Increased risk of hip fracture among Japanese hemodialysis patients

Abstract

Incidence of hip fracture in dialysis patients is significantly higher than that in the general population. As information is lacking about Asian dialysis patients, we compared the incidence of hip fracture in hemodialysis patients with that in the general population in Japan. We conducted a retrospective cohort study using panel data from the Japanese Society for Dialysis Therapy registry. The study included patients without history of hip fracture who received hemodialysis three times per week as of December 31, 2007. We compared the observed number of hip fractures to the expected number derived from a national survey, and calculated standardized incidence ratios (SIRs) and the incidence rate difference. Subgroup analysis was performed according to vintage and diabetic status. During the one-year study period, 1,437 hip fractures were recorded in the 128,141 hemodialysis patients (61.9 % male). The overall incidence was 7.57 and 17.43 per 1,000 person-years in men and women, respectively. The SIRs for male and female patients were 6.2 [95 % confidence interval (CI) 5.7–6.8] and 4.9 (95 % CI 4.6–5.3) compared to the general population, and remained nearly constant until 16 years vintage, but increased steeply thereafter. The incidence rate difference of hip fracture increased with age. The SIRs for diabetics of both genders were higher than those for non-diabetics. Our study provides additional evidence that hip fracture risk among Asian dialysis patients is also significantly higher than in the general population.

Appendix

In this Appendix we provide a mathematical basis for the estimation method of the incidence of hip fracture in this study.

If x is the observed number of individuals with the outcome of interest (i.e., hip fracture onset), and Py the number of person-years at risk, the incidence rate (IR) is given by

$$ {\text{IR}} = \frac{x}{\text{Py}} $$

The number of person-years at risk is obtained as a sum of the patient’s time at risk who suffered hip fracture (Pfy) and did not (Pny).

$$ {\text{IR}} = \frac{x}{\text{Py}} = \frac{x}{{{\text{Pny}} + {\text{Pfy}}}} $$

If a patient did not suffer a hip fracture, the patient’s time at risk of these patients (Pny) is equal to the patient’s survival time during the 1-year study period (Psny).

If a patient suffered a hip fracture, we assume the patient’s time at risk of these patients (Pfy) is equal to half of Psfy the patient’s survival time during the study.

$$ {\text{IR}} = \frac{x}{\text{Py}} = \frac{x}{{{\text{Pny}} + {\text{Pfy}}}} = \frac{x}{{{\text{Psny}} + (0.5 \times {\text{Psfy}})}} $$

This assumption is considered reasonable as no seasonal trends associated with the incidence of hip fractures have been detected in the general Japanese population [ref. 8].

In addition, as only a small number of hip fractures were reported, the total patient time at risk estimated using this method did not significantly change when this assumption was changed as below. For male hemodialysis patients,x = 595, Psny = 78,319.5388 years, Psfy = 592.75 years,

$$ {\text{IR}} = \frac{x}{\text{Py}} = \frac{x}{{{\text{Pny}} + {\text{Pfy}}}} = \frac{x}{{{\text{Psny}} + (0.5 \times {\text{Psfy}})}} = \frac{595}{78319.5388 + (0.5 \times 592.75)} = 7.57 \times 10^{ - 3} $$

If we assume that all fractured patients were suffered on the last day (i.e., equal to the patient’s survival time during the study), the IR was calculated as below

$$ {\text{IR}} = \frac{x}{{{\text{Psny}} + (1 \times {\text{Psfy}})}} = \frac{595}{78319.5388 + (1 \times 592.75)} = 7.54 \times 10^{ - 3} $$

If we assume that all fractured patients were suffered on the first study day (i.e., January 1st), the patient’s time at risk of these patients (Pfy) is calculated as follow

$$ {\text{Pfy}} = (1 \div 365) \times x $$

Thus, the IR was calculated as below

$$ {\text{IR}} = \frac{x}{{{\text{Psny}} + ((1 \div 365) \times x)}} = \frac{595}{78319.5388 + ((1 \div 365) \times 595)} = 7.60 \times 10^{ - 3} $$

Same as above, for female hemodialysis patients,x = 842, Psny = 47,879.5 years, Psfy = 832.58333 years,

$$ {\text{IR}} = \frac{x}{{{\text{Psny}} + (0.5 \times {\text{Psfy}})}} = \frac{842}{47879.5 + (0.5 \times 832.58333)} = 17.43 \times 10^{ - 3} $$

If we assume that all fractured patients were suffered on the last day (i.e., equal to the patient’s survival time during the study), the IR was calculated as below

$$ {\text{IR}} = \frac{x}{{{\text{Psny}} + (1 \times {\text{Psfy}})}} = \frac{842}{47879.5 + (1 \times 832.58333)} = 17.29 \times 10^{ - 3} $$

If we assume that all fractured patients were suffered on the first study day (i.e., January 1st), the patient’s time at risk of these patients (Pfy) is calculated as follow

$$ {\text{Pfy}} = (1 \div 365) \times x $$

Thus, the IR was calculated as below

$$ {\text{IR}} = \frac{x}{{{\text{Psny}} + ((1 \div 365) \times x)}} = \frac{842}{47879.5 + ((1 \div 365) \times 842)} = 17.58 \times 10^{ - 3} $$

In conclusion, application of this estimation method did not significantly affect the study results according to our trial calculation. Our assumption is considered reasonable (Fig. 4).

Fig. 4

Standardized incidence ratios (SIRs) stratified by diabetic status and vintage for men and women, standardized for age