INTRODUCTION
Every 3s, worldwide, a fracture occurs as a result of osteoporosis, mainly hip as well as non-vertebral and vertebral fractures, resulting not only in significant morbidity and mortality but a massive financial impact on society.(1,2) However, despite being such a common metabolic bone disease, osteoporosis has attracted little attention in many developing nations.
Osteoporosis is a progressive systemic disease of the skeletal system characterised by micro-architectural deterioration of the bone tissue as well as low bone mass, increasing bone fragility and susceptibility to fracture.(3) The diagnosis of osteoporosis is made by the measurement of bone mineral density (BMD) using dual-energy X-ray absorptiometry (DEXA) of the lumbar spine (LS), neck of femur and total upper femur.(4) The World Health Organisation definition of osteoporosis is a BMD 2.5 standard deviations (SD) or more below the young adult mean value for a woman (thus a T score ≤ −2.5 SD in postmenopausal women).(5) A rapid bone loss defined as an annualised bone loss rate of at least 1 SD of the sex-specific loss rate at each skeletal site is also used to determine a high risk of fractures.(6) Due to the low sensitivity and high specificity associated with the use of BMD, the Fracture Risk Assessment (FRAX®) tool is used to assess fracture risk. The FRAX® tool is an algorithm that incorporates multiple risk factors to provide a 10-year absolute fracture risk.(3)
The primary goal of pharmacological therapy in osteoporosis is to prevent fractures by improving bone strength, relieve symptoms of fractures and skeletal deformities as well as maintain normal physical function.(7) The gold standard and most widely used first line treatment for osteoporosis are bisphosphonates.(8) Unlike other therapies whose effects resolve after discontinuation, bisphosphonates are unique in that they bind to hydroxyapatite, and therefore their effects potentially persist for years after discontinuing this therapy.(9) As a result of a prolonged effect on bones, there is a concern with regards to the continued use of these drugs, mainly the possible development of osteonecrosis of the jaw (ONJ) and atypical fragility fracture (AFF).(10,11) The dose of bisphosphonate therapy in cancer patients is usually a higher accumulative dose compared to patients treated for osteoporosis which is at lower doses. However, in osteoporosis the incidence of ONJ is estimated to be only between 1/10,000 and 1/100,000, which is marginally higher than its incidence in the general population.(1)
As a consequence of these concerns, the Food and Drug Administration, in September 2011, reviewed the long-term safety and efficacy of bisphosphonate use and recommended that the use of bisphosphonates be reassessed after 3–5 years.(12,13) Thus, the concept of a drug holiday (DH) was developed. Currently, there is little evidence to support the need for a DH or to establish the effectiveness of treatment after restarting it. Furthermore, there is no evidence to guide how long to treat osteoporosis, the duration of the DH and when bisphosphonate therapy should be reintroduced.(14,15)
Three large clinical trials for the long-term use of alendronate, risedronate and zoledronic acid and their discontinuation were used as a guide on the duration of continuous therapy and the DH.(16–19) The recommendations by clinical guidelines regarding the length of treatment and initiation of a DH are based on these three trials.
The American Association of Clinical Endocrinology 2016 guidelines suggest a DH for high-risk patients after 10 years of oral bisphosphonates and 6 years of IV zoledronic acid.(20) In South Africa, the 2010 National Osteoporosis Foundation of South Africa guidelines recommend considering a DH after 5 years of bisphosphonate use for low-risk patients.(5) However, the real impact of a DH in a real-life setting versus the clinical trial setting is poorly established. Our study goal was to describe the effect of a DH on BMD at our facility.
METHODS
This study was a retrospective review of clinical records looking at the changes in BMD during a bisphosphonate DH in patients with osteoporosis attending a clinic at Charlotte Maxeke Johannesburg Academic Hospital. The data was collected during the period from January 2000 till December 2016. The population under study were all adults (above the age of 18 years) who had received zoledronic acid and/or the equivalent of oral alendronate for at least 5 years and subsequently gone on a DH with a DEXA scan done during the DH. The quantity of alcohol consumed by the patients was defined in the clinical notes by the attending physician as either nil, moderate or high. Premature menopause was defined as any woman who attained menopause before the age of 45 and male hypogonadism was defined by a recorded low testosterone level. Exclusion criteria were those who had received less than 5 yearly doses of intravenous and or oral therapy as well as those who did not have a follow-up DEXA scan after the DH. This study was approved by the University of Witwatersrand Medical Human Research Ethics Committee.
MEASUREMENTS
Demographics and clinical information, as well biochemical laboratory parameters such as serum creatinine and calcium, were collected from each of the patient's file. The presence and absence of clinical and radiological fractures as well as the site and timing together with the BMD by DEXA were also extracted from the records. The presence or absence of co-morbidities as well as other medication was noted. The effect of bisphosphonates on BMD during treatment was demonstrated by categorising BMD into yearly time points based on the duration of treatment with a bisphosphonate. The BMD values at different time points during the DH were obtained and also categorised into yearly time points at 1,2,3,4,5 and beyond. All scans were performed on the same HOLOGIC® DEXA machine. The precision error for the machine was unknown. Additional information was gathered on the indication for initiation of bisphosphonate therapy, type and duration of therapy, correct termination of therapy and adherence to the yearly blood tests, in order to assess the adherence to protocol by the attending doctors at the clinic.
STATISTICAL ANALYSIS
Demographics, risk factors and clinical features of patients included in the study were described using proportions for categorical variables, mean and SD for normally distributed continuous variables and median and inter-quartile range (IQR) for non-normal continuous variables. Characteristics of patients who had fractures during treatment and during the DH were compared to those of patients who did not get fractures using Fisher's exact tests for categorical variables and paired T-tests for a comparison of means. To investigate the effect of bisphosphonates on BMD of the radius and ulnar (RU), LS and total hip (TH), linear prediction graphs with confidence intervals of the respective BMD by time on treatment and duration of DH were plotted. Linear regression models were fitted to investigate whether the trends in BMD over time were statistically significant. Binary logistic regression models were fitted to compare patients with improved or maintained BMD versus those with a decrease in BMD during the DH.
RESULTS
A total of 97 patients (96.9% female; with 47.4% White, 30.9% Indian, 15.5% African and 6.2% Coloured) were included in this study (Table 1). The median age at the time initiation of therapy was 63 (IQR 56–68) years, with a median BMI of 26.9 (IQR 23.5–30.6) kg/m2. The median duration of treatment between the first and last BMD scan done before the initiation of the DH was 5 (IQR 4–5) years with the median duration of the DH being 2 (IQR 2–3) years. Regarding the type of therapy used, 79% used zoledronic acid, 16% alendronate and 5% used a combination of both treatments at different time points. The reasons for initiating therapy included BMD (T score ≤ −2.5) in 58%; osteopenia and a fracture history in 22%; increased bone turnover in 16% and rapid bone loss in 4% of the subjects. The majority of the patients had other co-morbidities (88.7%). Of these, 16% had rheumatoid arthritis, 16% had hypothyroidism, 6% had other connective tissues or autoimmune diseases, 6% had inflammatory bowel disease and the remaining 69% had other co-morbid conditions such as hypertension, diabetes mellitus, dyslipidaemia and asthma. With regards to glucocorticoid therapy, 29% of the patients had a history of its use. Of these, 43% had rheumatoid arthritis, 14% had asthma, and 43% had other connective tissues or autoimmune diseases.
Patient characteristics
| Demographic characteristics | N(%) |
|---|---|
| Gender | |
| Male | 3 (3.09) |
| Female | 94 (96.91) |
| Age (median, IQR) years | 63 (56–68) |
| Race | |
| African | 15 (15.46) |
| Indian | 30 (30.93) |
| White | 46 (47.42) |
| Coloured | 6 (6.19) |
| Risk factors | |
| BMI (median, IQR) (kg/m2) | 26.89 (23.52–30.61) |
| Smoking | |
| Yes | 14 (14.43) |
| No | 70 (72.16) |
| Ex-smoker | 12 (12.37) |
| Unknown | 1 (1.03) |
| Alcohol | |
| Nil | 36 (37.11) |
| Moderate | 4 (4.12) |
| Unknown | 57 (58.76) |
| Glucocorticoid use | |
| Yes | 28 (28.87) |
| No | 69 (71.13) |
| Family history | |
| Yes | 20 (21.05) |
| No | 75 (78.95) |
| Premature menopause | |
| Yes | 9 (9.28) |
| No | 85 (87.63) |
| Unknown | 3 (3.09) |
| Male hypogonadism | |
| Yes | 1 (33.3) |
| No | 2 (66.7) |
| Co-medical conditions | |
| Yes | 86 (88.66) |
| No | 11 (11.34) |
| Clinical features | |
| Duration of treatment (median, IQR) years | 5 (4–5) |
| Duration of drug holiday (median, IQR) years | 2 (2–3) |
Factors associated with decreased BMD during the drug holiday
| Demographic characteristics | Lumbar spine Odds ratio (95% CI) | P value | Total hip Odds ratio (95%) | P value |
|---|---|---|---|---|
| Age (at initiation of therapy) years | 0.94 (0.89–1.01) | 0.094 | 1.03 (0.98–1.10) | 0.233 |
| Race | ||||
| African/Coloured | 1 (Ref) | |||
| Indian | 0.50 (0.12–2.03) | 0.336 | 1 (Ref) | 0.878 |
| White | 0.58 (0.13–2.57) | 0.482 | 0.90 (0.22–3.61) | 0.640 |
| 0.70 (0.15–3.15) | ||||
| Risk factors | ||||
| BMI (kg/m2) | 0.97 (0.91–1.05) | 0.494 | (1.0.97–1.02) | 0.935 |
| Smoking | ||||
| No | 1 (Ref) | 1 (Ref) | ||
| Yes/ex-smoker | 1.43 (0.48–4.25) | 0.525 | 3.10 (0.91–10.60) | 0.071 |
| Glucocorticoid use | ||||
| No | 1 (Ref) | 1 (Ref) | ||
| Yes | 0.82 (0.24–2.79) | 0.773 | 0.88 (0.25–3.04) | 0.836 |
| Family History | ||||
| Yes | 1 (Ref) | 1 (Ref) | ||
| No | 1.24 (0.39–3.98) | 0.706 | 1.49 (0.42–5.55) | 0.537 |
| Premature menopause/male hypogonadism | ||||
| No | 1 (Ref) | 1 (Ref) | ||
| Yes | 0.67 (0.13–3.25) | 0.620 | 0.37 (0.08–1.74) | 0.207 |
| Co-medical conditions | ||||
| No | 1 (Ref) | 1 (Ref) | ||
| Yes | 2.07 (0.47–9.13) | 0.337 | 2.03 (0.43–9.54) | 0.370 |
Ref = Reference.
Effect of bisphosphonate on BMD
The mean ± SD BMD (T score) before treatment with bisphosphonates was −2.34 ± 1.01 at the LS, −1.56 ± 0.99 at TH and −2.29 ± 1.13 at the RU. The overall percentage change BMD (median and IQR) at the various sites on bisphosphonate therapy showed an increase at LS 12.90 (−2.38 − 25.93) P < 0.001, RU 8.00 (−50.00 − 14.89) P = 0.601 and TH 5.72 (−12.13 − 26.97) P = 0.297.
Figure 1 shows the relationship of the effect of time on treatment on BMD at the various sites. In relation to change in BMD and duration of treatment, there was a non-significant increase in BMD at TH (P = 0.193) and LS (P = 0.413). There was no significant change in BMD at RU (P = 0.967).
Effect of drug holiday on BMD
The overall percentage change on the effect of the DH on BMD, expressed as median (IQR), was assessed at the various sites. All three sites (Figure 2) showed a decrease in BMD with only the RU and TH being statistically significant [LS −3.13(−25.00 − 12.50) P = 0.398; RU −16.67(−26.67 − 2.13) P = 0.031 and TH −8.89(−27.78 − 10.71) P = 0.001]. There were no statistically significant changes in BMD with time on a DH at all sites [RU (P = 0.993), LS (P = 0.971)] except for a trend towards significance for TH (P = 0.058).
Predictors of change in BMD during drug holiday
The association of demographic, clinical variables and risk factors on BMD at the various sites during the DH was also assessed ( 2). There was no statistically significant risk factor associated with a decrease in BMD during the DH versus an increased or maintained BMD during this period.
Fractures during treatment and drug holiday
Nearly half of the patients, 46% (n = 45 of 97), had a history of a fracture at various sites: vertebral 13% (n = 13), non-vertebral 23% (n = 22) and at both sites 10% (n = 10). It was, however, not specified if they were fragility fractures and the exact sites of some of the non-vertebral fractures were not specified. Of these, 87% of the patients (n = 39) had fractures sustained before the initiation of therapy. Overall, 18% of the fractures (n = 8) were sustained during therapy. There was no significant risk factor associated with the fractures that occurred during bisphosphonate therapy. The remaining 9% (n = 4) of the fractures occurred during the DH, and age (P = 0.030) was the only significant risk factor in these patients. Only one patient who sustained a fracture during the DH had had a previous history of a fracture prior to initiation of treatment.
Biochemical data analysis
Of the 97 patients in the study population, only 19 had chronic kidney disease with 94.7% stage 3A (n = 18) and 5.3% stage 3B (n = 1). Prior to initiation of therapy (n = 3), 3.1% of the patients had hypocalcaemia, (n = 81) 83.5% had normal serum calcium levels, (n = 7) 7.2% had hypercalcaemia and a total of (n = 6) 6.2% were unknown. Of the seven with hypercalcaemia, only one patient had associated primary hyperparathyroidism. There were, however, 25 patients with high parathyroid hormone levels. The 25-OH-vitamin D levels for the patients prior to therapy were as follows: deficient (<30 nmol/l) (n = 22) 22.7%, insufficient (30–50 nmol/l) (n = 19) 19.6%, sufficient (>50 nmol/l) (n = 38) 39.2% and unknown (n = 18) 18.6%. Of the 22 with 25-OH-vitamin D deficiency (n = 8) 36.4% also had hyperparathyroidism and (n = 14) 63.6% had normal parathyroid hormone (PTH) levels. There were no patients with thyrotoxicosis prior to the initiation of therapy or secondary to thyroxine over replacement for hypothyroidism.
DISCUSSION
Outside the setting of clinical trials, there is limited information available in a real-life setting on the evolution of BMD and the prediction of fracture risk during bisphosphonate DH.(21,22) There is limited data in South Africa on the effect of bisphosphonate therapy on BMD in patients with osteoporosis as well as the effect of a DH on BMD. This study is the first as far as we are aware of that has assessed the effect of a DH on the BMD of patients with osteoporosis in South Africa.
In this study, there was an overall decrease in BMD during the DH at all sites measured, with a statistically significant decline at the RU and TH sites. This is in keeping with findings from the main clinical trials carried out to assess the effect of a DH on BMD.(17–19) This was also the case in a study that looked at the effect of a DH on BMD in a non-trial setting.(23) We found no significant risk factors associated with the decrease in BMD during the DH.
There is a substantial mortality associated with both hip and vertebral fractures.(24) Four patients sustained fractures during the DH, two of which were non-vertebral, one had both a vertebral and non-vertebral site and one sustained a vertebral fracture. Age was the only significant risk factor for fractures during the DH (P = 0.030). These finding differ for the Fracture Intervention Trial, Long Term Extension and Health Outcomes and Reduction Incidence with Zoledronic Acid Once Yearly Pivotal Fracture Trial extension trials which both demonstrated an increase in vertebral fracture in the placebo group during the extension period.(18,25) Caution should be taken in elderly patients with low BMD as they may fracture during the DH despite years on treatment.(26)
The efficacy of the use of bisphosphonate therapy in osteoporosis resulting in an increase in BMD and a reduced risk of fractures is well proven.(27) The greatest and only statistically significant increase in BMD on treatment, of 12.9%, was observed at LS (P < 0.001) compared to all the other sites. This is mainly because the LS is made up primarily of trabecular bone which exhibits a greater uptake of bisphosphonate during treatment than predominantly cortical bone, of which the TH and RU is made up of a greater proportion.(28)
In our study, there was a higher prevalence of osteoporosis in White and Asian population compared to the African and Coloured population. This is in keeping with previous studies that suggest an ethnic difference in the incidence of both osteoporosis and fractures among different races and ethnic groups.(5,29–31) The recently published FRAX®-based fracture probability for South Africa showed that the crude incidence of hip fractures in South Africa was highest in White (129.9 per 100,000) and Indian populations (111.7 per 100,000), then in Coloured (58.2 per 100,000) and Black populations (37.9 per 100,000).(32) However, the 10-year fracture risk probability was higher in Indian than in White South African men and women up to the age of 80 years, whereas Whites over the age of 80 years had the highest fracture risk of all ethnic groups. Coloured men and women, in turn, have a higher risk than Black South Africans.(32)
The main reason for the initiation of bisphosphonate therapy for the study cohort was a low BMD (T score ≤−2.5). However, 46% of the cohort sustained a fracture and of these 87% occurred before the initiation of bisphosphonate therapy. Precisely 22% were initiated based on the diagnosis of a history of a fracture with osteopenia. Despite the proven efficacy of bisphosphonate therapy in osteoporosis patients and fracture prevention, the treatment of patients only with osteoporosis has a limited capacity in total fracture prevention.(33) This is because fractures tend to occur in a larger group of women with BMD in the osteopenia range.(34) A recent study showed a significant risk reduction of vertebral fragility fractures in osteopenia women who received zoledronic acid than the women who received placebo.(33) Therefore, there is a need to consider early treatment with bisphosphonate therapy for patients who have high risk factors for fractures, independent of BMD as defined by FRAX®, as well as patients with osteopenia without a history of fractures. This may potentially reduce fracture risk.
The main reason for the need of a DH with bisphosphonate therapy is the fear of the severe adverse events, namely ONJ and AFF.(10,11) Studies have shown a decrease in the risk of AFF with discontinuing bisphosphonate therapy by 70% per year since the last use.(35) Hence the concept of a DH seems logical based on the pharmacokinetics of bisphosphonates and the concern for these long-term adverse events.(24) As a result, many physicians have discontinued bisphosphonate therapy automatically without taking into consideration a patient's individual risk of a fracture.(24) The importance of this is highlighted in a study that showed that the incidence per 100,000 person-years of major osteoporotic fractures in high, moderate and low risk women is 3100, 1600 and 650, respectively. This is very high compared to the risk of AFF after 2 and 8 years of bisphosphonate therapy of 2 and 78 per 100,000 person-years, respectively, and that of ONJ at just 1.03.(36) In the current study, there was no record of any of these severe adverse events. The incidence of these rare side effects is low.(1) As a result, high-risk patients with moderate-to-high risk of fractures, in whom the benefit of bisphosphonate therapy outweighs the risk of the rare adverse events, should be considered for on-going therapy.(24) The decision of who goes on a DH must be individualised, based on multiple risk factors such as a low BMD, clinical risk factors and a history of incident fractures.(37,38) Low-risk patients may, however, benefit from a DH.(20) However, there is still no consensus on the duration of a DH and how best to monitor patients once they are on a DH.(24) Despite the need for further research in this area, there is a suggestion that high-risk patients continue bisphosphonates for up to 10 years or change to an alternate therapy, and low-risk patients consider a DH after 5 years on oral bisphosphonates or 3 years on zoledronic acid.(1)
A limitation of the study was that it was retrospective in nature. There was no control group as all our patients went on a DH. Data was collected from clinical notes which at times had missing information. Also, the data obtained was in relation to patient visits which were inconsistent throughout the cohort. Despite this challenge, the information that was available was extensive and enabled the data collection process to take place. There were variable durations of bisphosphonate therapy, different bisphosphonates used, varied intervals between DEXA scans as well as the length of the DH between patients, compared to a clinical trial scenario. Another limitation was that although the same DEXA machine was used during the study period, various technicians operated the machine with potential observer error. Again, the precision error of the machine was unknown due to the retrospective nature of our study. Lastly, a single clinical practice was used and additional replication with a larger sample size from multiple centres is needed. Despite these limitations, this is the first and largest study of its kind reported in South Africa, and therefore its findings open the doorway for further, larger and possible prospective studies in this field.
CONCLUSION
The study showed a significant decrease in BMD at the TH and the distal forearm after a mean DH of 2 years. A DH should be very cautiously considered in patients with long-term use of bisphosphonates, especially in the elderly, those with a history of recent fractures, those with very low BMDs and on-going steroid use. An individualised approach based on risk factor assessment, fracture risk and BMD is key in assessing the duration of the DH. Therefore, only low-risk patients should be considered for a DH, and reassessment after 2–3 years is essential.
