Earlier studies have suggested that untreated coeliac disease may be associated with osteoporosis, but results are contradictory for the risk of long-term fractures.
To study the association between coeliac disease and fractures.
We used Cox regresson to examine the future risk of hip fracture and fracture of any type in more than 13 000 individuals with coeliac disease and 65 000 age- and sex-matched reference individuals in a general population-based cohort.
During follow-up, 1365 first hip fractures and 4847 fractures of any type occurred. Coeliac disease was positively associated with subsequent hip fracture (hazard ratio = 2.1; 95% CI = 1.8–2.4) (in children: hazard ratio = 2.6; 95% CI = 1.1–6.2) and fractures of any type (hazard ratio = 1.4; 95% CI = 1.3–1.5) (in children: hazard ratio = 1.1; 95% CI = 1.0–1.2). The absolute excess risk of hip fractures in children with coeliac disease was 4/100 000 person-years. Incidence ratios for hip fracture in individuals with CD were around two both prior to diagnosis of coeliac disease and afterwards; this risk increase remained 20 years after diagnosis of coeliac disease.
Individuals with coeliac disease, including children with coeliac disease, may be at increased risk of hip fracture and fracture of any type. Coeliac disease may be positively associated with long-term hip fracture risk.
Coeliac disease (CD) is an autoimmune disease characterized by small-bowel inflammation.1 In the last 20 years it has become increasingly clear that CD is associated with a number of disorders and complications,2–4 including lower bone mineral density.5 Decreased bone mineral density is a risk factor for osteoporotic fractures.6, 7 Osteoporotic fractures are common8 and associated with huge health costs for society.9 Individuals with osteoporotic fractures often have a decreased quality of life and are at an increased risk of death.10
Despite the established association between CD and lower bone mineral density5, 11 and osteoporosis,12, 13 earlier research on CD and fractures is inconsistent,14–20 with relative risks for any fracture ranging from 0.917 to 719 and for hip fractures from 0.6615 to 1.9.14 Because of these inconsistencies, the rationale for screening for CD in individuals with fractures has been questioned.21, 22 The heterogeneous results may partly be due to few positive events, a fact that may also have prevented stratification by age. Although, some of the above-mentioned studies have included individuals with CD diagnosed in childhood,14, 17 no specific relative risks for fractures are given for children in these studies.14, 17 It is therefore unclear to what extent CD diagnosed in childhood is a risk factor for subsequent fracture.
For all the above reasons, there is a need for further research on CD and risk of fractures. We used the Swedish National Registers to estimate the risk of hip fracture and any fracture in a large general population cohort. The long follow-up allowed us to also examine the relationship between time to diagnosis of CD and hip fractures, as well as to evaluate the long-term risk of hip fracture in individuals with CD.
Materials and methods
The Swedish National Board of Health identified all individuals with a hospital discharge diagnosis of CD between 1964 and 2003 by the Swedish National Inpatient Register (IPR). The IPR contains individual-based data from hospital discharge diagnoses in selected parts of Sweden since 1964. It has complete coverage of all hospital discharges in Sweden since 1987. Every record in the IPR can be identified through a Personal Identity Number. The Personal Identity Number is a unique number assigned to more than 99.9% of all Swedish residents and immigrants.23 Individuals who received any of the following International Classification of Disease (ICD) codes between 1964 and 2003 were defined as having CD: ICD-7: 286.00; ICD-8: 269.00, 269.98, ICD-9: 579A; ICD-10: K90.0).
For each individual with CD, Statistics Sweden used the National Total Population Register to select up to five reference individuals matched for age, sex, calendar year and county. The Total Population Register24 includes information on area of residence, vital status, and dates of immigration or emigration. We then used the IPR to define our two outcome measures: first hip fracture and first fracture of any type. We defined hip fracture and fracture of any type according to the relevant ICD codes (hip fracture: ICD 7–9: most diagnoses starting with ‘820’; and ICD-10: S72.0-S72.2). The IPR was also used to define diabetes mellitus (DM), thyroid disease and depression. The IPR does not distinguish between type 1 and type 2 DM.
The Swedish National Board of Health and Welfare identified 15 533 individuals with CD. We excluded 94 individuals with CD because of data irregularities (such as date of death preceding the date of diagnosis of CD). We then excluded 280 individuals with hip fracture within 1 year after diagnosis of CD (205 prior to diagnosis of CD), and 972 individuals who had a follow-up lesser than 1 year for other reasons, such as death. Similar exclusion criteria were applied to reference individuals. The analyses of subsequent hip fracture in the current study were hence based on 14 187 individuals with CD and 68 952 reference individuals who never had CD. Only first fractures were considered in the analyses. Characteristics of the study participants are given in Table 1.
|Characteristics||Reference (no coeliac disease) (%)||Coeliac disease (%)|
|Total||68 952||14 187|
|Age at first recorded diagnosis of coeliac disease (years)|
|Men||28 522 (41.4)||5876 (41.4)|
|Women||40 430 (58.6)||8311 (58.6)|
|1964–1973||2404 (3.5)||494 (3.5)|
|1974–1983||19 147 (27.8)||3918 (27.6)|
|1984–1993||30 327 (44.0)||6219 (43.8)|
|1994–2003||17 074 (24.8)||3556 (25.1)|
|Co-morbidity before end of follow-up|
|Diabetes mellitus||1892 (2.7)||926 (6.5)|
|Thyroid disorder||741 (1.1)||537 (3.8)|
|Depression||930 (1.3)||438 (3.1)|
|I||7275 (10.6)||1498 (10.6)|
|II||9204 (13.3)||2142 (15.1)|
|III||20 084 (29.1)||5145 (36.3)|
|Missing data||32 389 (47.0)||5402 (38.1)|
|Median age at first fracture (range)*|
|Hip fracture||77 (4–97)||81 (3–98)|
|Any fracture†||53 (1–95)||26 (1–99)|
|Median duration until first fracture (range)*|
|Hip fracture||6.0 (1.0–31.6)||9.5 (1.0–31.4)|
|Any fracture†||8.2 (1.0–37.7)||9.2 (1.0–34.2)|
The analyses of any subsequent fracture were based on 13 724 individuals with CD and 65 627 individuals who never had CD. All individuals in these analyses were free of prior fracture 1 year after study entry. The lower number of individuals in the analyses of fractures of any type is explained by the larger number of any fractures occurring in study participants prior to study entry.
Cases of fractures caused by suspected high-impact trauma were retained in all analyses because there are indications of comparable increased low- and high-impact trauma fracture risk with decreasing bone density especially in the elderly.25
In a subset of individuals (n = 45 348) we had data on socioeconomic index (SEI) (Table 1) based on a three-category occupational classification from 1968.26 Some 6500 of these were children born after 1990. These children were assigned a socioeconomic code on the basis of the occupation of the mother. We used the SEI of the mother and not that of the father because in divorced families, the children more often live with their mother than with their father.
CD and subsequent fracture
Cox regression analysis estimated the risk of subsequent fracture in individuals with CD. Risk estimates are given as hazard ratios (HRs). The follow-up time began 1 year after study entry (date of first recorded diagnosis of CD and the corresponding date in reference individuals) and ended on the date of first discharge diagnosis of fracture, date of emigration, death or the end of the study period (31 December 2003), whichever occurred first. We excluded the first year after study entry to minimize the risk of detection bias. Log minus log plots were used to test the proportional hazard assumption (Appendix).
The Cox regression model was internally stratified, i.e. individuals with CD were matched with their corresponding reference individuals by age at study entry, sex, calendar year and county of residence at study entry. Individuals with CD were only compared with their matched reference controls.
In separate analyses, we stratified by sex and age at first recorded diagnosis of CD (≤15 years vs. ≥16 years). We used this age cut-off because it is often used in Sweden's official population statistics. In a subset of individuals with data on SEI we adjusted for SEI. We also adjusted for the following disorders because they are associated with an increased risk of both CD and fractures and therefore constitute potential confounders: thyroid disorders,27, 28 diabetes mellitus13, 29 and depression.30, 31 We did not adjust for hyperparathyroidism because high levels of parathyroid hormone are often caused by CD itself.32 Hyperparathyroidism could therefore be regarded as an intermediate between CD and osteoporosis.33 Neither did we adjust for the presence of epilepsy because recent research has indicated that epilepsy is not associated with CD.34 To evaluate if inclusion of fractures in the first year after study entry would affect our risk estimate, we performed an additional analysis including these fractures.
In order to validate our findings, we also looked at our main outcome measure, hip fracture, when (i) study participants were free of any fracture at time of study entry, (ii) study participants with hip fractures prior to study entry were included (our outcome measure was then first hip fracture after study entry and diagnosis of CD), and (iii) hip fractures occurring before the age of 30 years were excluded.
CD and prior history of fracture
Conditional logistic regression estimated the association of CD (the dependent variable) with prior hip fracture or prior fractures of any type. These analyses were performed to evaluate if the associations between CD and fractures were independent of sequence, i.e. if there was only an association with hip fracture or any fracture after a diagnosis of CD and not before. The end of follow-up was defined as the date of first diagnosis of CD and the same date for the matched reference individuals without a diagnosis of CD. Those with 1 year or less between the dates of first hip fracture (or any fracture) and study entry (equivalent to date of diagnosis of CD) were excluded. Risk estimates are given as odds ratios (OR).
Incidence of hip fracture in CD diagnosed in adults
The risk of complications may be greatest in the immediate years prior to or after a diagnosis of CD,35 but less so in the distant past or after many years with gluten-free diet. For this reason we also investigated the incidence of hip fractures and incidence ratios (IRs) according to CD status in a number of pre-specified time periods. In these analyses, we only considered a person’s first hip fracture in each time period. To make data prior to CD and afterwards comparable, we restricted the IR analyses to individuals entering the study after the age of 15 years (CD diagnosed in adulthood). These individuals had been at risk of hip fractures for more than 10 years both prior to diagnosis of CD and after such diagnosis. We used conditional logistic regression to calculate statistical significance for hip fracture incidence in time periods preceding diagnosis of CD and Cox regression for time periods following diagnosis of CD.
Statistical significance was defined as 95% confidence intervals for risk estimates not including 1.0. Statistical values were calculated using SPSS 11.0 (Chicago, IL, USA).
We had 80% power to detect an increased risk of subsequent hip fracture (our main outcome measure) at a significance level of 5% in individuals with CD if more than 190 of 14 187 of these individuals had a subsequent hip fracture compared with 1037 of 68 952 individuals without a diagnosis of CD. This corresponds to a HR of 1.2.
This study was approved by the Research Ethics Committee of the Karolinska Institutet. None of the participants was contacted. Patient information was anonymized prior to the analyses.
The median age at study entry was 2 years in both individuals with CD and those who never had a diagnosis of CD (range = 0–93 years). Some 4800 (34%) individuals with CD had been diagnosed in adulthood. The mean age at first diagnosis of CD was 2.3 years in the paediatric population and 53 years in the adult population. The majority of study participants were female (Table 1). The median duration from study entry to fracture and the median age at time of fracture are given in Table 1; 96.8% of all hip fractures occurred at the age of 50 years or later (1323/1365).
CD and subsequent hip fracture
Individuals with CD were at increased risk of subsequent hip fracture (HR = 2.1; 95% CI = 1.8–2.4; see also Table 2 and Figures 1 and 2). Hip fracture incidence was 172/100 000 person-years in individuals with CD compared with 108/100 000 person-years in reference individuals (Table 2). There was no notable difference in risk estimates for hip fracture stratified by sex or age at first recorded diagnosis of CD (Table 2), and children with CD were at a statistically increased risk of subsequent hip fracture. The hip fracture incidence in children with CD was 6/100 000 person-years, i.e. an absolute excess risk of 4 hip fractures/100 000 person-years (Table 2). There was no difference in risk estimates over time (HRs per calendar period: 1964–1973: 2.0; 1974–1983: 1.9; 1984–1993: 2.2; 1994–2003: 2.1; all P < 0.05). Adjustment for SEI in a subset of individuals with data on SEI did not affect our risk estimate (both crude and adjusted HR = 2.4). When we adjusted for a diagnosis of DM, depression and thyroid disease before the end of follow-up, the adjusted HR for subsequent hip fracture in individuals with CD was 2.0 (95% CI = 1.7–2.3; P < 0.001). When we included the first year after study entry, the risk estimate for subsequent hip fracture was 2.2 (Table 2).
|Hip fracture||Any fracture|
|Participants||Events/person-years*||HR, 95% CI||Participants||Events/person-years*||HR, 95% CI|
|With CD||No CD||With CD||No CD||With CD||No CD||With CD||No CD|
|First year excluded|
|Any coeliac disease||14 187||68 952||328/189 892 (172)||1037/960 471 (108)||2.1; 1.8–2.4||13 724||65 527||1011/181 095 (558)||3836/904 557 (424)||1.4; 1.3–1.5|
|0–15 years||9368||46 178||8/141 519 (6)||15/696 674 (2)||2.6; 1.1–6.2||9294||45 548||438/136 672 (320)||1942/670 538 (290)||1.1; 1.0–1.2|
|≥16 years||4819||22 774||320/48 373 (662)||1022/263 797 (387)||2.1; 1.8–2.4||4.430||19 979||573/44 423 (1290)||1894/234 019 (809)||1.8; 1.6–2.0|
|Male||5876||28 522||95/79 352 (120)||281/403 440 (70)||2.2; 1.7–2.9||5678||27 008||445/74 974 (594)||1710/375 724 (455)||1.4; 1.2–1.5|
|Female||8311||40 430||233/110 540 (211)||756/557 031 (136)||2.0; 1.7–2.4||8046||38 519||566/106 121 (533)||2126/528 833 (402)||1.4; 1.3–1.6|
|First year included|
|Any coeliac disease||15 234||74 615||403/204 665 (197)||1357/1 057 316 (128)||2.2; 2.0–2.5||14 753||70 884||1173/195 390 (600)||4433/997 873 (444)||1.5; 1.4–1.6|
When we restricted our analysis to individuals free of any fracture 1 year after study entry, the risk estimate for subsequent hip fracture remained statistically significantly increased (HR = 2.0; 95% CI = 1.7–2.3; P < 0.001). Moreover, when we excluded first hip fractures occurring before the age of 30 years, there was a positive association of CD with subsequent hip fracture (HR = 2.1; 95% CI = 1.8–2.4).
In the above analyses, we only looked at first hip fracture. Risk estimates were similar when we included individuals with hip fracture prior to study entry in the analyses. Individuals with CD were then at a twofold increased risk for hip fracture after study entry (and diagnosis of CD) both when we excluded the first year of follow-up (HR = 2.1; 95% CI = 1.8–2.4; P < 0.001; based on 344 and 1125 hip fractures) and when this year was included (HR = 2.3; 95% CI = 2.0–2.5; P < 0.001; 433 and 1511 hip fractures).
CD and any subsequent fracture
There was also an increased risk of any type of fracture after diagnosis of CD (HR = 1.4; 95% CI = 1.3–1.5; P < 0.001; see also Table 2 and Figures 3 and 4). The risk estimates were not affected by stratification for sex or age at diagnosis (Table 2), although the HR for subsequent fracture of any type for those with CD diagnosed in childhood did not attain statistical significance (P = 0.052). There was no difference in risk estimates over time (HRs per calendar period: 1964–1973: 1.4; 1974–1983: 1.4; 1984–1993: 1.3; 1994–2003: 1.6; all P < 0.05). Adjustment for SEI in a subset of individuals with data on SEI did not affect our risk estimate (both crude and adjusted HR = 1.4). When we adjusted for a diagnosis of DM, depression and thyroid disease before the end of follow-up, the adjusted HR for subsequent hip fracture in individuals with CD was 1.4 (95% CI = 1.3–1.5; P < 0.001). Inclusion of the first year after diagnosis did not affect the HR for any subsequent fracture (HR = 1.5; see also Table 2).
CD and prior history of fracture
We found an increased risk of CD after hip fracture (OR = 2.0; 95% CI = 1.6–2.5; P < 0.001) and after any type of fracture (OR = 1.6; 95% CI = 1.5–1.8; P < 0.001).
Incidence of hip fracture in CD diagnosed in adulthood
The incidence figures for hip fracture per 100 000 person-years before and after diagnosis of CD are given in Table 3. We saw no notable influence on the hip fracture incidence with introduction of gluten-free diet. Instead, the incidence ratios were around 2 both prior to diagnosis of CD and afterwards, with the lower level of the 95% lower confidence interval for the incidence ratio at 1.3 or above in all time periods (Table 3). Even 20 years after diagnosis of CD did this study find an increased risk of hip fracture (IR = 2.0; 95% CI = 1.3–2.3; P = 0.001; based on 29 hip fractures in individuals with CD and 59 fractures in individuals without a diagnosis of CD). The risk estimates for any fracture showed a similar pattern (incidence ratios according to time to diagnosis of CD): ≥10 years prior: 1.4; 0–2 years prior: 2.1; 0–2 years after diagnosis: 2.3; ≥10 years after: 1.2).
|Years before or after diagnosis of coeliac disease||Person-years||No. events||Incidence/100 000 person-years||IR; 95% CI||P-value|
|10 years or more before diagnosis of CD|
|No coeliac disease||451 428||88||19||1.9; 1.3–2.9||0.001*|
|Coeliac disease||90 833||34||37|
|5 to 9.99 years before diagnosis of CD|
|No coeliac disease||142 897||133||93||1.8; 1.3–2.5||<0.001*|
|Coeliac disease||28 865||48||166|
|2 to 4.99 years before diagnosis of CD|
|No coeliac disease||87 061||158||181||1.9; 1.4–2.6||<0.001*|
|Coeliac disease||17 591||62||352|
|0.01 to 1.99 years before diagnosis of CD|
|No coeliac disease||57 774||179||310||1.9; 1.4–2.5||<0.001*|
|Coeliac disease||11 713||68||581|
|0 to 1.99 years after diagnosis of CD|
|No coeliac disease||53 114||237||446||3.0; 2.4–3.7||<0.001†|
|Coeliac disease||10 007||134||1339|
|2 to 4.99 years after diagnosis of CD|
|No coeliac disease||58 710||225||383||2.3; 1.8–2.9||<0.001†|
|Coeliac disease||11 934||104||871|
|5 to 9.99 years after diagnosis of CD|
|No coeliac disease||69 551||226||325||2.2; 1.8–2.8||<0.001†|
|Coeliac disease||14 473||104||719|
|10 years or more after diagnosis of CD|
|No coeliac disease||86 825||287||331||1.7; 1.4–2.1||<0.001†|
|Coeliac disease||18 402||104||565|
We found a statistically significantly positive association of CD with hip fracture or any fracture, irrespective of whether the fracture occurred prior to or after the diagnosis of CD. The overall risk estimates for subsequent hip fracture and any fracture were almost identical to those of West et al.14 Individuals with CD diagnosed in childhood were at increased risk of subsequent hip fracture and may also be at increased risk of fracture of any type. The duration to and after diagnosis of CD did not notably influence the risk of hip fracture. Hence, this study found no evidence that a gluten-free diet lowers the risk of hip fracture. Even 20 years after diagnosis of CD were individuals with CD at increased risk of hip fracture.
To our knowledge, this is the largest study of CD and fractures so far. The number of fractures in individuals with CD exceeded that of all earlier studies combined14–20 and hip fractures in individuals with CD were ascertained during 200 000 person-years of follow-up. We used hip fracture as our main outcome measure because it is a typical osteoporotic fracture,36 for which hospitalization is mandatory and the specificity is high (the specificity for hip fracture in the Swedish IPR is close to 100%).37, 38
We found a positive association between CD and hip fracture, and this association was independent of sex and age at diagnosis. Although, West et al.14 included individuals with CD diagnosed in childhood (13% had a diagnosis of CD made before the age of 15 years) the authors provided no risk estimate for fractures for this age stratum. Our study included more than 9000 individuals with CD diagnosed when aged 0–15 years. Fractures are common in children, but the role of bone mineral density for the risk of fractures is more debated for child fractures39 than for fractures in elderly people.40 Recent data do however indicate that childhood bone mineral density predicts the risk of future fractures.41 The current study found an increased risk of subsequent hip fracture (HR = 2.6) in individuals with CD diagnosed in childhood. To our knowledge, these findings are new. We do however urge caution when interpreting this result, because hip fractures in individuals with CD diagnosed in childhood were few. The 2.6-fold risk increase is based on 6 hip fractures/100 0000 person-years in individuals with CD in childhood compared with 2/100 000 among reference individuals. Hence, the absolute excess hip fracture risk in childhood CD is only 4/100 000 person-years. The risk estimate for any subsequent fracture in individuals diagnosed with CD in childhood had a borderline statistical significance (P = 0.052). Children often suffer high-impact trauma fractures, to which osteoporosis may play a subordinate role. Different fracture circumstances may therefore explain the difference in any fracture risk between children (HR = 1.1) and adults with CD (HR = 1.8).
The similar risk estimates for hip fracture both prior to diagnosis of CD and afterwards found in this study confirm earlier data from West et al.14 and Vestergaard et al.17 The likely mechanism for the increased risk of hip fracture in CD seen in this study is lower bone mineral density, as bone mineral density in the femoral neck is a predictor for fractures in the hip region,6 even 10 years after the measurement of bone mineral density.7 Individuals with untreated CD have lower bone mineral density at diagnosis than the general population,5, 42 and although bone mineral density improves with gluten-free diet it remains lower in individuals with CD on a gluten-free diet than in healthy controls.43 Longitudinal data suggest that little bone mass is gained after the first year of gluten-free diet,44 and this would indicate that individuals on a gluten-free diet continue to be at increased risk of fractures. This is supported by our findings of a statistically significantly increased risk of hip fractures even 20 years after diagnosis of CD. The highest HRs for hip fracture and any fracture were however seen at time of diagnosis of CD.
We used a hospital-based register to identify individuals with CD and cannot exclude the possibility that our individuals with CD had more active disease than outpatient-treated individuals with CD. Initial investigation for CD, including small-bowel biopsy or upper endoscopy, is however often carried out in a hospital setting. This is especially so in children where general anaesthesia may be required for upper endoscopy; and children constituted the majority of study participants in our study. Hospital admission during diagnostic work-up used to be common also in adults in the first part of our study period. Risk estimates for hip fracture and any fracture were similar in early and late study years, and we therefore argue that selection bias because of the use of hospital-based register data is limited. The risk estimates for fracture in this study are actually very close to those reported in a recent cohort study.14 Still, co-morbidity of individuals with CD is a potential concern of this study.3, 4, 13, 28
Although, type 1 DM is associated with a seven- to ninefold increased risk of hip fracture,29 co-morbidity with type 1 DM is unlikely to explain our findings because the risk increase for hip fracture and any fracture in this study remained after adjustment for DM, thyroid disorder or depression. Individuals with CD are also at increased risk of certain cancers,4 and certain cancers are associated with fractures.45 We were not able to reliably adjust our data for existing cancers since we did not have access to cancer data from the Swedish national Cancer Register. The lack of cancer data should however be no major drawback because the overall risk increase for cancer in individuals with CD is low (standardized incidence ratio = 1.3),4 and Michaelsson et al. have shown that only 1% of possible incident cases of hip fractures in a Swedish setting could be due to malignancy.38
The specificity for chronic diseases is generally high in the IPR. In a subset of individuals with lymphoma, the specificity for concomitant CD was above 85%.46 The sensitivity for CD in the IPR is naturally lower. Still, our study had considerable power to detect associations, as there were more than 13 000 individuals with CD participated in our main analyses. We may not have identified every case of CD, but a large proportion is likely to have been admitted considering that the population of Sweden is some 9 000 000 people and the prevalence of diagnosed CD until adulthood in a Swedish study in 1999 was only slightly more than 1/1000.47 The number of false-negative controls should not exceed 1%13 and therefore not affect our risk estimates more than marginally.
We used an internally stratified Cox regression model, i.e. a conditional statistical approach for our main analyses. This means that individuals with CD were only compared with their reference individuals matched for sex, age, calendar period and geographic area. This is important since female sex is a risk factor for both CD48 and osteoporosis; and studies of CD and fractures14 (and other outcomes)4 have indicated lower risk of complications in individuals diagnosed in recent years than in the distant past. Hence, through our study design we eliminated the potential confounding influence of age, sex, calendar period and geographic area, but we were not able to consider a number of other potential confounders such as smoking, body mass index, and the use of biphosphonates or hormone-replacement therapy. Smoking is a risk factor for hip fractures in both men49 and women,50 but inversely related to CD.51, 52 Smoking is therefore unlikely to explain our findings. This is further supported by the fact that most studies of CD and fractures have reported no difference in risk estimate after adjustment for smoking.14, 15 Besides, smoking status is strongly associated with socioeconomic position in Sweden53 and adjustment for SEI in more than 40 000 individuals did not influence our results. Nevertheless, we cannot exclude the possibility that inclusion of smoking in our model would have affected our risk estimates. In contrast, use of hormone replacement therapy may be positively associated with CD,14, 15 but negatively associated with fracture.38 Frequent use of hormone replacement therapy as reported by West et al.14 and Thomason et al.15 could therefore not explain our findings. We did not adjust for body mass index. Body mass index is often low in individuals with CD43 and is an important risk factor for fractures in elderly, and thus anthropometric data have often been regarded as potential confounders in studies of CD and fractures.14, 15, 18 However, the low body mass index in individuals with CD is probably a consequence of lower energy intake,43 malabsorption and inflammation.54 All these factors are themselves due to CD. Low body mass index could therefore be regarded as an intermediate with regards to nutritional deficiency and osteoporosis,33 and inclusion of body mass index in an analysis of CD and fracture risks underestimating the risk of fracture.
In conclusion, our study confirms previous reports of a positive association of CD with hip fracture or any fracture. This risk increase was seen both prior to CD and afterwards and is most likely due to low bone mineral density. We also suggest that CD diagnosed in childhood is associated with an increased risk of hip fracture. The lack of change of hip fracture risk over time suggests that gluten-free diet does not influence the risk of hip fracture in individuals with CD.
JFL was supported by grants from the Swedish Research Council and Örebro University Hospital while writing this article. This project was supported by a grant from The Swedish Society of Medicine, the Juhlin Foundation, the Clas Groschinsky Foundation, the Sven Jerring Foundation, the Örebro Society of Medicine, the Majblomman Foundation, the Karolinska Institutet funds and the Swedish Coeliac Society.
This project (04-030/1) was approved by the Research Ethics Committee of the Karolinska Institute, Stockholm, Sweden on the 18th March 2004.
Jonas F. Ludvigsson.
[ Log minus log plots for testing proportional hazard assumption. Parallel lines indicate that the proportional hazard assumption is not violated. y-axis scales are not identical. ]