Background Management of osteoporosis is imperfect because patients may not start, persist or comply with treatment. This study was aimed to identify baseline variables associated with women failing to start, persist or comply with bisphosphonate treatment.
Methods 254 women >50 years old were selected 5 years after having a bone densitometry (bone mineral density (BMD)) diagnosis of osteoporosis. At the outset, information about osteoporosis was sent to the general practitioner (GP). Women were not under pressure at the outset to start or comply, and they and their GP were unaware that follow-up studies would take place. Patient survival was identified from the National Health Service Strategic Tracing Service, prescription data from GP records and baseline data from the initial questionnaire. Persistence was defined as at least one prescription issued per year and compliance as having a medicine possession ratio of ≥80% for each of 5 years.
Results 38% failed to start treatment. Failure was associated with higher BMD Z score and residence in a nursing/residential home. Half of those starting and alive at 5 years persisted with treatment, whereas only 23% achieved a medicine possession ratio of ≥80%. Persistence was associated with a lower comorbidity index and compliance with a lower BMD Z score and a fall before starting treatment.
Conclusions Treatment was low, especially in nursing/residential homes where known low treatment prevalence appears to be associated with non-initiation. The degree of depression of BMD (not just low BMD) was associated with better initiation and compliance. The association of falls with compliance suggests that fall clinics may be able to play a part in improving osteoporosis management.
- forearm bone density
Statistics from Altmetric.com
Large studies interrogating dispensing databases have shown that a significant proportion of patients starting oral bisphosphonate therapy fail to carry on taking this drug after 1 year.1–3 Continuation with therapy after 2 years has been the subject of a few studies, even though persistence with treatment is important for fracture prevention.4 5 While persistence is important for a drug effect, maximising compliance is important for optimal fracture reduction.6 Compliance may be assessed from dispensing database studies,1 6 but the use of these facilities may introduce significant biases. The socioeconomic status of the population that can access drugs may be restricted, and in many studies that make use of drug therapy databases, the reason for commencing drug treatment, or a dual-energy x ray absorptiometry (DXA) measurement, is not recorded. The only study investigating compliance over 5 years came from the USA1 and was limited by the use of patients from a specific (and low) socioeconomic group; however, although all were undergoing osteoporosis treatment, only one-fifth of patients had had a bone mineral density (BMD) scan.
Previous UK studies have examined early persistence at 1 year by telephone calls,7 the importance of dosing frequency on compliance8 or the use of general practitioner (GP) databases to identify compliance over 1 year.9 To date, there is no accepted standard for establishing when enough medication has been taken to be clinically effective. However, previous studies suggested that patients reporting compliance for at least two-thirds of days had better BMD than those patients reporting less.10 11 Compliant use (medicine possession ratio (MPR)≥80%) with bisphosphonates has been shown to be associated with a 16%–45% decrease in the risk of osteoporotic fracture compared with non-compliant use.4 5 There is further evidence that compliant use of bisphosphonates (MPR≥80%) beyond 3 years may be necessary before the possibility of discontinuation can be considered.12
The aims of this study were to identify, in women found to have osteoporosis on distal forearm scanning (determined on the basis of 90% device-specific cut points as described later), factors that were associated with starting treatment, persisting with treatment and complying with treatment for osteoporosis over 5 years. We have done this by exploiting a large database of patients scanned at the forearm between 1996 and 2002 and studied a sample in detail.
Patients and methods
Between 1996 and 2002, approximately 13 000 women aged ≥50 years had been identified as requiring densitometry by their GP. They had undergone forearm bone densitometry at their GP's surgery and had completed a general health and lifestyle questionnaire. The GP was advised that women with osteoporosis should be started on appropriate treatment. Other than stating that their patient had osteoporosis and that treatment was advised, no further directions (eg, type of treatment) were given to the GP. In the population studied, no charge was made for the forearm scan and women 60 years old and above received treatment free of charge. No indication was given either to the GP or to the patients that any follow-up would take place.
At least 5 years later, a sample of 10% of these women were identified. The GP surgery was contacted, and information was obtained about whether a patient was alive or had died. Patient records were accessed at the surgery, and treatment data were recorded.
Bone mineral density
BMD had been measured at the non-dominant distal forearm site (unless a wrist fracture had occurred on that side) by single-energy x ray absorptiometry (Osteometer DTX100, Osteometer, Rodøvre, Denmark). Reference values had been established previously for healthy female participants (no fracture history and no bisphosphonate, corticosteroid or hormone replacement therapy use ever) aged 50–98 years comprising 342 participants aged 50–59 years, 295 aged 60–69 years, 479 aged 70–79 years, 167 aged 80–80 years and 10 participants aged >89 years.
A distal forearm BMD of <0.34 g/cm2 was considered to indicate osteoporosis. This value had been previously ascertained as including 90% of participants whose BMD at L2–4 or femoral neck was more than 2.5 SDs below peak mass (<T−2.5) as determined by axial DXA and hence having osteoporosis.13
A previously validated and detailed questionnaire filled out at the time of the bone scan collected information concerning health, lifestyle and current medication, both medically prescribed and self-purchased drug use. For each patient, a Townsend score14 (a socioeconomic deprivation index) was determined using the individual's postcode; a Charlson comorbidity index15 16 (the presence of comorbid diseases) was determined using the list of diseases stated in the completed questionnaire; BMD Z score was calculated (as described above), and the date of death, where applicable, was obtained from the National Health Service Strategic Tracing Service.
At follow-up 5 years later, electronic GP records were used to collect information on bisphosphonate prescriptions issued for all women originally identified as having osteoporosis by forearm BMD. Medication information included the number of prescriptions written by the GP, drug name, dosage, number of pills supplied and date of each prescription refill. Each patient was assigned an index date based on the first prescription written by the GP for a bisphosphonate after their bone scan. No information was available on whether bisphosphonate prescriptions issued by the GP were subsequently dispensed to the patient at a pharmacy or indeed whether the patient actually took the drug. The issue of a bisphosphonate on GP records was used as evidence of persistence or compliance.
Patients were evaluated for both long-term persistence and compliance with bisphosphonate treatment over 5 years. Persistence with treatment (ie, continuous use) was defined as the time from initiation (index date) to discontinuation of treatment (first failure) either by a complete termination or by an extended interruption in treatment (of greater than 12 months). The evidence of at least one bisphosphonate prescription issued per year of follow-up was a measure of persistence.
Prescription issue data from the first year after the index date and all subsequent years were used to measure annual rates of bisphosphonate compliance. Compliance with bisphosphonate therapy was measured based on the medicine possession ratio (MPR). We have defined MPR as the number of prescriptions written by the GP for bisphosphonate medication divided by the number of prescriptions expected to be written in a fixed time (eg, 1 year). In this study, patients with an MPR >80% for each of the 5 years of follow-up were defined as having a high level of compliance.
Forearm BMD measurements were expressed as areal density or as Z scores (number of standard deviations above or below the age-related mean for healthy participants). Differences between two groups were determined by Pearson χ2 statistic for categorical variables (or if appropriate, Fisher's exact test) and for multiple groups by analysis of variance for continuous variables. Non-parametric comparison of two groups was determined using the Mann–Whitney U test.
Characteristics examined in univariate analysis were living in a nursing/residential home, history of diabetes, suffering from indigestion even intermittently; current prescription medication ≥3 items; “ever use” of oral corticosteroids, lifetime history of fracture and after age 45, history of fracture in first degree relatives, current smoker, alcohol use (>2 units/week), a fall in the 3 months before bone scan, maximum exercise per week (immobile/potter around house vs walk outside/go to shops at least once) and use of a walking aid as dichotomous variables. Age, BMD (as Z score), weight, Townsend score and Charlson comorbidity index were included as continuous variables.
Multivariate analysis examined the association between independent factors recorded at the time of forearm BMD scanning with subsequent initiation, persistence and compliance of bisphosphonate treatment. Because of the relatively small number of participants, only those factors found to be significant in the initial analysis were included in backward logistic regression analysis. The regression analyses were age adjusted when appropriate. Odds ratio (OR) and corresponding 95% confidence intervals (CIs) are reported. Relative risks (RRs) were also calculated for those categorical variables that were found to be positive on regression analysis to investigate how many more times more likely an outcome may be if the variable was present.
All treatment comparisons were two sided, and statistical significance was taken at the 5% level in all analyses unless otherwise stated in the text. All analyses including correlation and regression analyses were performed using SPSS V.16.0.
Data for 19 out of 1299 women randomly selected for follow-up were unavailable. Out of 1280 women, 254 had BMD at the forearm <0.34 g/cm2 (having osteoporosis). Women with BMD<0.34 g/cm2 were older (77.2 (8.3) vs 64.5 (9.7) years, p<0.001) and lighter (65.5 (12.2) vs 69.3 (13.6) kg, p<0.01).
Initiation of treatment after forearm BMD scan
After scanning, 157 (62%) out of 254 women with osteoporosis were started on bisphosphonate treatment by their GP. Characteristics of women who never started a bisphosphonate appear in table 1. These women (n=97) were older, had a higher mortality rate over the subsequent 5 years, used a walking aid and had greater morbidity.
After backward stepwise elimination of variables (see the Methods section) using those significant in table 1 (excluding alive at 5 years, see footnote), only non-residence in nursing/residential homes (OR, 21.71 (95% CI, 6.36 to 74.11), RR 7.9) and a low Z-score (1.88 (1.19 to 2.98)) remained significant.
Discontinuation of bisphosphonate therapy
One hundred and fifty-seven women had at least one prescription for a bisphosphonate after their bone density scan (table 1). Of these 157 women, 37 (22.9%) had only one prescription for a bisphosphonate. Out of 34 women in nursing/residential homes, only 3 received a prescription, and for each, only one prescription was written.
At the end of the first year of treatment, 149 women were alive, of whom 74% were persisting with treatment and 44% were achieving an annual MPR ≥80% (table 2).
Five-year persistence with bisphosphonate therapy
Persistence with treatment (defined as at least one prescription per year, see the Methods section) decreased with time (table 2). After 5 years, 111 out of the original 157 participants who had started treatment were still alive and remained in the study. Of these 111 women, 50% were persistent with treatment over 5 years including patients who transferred from one drug to another (figure 1).
Univariate analysis of characteristics of women who persisted with treatment for at least 5 years included younger age (73.3 (7.1) vs 76.2 (5.9) years, p<0.05), not smoking (5/55 vs 16/47, p<0.01) and a lower Charlson index (categories 1–3 vs categories 4 and greater, p<0.01) compared with women who only persisted with treatment for a maximum of 1 year. In subsequent multivariate analysis, only a lower Charlson Index at the time of BMD scan remained significant (OR, 2.33 (1.36–3.97)). RR of starting with lower Charlson index categories was 1.86 compared with higher categories.
Compliance over 5 years
Compliance with bisphosphonate treatment was assessed annually for MPR ≥80% (see the Methods section) and continued to fall over the study period (table 2). At year 5, an annual compliance of MPR ≥80% was recorded in 23% of women (table 2). A flow chart of treatment is shown in figure 1.
Univariate analysis did not reveal any significant characteristics associated with a high level of compliance. Backward elimination of variables with a 20% level of significance (p<0.2) showed that lower Z score (2.93 (1.07–8.00)) and a fall in the 3 months before bone scan (OR, 12.20 (1.17–125.00), RR 1.9) were significantly associated with an MPR ≥80%.
A low bone mineral density, even within a population already identified as having osteoporosis, was a significant associate of starting treatment. The proportion starting treatment (62%) compares well with a US study of similar numbers, in which only 40.6% (140 women starting drug treatment out of 344 determined to have osteoporosis) of women identified as having osteoporosis started treatment, although this proportion increased to 56% if an interview with a healthcare provider materialised.17 Our study did not identify whether an interview with the GP took place, but the uptake compares favourably with the US experience.
Other risk factors for fracture that we studied, which included those used in the FRAX questionnaire,18 were not significantly associated with initiation. Both our results and those of the US study17 show that neither a history of fracture nor a family history of fracture is associated with starting treatment. Both studies might have achieved significance for history of fracture with greater numbers. The CIs of the calculated OR in the US study of 247 women only just fell below unity, and using our data, a study of approximately 480 women would be required to achieve significance with a power of 80%. Family history of fracture seems less likely to be a factor, with the US study not achieving any significance at all. Other associations found in the USA are not directly comparable with our study. Patients with higher income in the USA were likely to start treatment, but the Townsend score (an index of deprivation) was not associated with initiation in the present study. This may reflect the difference in accessibility and cost in healthcare systems. However, a low bone density result appears to be a greater encouragement to start treatment. A Canadian study, which included patients starting treatment when either having osteopenia or osteoporosis, also showed that results from a DXA scan were associated with initiation,19 far exceeding other factors such as low body weight, short stature or a previous hip, wrist or vertebral fracture. The performance of a BMD measurement may be an important consideration in the overall management of osteoporosis. In some situations (eg, women older than 75 years), knowledge of bone density is deemed unnecessary before starting treatment.20 Our data suggest that knowledge of having a low bone density encourages initiation of treatment in a population in the eighth decade of life, and lower BMD appears to be a common factor across nations for encouraging initiation of treatment. A factor militating against initiation of treatment was observed in a nursing residential home (table 1). Aspray et al21 found that few people in nursing/residential homes were on osteoporosis treatment, and the present study indicates that this may be because very few patients are even started on treatment although BMD is low. Discrimination against this population is not appropriate since many live for 3–5 years,22 23 the incidence of hip fracture is high24 and alendronic acid increases BMD in women in long-term care facilities.25
Persistence with treatment declined with time, particularly in year 1, as in other studies.3 26 27 Over the full 5-year period, the proportion being issued at least one prescription per year had fallen to 50%. This proportion is similar to that of 48% in a study of persistence over 5 years in the USA.1 This study was confined to patients with a low income ($10,000–20,000) and included both men and women. In common with the study of Solomon et al,1 we also identified that a higher comorbidity index was associated with worse persistence. Non-modifiable baseline variables that were used by Solomon et al1 to assist in improvement in persistence, such as sex and race, were not applicable to our study. A history of fracture was not a factor in our study. Few patients started treatment in nursing/residential homes, but the finding that none had more than one treatment suggests that, together with the poor initiation rate, oral treatment of osteoporosis is a problem in this setting.
Whereas persistence with treatment confers some benefits, optimal efficacy is only achieved with high compliance rates with an MPR ≥80%. High levels of MPR are necessary for reduction in osteoporotic fracture,4–6 and achievement of such levels may allow an interval in treatment without prejudicing future fracture rates.12 Such intermissions in treatment may be beneficial because of the risk of subtrochanteric fracture with alendronate.28 In our study, only 23% of patients achieved an MPR ≥80% compared with approximately 36% having an MPR>66% in the US study.1 The value of a bone density is highlighted both by the US study in which a BMD measurement contributed to a high MPR1 and by the present study in which a low BMD value at the outset encouraged compliance.
Good compliance was also associated with a previous fall as well as with low bone density. This combination generates a powerful tool for osteoporosis management that could be exploited in fall clinics and develops the interaction of falls and osteoporosis. Bone densitometry measurements will serve to enhance both initiation of and compliance with treatment in a setting which already captures a population with a factor (falling) that is associated with compliance with treatment for osteoporosis. A search for osteoporosis is a worthwhile quest in fall clinics for both primary and secondary prevention. It is notable that a previous fracture was not associated with compliance, although it only achieved significance (OR 1.13) in the USA.1 As with initiation, bone densitometry is of greater significance than a previous fracture for compliance.
This work highlights low initiation rates, poor persistence and compliance with oral bisphosphonate treatment over 5 years, all of which are likely to result in suboptimal fracture prevention.6 If oral treatment is pursued, further work is required both to ascertain whether failure of initiation relates to the prescriber or the patient and to establish causes of discontinuation and to investigate methods of ensuring high compliance rates. Previous studies over 1 year with the selective oestrogen receptor modulator, raloxifene, indicated that use of a bone turnover marker improved compliance and that intervention of a nurse was still better.29 If high compliance cannot be achieved by the oral route, annual intravenous injection of a bone resorptive agent might be studied as a more cost-effective strategy.
Our study does, however, have several limitations. We were unable to determine the reasons for non-initiation of treatment and could not establish whether the decision not to start treatment was taken by the GP, by another person with a medical background, by the patient or by a patient's relative. The reasons for this decision, such as side effects, difficulty with following instructions for taking the drug or a desire to avoid polypharmacy, were not sought either. The same applied to discontinuation of treatment. Compliance with bisphosphonate treatment may have been overestimated, as we were unable to validate whether patients actually used their bisphosphonate prescription correctly.
This may be a highly selected cohort of patients and may not be representative for the whole of the UK. Inclusion biases may have been introduced because of the practitioners being interested in the subject.
Our study has several strengths. We have specifically chosen women in whom diagnosis of osteoporosis has been positively made by bone densitometry. Unlike clinical trials, this retrospective database analysis did not directly interfere with the GP's intention to treat patients diagnosed with low bone density nor did this study influence the patient's medication-taking behaviour. Using electronic records to conduct follow-up of all patients, we were more certain about associations for discontinuation. In studies based on health insurance statistics, those becoming uninsured (for whatever reason) are rejected and can account for 30% of those on the database.3 We were also able to investigate a cross-section of society, whereas those insured are probably drawn from higher socioeconomic groups.
What is already known on this subject
A BMD measurement encourages starting treatment for osteoporosis.
Few patients in nursing/residential homes are on any treatment for osteoporosis.
One follow-up compliance study over 5 years has been reported. Not all patients had bone density, and all came from a low socioeconomic group.
What this study adds
Patients in nursing/residential homes are not on treatment because they are never started even though they have osteoporosis.
Low BMD measurements, within a population with osteoporosis, are associated with greater initiation of treatment.
Low BMD measurements are the most important factor associated with starting and complying with treatment, more so even than a previous fracture.
Less than a quarter of patients starting treatment have optimal compliance after 5 years.
A history of falls is associated with better compliance, further supporting the concept that fall clinics may have a role in encouraging osteoporosis treatment.
Funding This work was supported by funding from the Bone Disease Foundation.
Competing interests MWJD has acted as consultant, given lectures supported by and served on advisory boards for MSD, Procter & Gamble, Novartis, Roche/GSK.
Ethical approval This study was scrutinised and accepted by the RJAH Institutional Review Board and the North East Wales Local Research Ethics Committee. REC:06/WNo03/2.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.