/cm2)
RR (95% CI)
12,639
45.4
25.0
12,639
64.0
28.0
12,639
61.2
28.6
89
12,639
66.7
27.3
29
14,065
43.5
24.3
198
14,065
68.2
29.3
78
14,065
65.7
30.5
120
14,065
69.7
28.6
52
12,639
45.4
25.0
Hip
175
12,639
64.0
28.0
Men
Hip
86
12,639
61.2
28.6
Men
Hip
89
12,639
66.7
27.3
Women
Hip
29
14,065
43.5
24.3
T2D
Women
Hip
198
14,065
68.2
29.3
T2D insulin treated T2D non insulin treated
Women
Hip
78
14,065
65.7
30.5
Women
Hip
120
14,065
69.7
28.6
3.06 (1.27–7.38) 1.19 (0.61–2.31) 1.10 (0.41–2.95) 1.28 (0.53–3.11) 3.03 (0.98–9.44) 0.89 (0.59–1.35) 0.87 (0.43–1.74) 1.39 (0.90–0.54) 17.79 (5.57– 56.75) 1.45 (0.53–3.99) 1.77 (0.43–7.22) 1.23 (0.30–5.03) 8.55 (1.19–61.49) 1.72 (0.97–3.02) 1.72 (0.64–4.64) 1.71 (0.87–3.36)
2,688
>65
N/A
Women
Hip Hip
316
T2D T2D oral antidiabetics T2D insulin
Women
Hip
Women
Hip
T2D
Men
Hip
T2D oral antidiabetics T2D insulin
Men
Hip
Men
Hip
T2D
Both Both
Reference
Type of diabetes
Gender
Fracture type
Ahmed [46] 2005, Norway, self report, screening of HbA1C and register data
T1D
Men
T2D
Men
T2D insulin treated T2D non insulin treated T1D
Men
Non 52 vertebral Non 175 vertebral Non vertebral 86
T2D
Women
T2D insulin treated T2D non insulin treated T1D
Women Women Men
Non vertebral Non vertebral Non vertebral Non vertebral Non vertebral Hip
T2D
Men
T2D insulin treated T2D non insulin treated T1D
Cortes-Sancho [48] 2004, Taiwan, screening, FPG and HbA1C
De Liefde [29] 2005, Netherlands, screening, OGTT)
Men Women
N DM
1.4 (1.1–1.7) 1.8 (1.3–2.6) 2.9 (1.9–4.5) 1.5 (1.1–2.0) 2.6 (1.7–4.1) 4.2 (2.0–8.6) 792
Both
Non vertebral Hip
Both
Wrist
5,863
73.8
26.8 1.09 (0.88–1.34) 1.16 (0.81–1.65) 1.08 (0.71–1.66)
430
Osteoporos Int (2007) 18:427–444
Table 1 (continued) Reference
Type of diabetes
Gender
Fracture type
N DM
N controls
Age BMI (years) (kg/cm2)
RR (95% CI)
Forsen [3] 1999, Norway, self report, screening of non-fasting blood glucose and HbA1C
T2D ≤5 years duration T2D ≤5 years duration T2D ≤5 years duration T2D ≤5 years duration T2D ≤5 years duration T2D ≤5 years duration T2D ≤5 years duration T1D
Women 50– 74 years Women 50– 74 years Women ≥75 years Women ≥75 years Men 50– 74 years Men ≥75 years Men ≥75 years Women 50– 74 years Men 50– 74 years Women
Hip
221
13,685
66
29
Hip
274
13,685
67
30
Hip
144
3,831
82
28
Hip
196
3,831
81
27
Hip
148
13,210
66
27
Hip
220
2,868
81
26
Hip
7
2,868
81
25
Hip
23
13,685
59
27
Hip
28
13,210
60
25
Any fracture 74
1,058
N/A
N/A
Women
Hip
74
1,058
N/A
N/A
Women
Wrist
74
1,058
N/A
N/A
Women
Spine
74
1,058
N/A
N/A
Women
Vertebral deformity Vertebral deformity Vertebral deformity Vertebral deformity Any fracture Wrist
63
5,156
67
N/A
35
5,156
63
N/A
347
1,970
67
N/A
182
1,970
63
N/A
147
3,450
N/A
N/A
N/A
N/A
0.7 (0.4–1.5) 1.7 (1.1–2.7) 0.9 (0.5–1.5) 1.4 (0.9–2.0) 1.0 (0.4–2.6) 1.8 (1.0–3.4) 1.1 (0.6–2.5) 5.7 (1.8–17.9) 4.0 (0.6–28.2) 0.83 (0.52–1.33) 0.54 (0.20–1.49) 0.62 (0.33–1.16) 0.52 (0.19–1.42) 1.24 (0.61–2.51) 0.97 (0.34–2.77) 0.92 (0.67–1.26) 0.77 (0.49–1.22) (0.4–3.0) 1.4 (0.2–10.2) 2.7 (0.3–21.7) 7.6 (5.9–9.6) 4.1 (2.7–6.0) 17.4 (12.5–23.5) 4.5 (3.2–6.3) 31.6 (21.7–44.3) 4.6 (3.3–6.4) 32.6 (22.3–46.0) 6.6 (5.0–8.5)
T1D Gerdhem [39] 2005, Sweden, self report
Hanley [49] 2003, Canada, self report
Ivers [42] 2001, Australia, self report
Miao [37] 2005, Sweden, register data
T2D, no insulin treatment T2D, no insulin treatment T2D, no insulin treatment T2D, no insulin treatment T1D T1D
Men
T2D
Women
T2D
Men
T2D diet or tablets
Both
T1D
Men
Proximal humerus 12,551
T1D no eye complications T1D eye complications T1D no nephropathy
Men
Hip
12,551
Population N/A
N/A
Men
Hip
12,551
Population N/A
N/A
Men
Hip
12,551
Population N/A
N/A
T1D nephropathy
Men
Hip
12,551
Population N/A
N/A
T1D no neurological complications T1D neurological complications T1D no cardiovascular complications
Men
Hip
12,551
Population N/A
N/A
Men
Hip
12,551
Population N/A
N/A
Men
Hip
12,551
Population N/A
N/A
Population N/A
Osteoporos Int (2007) 18:427–444
431
Table 1 (continued) Reference
Michaelsson [50] 1995, Sweden, self report Nicodemus [51] 2001, USA, self report
Poor [52] 1995, USA, files Seeley [53] 1996, USA, self report Strotmeyer [43] 2005, USA, self report or FPG screening Taylor [54] 2004, USA, self report Van Daele [27] 1995, Netherlands, drug use or OGTT Vestergaard [38] 2005, Denmark, register data
Type of diabetes
Gender
Fracture type
N DM
N controls
Age BMI (years) (kg/cm2)
T1D Cardiovascular complications T1D
Men
Hip
12,551
Population N/A
N/A
Women
Hip
12,054
Population N/A
N/A
T1D no eye complications T1D eye complications
Women
Hip
12,054
Population N/A
N/A
Women
Hip
12,054
Population N/A
N/A
T1D no nephropathy
Women
Hip
12,054
Population N/A
N/A
T1D nephropathy
Women
Hip
12,054
Population N/A
N/A
T1D no neurological complications T1D neurological complications T1D no cardiovascular complications T1D cardiovascular complications T2D dietary treatment
Women
Hip
12,054
Population N/A
N/A
Women
Hip
12,054
Population N/A
N/A
Women
Hip
12,054
Population N/A
N/A
Women
Hip
12,054
Population N/A
N/A
Women
Hip
27
1,066
N/A
N/A
T2D oral antidiabetics
Women
Hip
47
1,066
N/A
N/A
T1D
Women
Hip
47
30,377
60.9
25.8
T2D
Women
Hip
1,682
30,377
62.3
30.5
T2D ≤4 years duration T2D 5–12 years duration T2D 13–40 years duration T2D insulin treated
Women
Hip
N/A
N/A
N/A
N/A
Women
Hip
N/A
N/A
N/A
N/A
Women
Hip
N/A
N/A
N/A
N/A
Women
Hip
N/A
N/A
N/A
N/A
T2D oral antidiabetics
Women
Hip
N/A
N/A
N/A
N/A
T2D no pharmacological Women treatment T2D Men
Hip
N/A
N/A
N/A
N/A
Hip
42
190
78.5
N/A
T1D
Women
Foot
101
9,441
71.4
N/A
T2D
Both
Any fracture 566
2,236
73.5
28–30
T2D
Women
Hip
443
7,437
73.5
N/A
T2D
Women
Any fracture 335
3,115
72.3
27.8
T1D
Both
Any fracture 4,362
484,657
N/A
N/A
T2D
Both
Any fracture 9,598
484,657
RR (95% CI) 28.6 (15.2–48.8) 9.8 (7.3–12.9) 4.1 (2.3–6.9) 20.5 (14.5–28.3) 6.4 (4.3–9.2) 32.6 (20.4–49.4) 5.7 (3.7–8.3) 41.6 (26.9–61.4) 8.1 (5.8–11.0) 29.2 (15.1.51.1) 1.09 (0.41–2.91) 2.04 (1.09–3.82) 14.1 (5.85–34.20) 1.75 (1.25–2.43) 1.47 (0.81–2.67) 1.82 (1.05–3.16) 1.21 (0.68–2.14) 2.79 (1.61–4.85) 1.82 (1.05–3.16) 1.21 (0.68–2.14) 0.9 (0.5–1.7) 2.36 (0.97–5.73) 1.23 (0.82–1.86) 1.52 (1.14–2.02) 0.63 (0.44–0.90) 1.30 (1.16–1.46) 1.19 (1.11–1.27)
432
Osteoporos Int (2007) 18:427–444
Table 1 (continued) Reference
Type of diabetes
Gender
Fracture type
N DM
N controls
T1D
Both
Hip
4,362
484,657
T2D
Both
Hip
9,598
484,657
T1D
Both
Wrist
4,362
484,657
T2D
Both
Wrist
9,598
484,657
T1D
Both
Spine
4,362
484,657
T2D
Both
Spine
9,598
484,657
Age BMI (years) (kg/cm2)
RR (95% CI) 1.70 (1.31–2.21) 1.38 (1.18–1.60) 1.04 (0.76–1.44) 1.21 (1.01–1.45) 2.48 (1.33–4.62) 1.34 (0.97–1.86)
The mode of diagnosis of diabetes is mentioned under reference. Under reference, country of origin, date of publication, and method of diagnosis of diabetes has been mentioned. OGTT: Oral glucose tolerance test, FPG: fasting plasma glucose. N DM is number of patients with diabetes, N controls is number of control subjects, BMI: Body mass index, RR: relative risk with 95% confidence intervals (95% CI).
excluded. A common weighted risk estimate for relative fracture risk was calculated as a derSimonian and Laird estimator using a random effects model [34]. Tests for heterogeneity were performed. Publication bias was tested using funnel plots. Sensitivity analyses were performed excluding large studies. For BMD a weighted mean difference (WMD) estimate was calculated for Z-score in the spine or hip [34]. Analyses of pooled relative fracture risk and WMD were only performed providing that at least three studies were available. The observed BMD was converted into an expected relative fracture risk using the estimates of Marshall et al. [35]. The expected relative fracture risk was calculated as c−z, where z is the observed zscore, and c is a constant. To assess the effects of diabetes type, age, disease duration, BMI and average glucose level (HbA1C) a meta-regression analysis was performed using STATA 8.0 [36].
Results Fracture risk Table 1 shows details of studies on fracture risk in patients with diabetes compared to controls. Most studies focused on hip fracture risk. A trend towards an increased fracture risk was reported at most skeletal sites for T1D, whereas for T2D many studies also reported a trend towards an increased fracture risk at most skeletal sites, although some studies also reported a trend towards a decreased fracture risk. There was a marked trend towards a higher relative fracture risk with the presence of complications in T1D (study by Miao et al. [37]). No studies on the effect of complications were present for T2D. Table 2 shows the combined results for relative fracture risk. Only for hip fractures were the studies numerous enough to allow analysis in T1D. There was a highly
Table 2 Observed risk of fractures. Only studies with all treatment modalities represented Diabetes type
Fracture type
RR (95% CI)
P
Number of studies
T1D
Hip Hip
6.94 (3.25–14.78)* 8.65 (7.26–10.30)*