History of fracture after the age of 45–50 years is associated with a doubling of the risk of a new fracture, the risk being even higher after a vertebral fracture.1 Some studies found a recent history of fracture to be a greater risk factor for subsequent fracture than a fracture that had occurred longer ago.2^–7 These were studies on short-term fracture risk prediction,5 6 limited to selected subgroups of fractures6 7 or to low-trauma fractures in subjects older than 60 years.3

As these studies suggested clustering of fractures in time, we analysed the time that elapses between a first and subsequent clinical vertebral and non-vertebral fracture in a long-term study from menopause onwards, including low-trauma and high-trauma fractures. The same cohort was used as in two previous studies, but for the present study we analysed the time relation annually and for all fractures from menopause onwards.2 4

METHODS

Participants

In a cross-sectional population-based study (1992–1994), 4140 postmenopausal women aged 50 to 80 years registered with general practices completed a questionnaire on risk factors for osteoporosis, fracture history and onset of menopause. The methodology of this study has been described in detail elsewhere.8 At 10 years after baseline, all women who were traceable and alive (n = 2847) were invited to complete a questionnaire about fracture history and 2372 women participated (83%).

Measurements

The women were invited to complete the questionnaire about their fracture history2 4 during the 5-year and 10-year follow-up study.2 4 The research assistant confirmed all fractures reported by the patients by searching the medical files in the participating general practice centres. In the Netherlands, clinical fractures are treated in the hospital and then reported to the patient’s doctor.

Statistics

Cox regression models were used for the statistical analyses, taking into account the fracture and death hazard. For the first fracture risk analysis, the follow-up period was calculated as time between the onset of menopause and first fracture. For the re-fracture analysis, the follow-up period was calculated as time between the first and the subsequent clinical fracture. All fractures that could be confirmed were used in the analyses, including the fractures of the women who dropped out. The dropout was known after 5 and 10 years, however, the yearly dropout had to be estimated because data about death were only collected during the 5-year and 10-year follow-up studies and not in the time between these two follow-up studies.

First, relative risks (RR) with 95% CIs of subsequent fractures compared with first fracture during follow-up were calculated for intervals of 0–1, 0–5, 0–10 and 0–20 or more years after the first fracture as well as for specific intervals of 2–5 years, 6–10 years and >10 years after the first fracture.

Second, we calculated the percentages of all first fractures that occurred as well as the percentages of all subsequent fractures based on the time intervals from menopause or first fracture onwards, respectively.

Third, one-way analysis of variance (ANOVA) was used to analyse whether the follow-up period was significantly different by women with multiple first fractures and first fracture location. Fractures were classified according to the groups used by the World Health Organization (WHO)9 into hip, major (clinical vertebral, forearm, humerus) and minor (other) fractures. In addition, age-adjusted multiple Cox regression models were performed to analyse the effect of the variables mentioned above on a second fracture.

SPSS software (V. 15.0; SPSS, Chicago, Illinois, USA) was used for the statistical analyses.

Ethics

The Medical Ethics Committee of Maastricht University and the Maastricht University Hospital approved the study (reference no. MEC 94–196.1).

RESULTS

A total of 4140 postmenopausal women with a known fracture history participated in the present study. The women who completed all questionnaires had significantly more first and subsequent fractures than the women who only completed the baseline questionnaire (25.5% vs 17.4%, p value <0.001) and were significantly younger at baseline (61.6 years vs 64.5 years (p<0.001). The calculated dropout rate was 1.6% per year and resulted from death or non-participation by participants or doctors.

Of the 4140 women, 924 (22.3%) sustained a first fracture from menopause onwards, with a maximum follow-up of 45 years (mean (SD) 19.2 (9.3) years) and included 79 620 person years. The absolute risk for a first fracture increased linearly with an annual increase of 1.0% per year (tables 1 and 2).

Of the 924 women with a first fracture, 243 (26.3%) sustained a subsequent fracture, with a maximum follow-up of 40 years (mean 8.9 (7.6) years) and included 8181 person years. The 1-year absolute risk for subsequent fracture during the first year after the first fracture was 6.1% (95% CI 4.5 to 7.6) and changed exponentially over time, resulting in 40.2% (95% CI 35.4 to 44.0) over a 20-year period (tables 1 and 2).

Of all first fractures, 4.3% (95% CI 3.0 to 5.6) occurred annually from the onset of menopause onwards. A total of 23% of all subsequent fractures (95% CI 18.1 to 28.0) occurred in the first year after the first fracture, 31.3% (95% CI 25.8 to 36.8) in the next 4 years and 25.9% (95% CI 20.8 to 31.1) during the next 5 years (tables 1 and 2 and fig 1A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 (A) Percentage of all first (grey line) and subsequent (black line) fractures. (B) Relative risk of all subsequent fractures calculated as a mean from the first fracture (grey line) and per separate year of follow-up (black line).

Compared with the risk of a first fracture, the RR of subsequent fracture over the entire follow-up period was 2.1 (95% CI 1.7 to 2.6) (fig 1B). The RR was highest in the first year after the first fracture (RR 5.3; 95% CI 4.0 to 6.6) and remained significantly elevated during 15 years of follow-up (RR 1.7; 95% CI 1.2 to 2.2). Compared with the risk of a first fracture, the RR for the specific intervals of 0–1, 2–5, 6–10 and >10 years was highest in the first year and remained significantly elevated within the next 4 years (RR 2.8, 95% CI 2.0 to 3.6), the next 5 years 1.4 (95% CI 1.0 to 1.8), but not after >10 years (RR 0.4; 95% CI 0.3 to 0.5).

Using the data of the 924 women with a first fracture, the years of follow-up between a first and second fracture were not significantly different for single or multiple fractures (mean years: 8.9 and 7.5, respectively; p = 0.279) and fracture location (mean years: 7.8 for hip, 8.7 for major (clinical vertebral, forearm, humerus) and 9.3 for minor (others); p = 0.144).

Subsequent fracture risk was similar in patients whose first fracture was a hip fracture (n = 37, hazard ratio (HR): 1.1; 95% CI 0.8 to 1.6), a major fracture (n = 118, HR: 1.2; 95% CI 0.9 to 1.5) or a minor fracture (n = 88, reference). The age-adjusted analyses showed the same results.

DISCUSSION

The results show that clinical vertebral and non-vertebral fractures cluster in time from menopause onwards. A total of 23% of the subsequent fractures occurred within 1 year after the first fracture, and 54.3% occurred within 5 years.

This increased fracture risk immediately after a first fracture has been reported by others after vertebral fractures,6 osteoporotic fractures of hip, humerus and clinical spine7 and low trauma fractures in men and women older than 60 years.3 When calculated from first fracture onwards our findings are comparable with the studies by Johnell et al7 and Center et al3 When calculated for specific time intervals (year 0–1, 2–5, 6–10, >10) our results are similar to the results of Center et al who found that the subsequent fracture risk was no longer different from the first fracture risk by the end of a 10-year follow-up period, possibly reflecting a healthier subset, as also suggested by Center et al3 Our current results extend these observations to all clinical fractures, from menopause onwards, on a yearly basis and for low-trauma and high-trauma fractures.

Klotzbuecher et al1 who calculated pooled RRs between 1.8 and 2.0, as well as the developers of the new WHO algorithm,9 10 indicated that risks could wane off over time. Our results show that the long-term RR of 2 is driven by the high short-term RR (fig 1).

The time intervals from menopause and first fracture onwards were used, but also the specific time intervals. This distinction has clinical consequences. For example, for patients presenting immediately after a first fracture, it is possible to calculate the risk of subsequent fracture after 1, 5, 10 years or lifetime risk. However, if a woman does not seek medical attention immediately after her first fracture, the risk over the next years will be lower, as many subsequent fractures may already have occurred. In addition, in most randomised clinical trials the patients are selected based on a low T score and/or a prevalent vertebral fracture. In only one study, patients were selected based on a prevalent hip fracture. In that study zolendronate decreased not only subsequent fracture risk, but also mortality. We therefore need more studies on fracture prevention in patients selected only based on a prevalent non-vertebral fracture, other than the hip. Meanwhile, most guidelines propose a BMD measurement to decide about drug treatment in patients presenting with a fracture.11

This fluctuation in fracture risk might be caused by falls and fall-related risks that increase the subsequent fracture risk after a first fracture. We have no data on fall-related risk factors from the baseline study or on the number of falls during follow-up. However, Van Helden et al12 showed that patients with an early second fracture after the first fractures were recurrent fallers, as they found a temporarily increase in fall accidents after a first fracture. Other possibilities for fractures might by immobility (due to plasters, bed rest), comorbidity, increased bone turnover and different coping styles of the patients with a fracture.12 13

Our study had several limitations. First, 274 women (6.6%) received bisphosphonates after the baseline assessment. However, excluding these women from the analysis did not change the results (data not shown). Second, only the fractures reported by the patients were confirmed in the medical files, without validating the negative reports. However, Ismail et al14 concluded that self-reported incidence of fractures produced accurate information about the occurrence of most fractures.14 Third, since we only included clinical vertebral fractures, our results are not valid for morphometric vertebral fractures.15 Fourth, 98.5% of all women in our study population were Caucasian, and our results need confirmation in other ethnic groups. Finally, we do not know the exact dropout rate. The dropout was only checked every 5 years, which resulted in an unknown date of death for most deceased. Our calculated dropout rate was low, presumably because we included women from menopause onwards, so their mean age was lower than in other studies.

In conclusion, our results indicate that from menopause onwards, clinical fractures cluster in time. One out of four of all subsequent fractures occurred within 1 year after a first fracture, and one in two occurred within 5 years, indicating the need for early action after a first fracture to prevent subsequent fractures.