Partial adherence: a new perspective on health economic assessment in osteoporosis
- First Online:
- Cite this article as:
- Kanis, J.A., Cooper, C., Hiligsmann, M. et al. Osteoporos Int (2011) 22: 2565. doi:10.1007/s00198-011-1668-0
- 275 Views
Partial adherence in osteoporosis increases the risk for fragility fracture and has considerable impact on cost-effectiveness. This review highlights a number of avenues for further research, such as improved definition of thresholds of compliance and persistence, as well as gap length, offset times, and fraction of benefit.
A number of economic models have been developed to evaluate osteoporosis therapies and support decisions regarding efficient allocation of health care resources. Adherence to treatment is seldom incorporated in these models, which may reduce their validity for decision-making since adherence is poor in real-world clinical practice.
An ad hoc working group of the European Society for Clinical and Economic Aspects of Osteoporosis and Osteoarthritis met to review key issues concerning the incorporation of partial adherence in health economic models.
Observational data have shown that poor adherence is associated with an increase in the risk for fragility fracture. Health economic modelling indicates that full adherence is associated with more quality-adjusted life years gained than partial adherence, as well as higher treatment costs and lower fracture-related costs. Although adherence appears as an important driver of cost-effectiveness, the effect is dependent on a range of other variables, such as offset time, fraction of benefit, fracture risk, fracture efficacy, fracture-related costs, and drug cost, some of which are poorly defined. Current models used to evaluate cost-effectiveness in osteoporosis may oversimplify the contributions of compliance and persistence.
Partial adherence has a significant impact on cost-effectiveness. Further research is required to optimise thresholds of compliance and persistence, the impact of gap length, offset times, and fraction of benefit.
Osteoporosis constitutes a major economic burden. It affects about a third of the female population aged over 50 years and accounts for more days in hospital and greater disability than diseases like diabetes, myocardial infarction, and breast cancer . European estimates indicate that there were 3.79 million osteoporotic fractures and 0.89 million hip fractures in the year 2000 , with total direct costs of €31.7 billion . These costs are expected to rise to €76.7 billion by 2050 due to demographic changes, notably the ageing of European populations. Health economics is becoming more important in osteoporosis due not only to the increasing requirement for the evaluation of cost-effectiveness of treatments by decision- and policymakers, but also to inform guidelines for the management of disease. An aim of cost-effectiveness analysis in the context of osteoporosis is to estimate the number of fracture events avoided by treatment and translate this into savings in terms of cost, morbidity, and mortality (commonly expressed as quality-adjusted life years, QALYs, gained) .
Poor adherence would be expected to have two effects in health economics: it would increase the risk of outcome events (e.g. the risk of osteoporosis-related fracture and subsequent morbidity) and reduce the cost of treatment. Despite this, health economic evaluation often fails to take adherence into account . This is partly because both cost and effectiveness are decreased so that the impact on cost-effectiveness is limited, and partly because health economic assessments are commonly based on efficacy data from randomised controlled trials (RCTs), which were performed according to intention-to-treat principles in populations with quite remarkable levels of adherence compared with those observed in daily practice. The absence of more reliable naturalistic data on adherence therefore reduces the validity of the conclusions that can be drawn .
It is important to understand how partial adherence affects cost-effectiveness for the proper evaluation of new drugs appearing in the therapeutic armamentarium. This is particularly relevant in osteoporosis, since there are large variations in dosing intervals, drug delivery, and tolerability between the different treatment alternatives, which may have an impact on adherence. This paper is the result of discussions by an ad hoc Working Group of the European Society for Clinical and Economic Aspects of Osteoporosis and Osteoarthritis (ESCEO) on the issue of how partial adherence could be better incorporated into health economic models in osteoporosis.
Health economics in osteoporosis
An important component of health economic assessment is the comparison of two or more treatment alternatives in patients at the same level of risk with the ultimate target of balancing benefit against cost . A related objective may be identification of subgroups of patients in whom a treatment is particularly cost-effective. The main output in this type of analysis is the incremental cost-effectiveness ratio (ICER), which is defined as the ratio between the difference in total cost of two treatments and the difference in benefits. Cost is measured in monetary terms, whilst health benefit may be measured in different ways. For instance, “cost minimization” is only meaningful for agents with similar efficacies or side effects, which is difficult to apply in a heterogeneous class like the osteoporosis drugs. Similarly, “cost per fracture saved” is unsuitable since it provides no information on whether it is better to prevent a forearm fracture or a hip fracture. Finally, “cost per life year saved” does not account for the variable loss of quality of life associated with different fracture outcomes. Moreover, none of these parameters are comparable with equivalents in other disease areas, such as cardiology or infectious disease.
With this in mind, “cost per QALY gained” has gained wide acceptance as the most utilitarian expression of ICER for osteoporosis since it combines life years lost and life years disabled (i.e. morbidity and mortality consequences in a single measure), and results can be compared across disease states. Measurement of quality of life using QALYs effectively combines mortality (where one QALY equals 1 year of life gained) and disability expressed with a utility measure (where, for example, one QALY equals 2 years of life with a utility of 0.5). Such an assessment is not without problems, such as who measures the quality of life (the patient or the physician), which scale to use, and which precautions should be applied in reducing the complex concept of quality of life to a single figure. The current practice in osteoporosis studies is to derive aged-weighted estimates using instruments such as the self-reported, preference-based European Quality of Life 5 Dimensions Index, which appears to provide a comprehensive framework to determine health status and quality of life.
There are a large number of interrelated variables that determine cost-effectiveness. These include those related to the population or patients themselves, such as the underlying risk of fracture, the age of the population, and the utilities assumed (i.e. a hypothetical or actual measure of the value that the osteoporotic patient gives to his or her health care status). The cost of treatment is also incorporated into the cost-effectiveness evaluation, as are the cost of case finding in screening programmes and the cost of care in the event of fracture. Other factors are related to the intervention itself, such as the effectiveness of the treatment, the nature of any associated side effects, the duration of treatment, and the offset time. Among this array of variables, adherence (i.e. compliance and persistence) is important, but it may also influence other parameters, or may be affected by them. For instance, non-compliance would be expected to reduce the cost of intervention and the side effects, whilst a longer offset time would attenuate the impact of poor persistence on outcomes. On the other hand, full adherence may reduce utility for the patient due to constraints associated with complying with drug intake, though this may be offset by a gain in utility due to fractures avoided.
The cost-effectiveness of treatment as measured by the ICER should be weighed against health care spending in terms of the willingness and ability to pay , which may vary among countries . Cost-effectiveness acceptability curves in a UK setting have shown no marked difference between osteoporosis agents in women aged 70 years with a prior fracture and no test of bone mineral density (BMD) . This highlights an important step of any health economic assessment, which is to identify individuals at high risk. Treating populations at greater risk for fracture would be expected to reduce the burden of disease and health care costs. In this context, the fracture risk assessment tool FRAX [3, 8] is set to improve the identification of high-risk patients and is likely to have an increasing impact in health economics.
Defining and measuring adherence in osteoporosis
Comparison of adherence across studies and treatments is not always simple due to the wide variety of differing definitions in the field. For the purposes of this paper, we define adherence and persistence below, as well as some related terms [9, 10]. Adherence is used as a generic term to cover compliance and persistence.
Compliance refers to the degree or extent of conformity to the recommendations regarding day-to-day treatment with respect to the timing, dosage, frequency, and mode of administration, i.e. it is the extent to which a patient acts in accordance with the prescribed interval and dose of a regimen . Compliance is generally measured over a period of time and reported as a percentage. In randomised controlled trials, compliance is often expressed as a pill count (e.g. the proportion of patients taking more than 80% of dispensed medication), whereas claims studies report the medication possession ratio (MPR), which is the number of doses dispensed in relation to the dispensing period. MPR is designed for oral medication, but may be adapted for once-yearly treatments by using the number of visits missed. However, data from RCTs cannot supply reliable measures of compliance in real life but may be used to create estimates and hypotheses that can be tested in real clinical practice, for example, in the phase-4 setting. Naturalistic trials on compliance are very difficult to perform but are required to evaluate the impact of adherence on cost-effectiveness.
One disadvantage of the MPR is that it fails to take into account how the patient takes the treatment. There are currently no tools that include variables such as intake instructions, though a Medication Event Monitoring System® has been used to evaluate frequency and timing of intake . In osteoporosis, this may be relevant considering the precise intake instructions for some of the agents concerned. Examples include the necessity for overnight fasting and an upright position for 30 min after intake for the oral bisphosphonates, avoiding the concurrent intake of calcium supplements and bisphosphonates, and the precise timing of drug intake for strontium ranelate. There remains room for improvement and better definition of thresholds for compliance in osteoporosis, including more accurate ways of measuring items such as missing doses, wrong timing, and non-observance of intake instructions.
Persistence is the act of continuing the treatment for the prescribed duration  and is defined as the length of time from initiation to discontinuation of therapy. Evaluation of persistence should always include a pre-specified limit on the length of time permitted between doses (gap length) to define when non-compliance becomes cessation of treatment, i.e. non-persistence. Defining these intervals is critical, since persistence can be very low if the gap length is set low and very high if it is set high. For example, in a study comparing alendronate and ibandronate, persistence was judged to be 39% and 57%, respectively, when allowing a 2-week period of grace, even though at 6 months, 73% and 80% of patients had refilled five of the six monthly prescriptions . Non-persistence is commonly considered as stopping medication for 1 month for daily or weekly treatment . Gap lengths for osteoporosis treatments with longer dosing intervals are less well defined, though stopping treatment for 2 months may be a suitable definition for a monthly treatment, and a delay of more than 3 months in the case of yearly injections was discussed by the working group.
The impact of non-persistence on outcome depends on the onset and offset times of the treatment. As regards onset time, some evidence suggests that patients who stop osteoporosis treatment before 6 to 12 months cannot expect to receive full treatment benefit . On the other hand, the offset time, i.e. the residual treatment effect after stopping treatment, is often assumed to be equal to the length of time on therapy (up to a maximum offset time of 5 years), with no difference between the agents in the osteoporosis class. This means that a patient who has been on treatment for 2 years can expect some benefit from some fracture efficacy for a further 2 years after stopping treatment, which effectively reduces the impact of non-persistence on outcome. The impact of differing modes of action on offset time is not well characterized, though current knowledge suggests that the above assumptions are reasonable for all treatments in osteoporosis [7, 15–17]. Further research should improve our understanding of offset times and the possible impact of variations in mode of action.
Determinants of adherence
There are a number of determinants of adherence [18–22]. The determinants of persistence have been studied from the patient's perspective  and found to include both internal and external factors. Internal factors include general health behaviour, habits, perceived severity and risk of osteoporosis, expectations from treatment, benefit of treatment, and barriers to treatment; external factors include the doctor–patient relationship, which is also considered to be important by the patients themselves . Poor tolerability may also have a negative effect on adherence. In this context, the impact of poor gastrointestinal tolerability of the once-daily oral bisphosphonates and the corresponding constraints of dosing instructions was one of the principal drivers behind the development of formulations with longer dosing intervals . Qualitative studies involving behavioural and psychological approaches do provide helpful information on predictors, which can help define programmes to improve adherence in patients with osteoporosis. These programmes should involve a clear explanation of the disease by the doctor and a description of possible treatment alternatives, ideally leading to the patient deciding which treatment should be prescribed [19, 24] and with subsequent close monitoring of adherence .
Impact of poor adherence on fracture efficacy
The failure of patients to adhere with treatment compromises the overall efficacy of any management strategy. Evidence from RCTs indicates that osteoporosis treatments are associated with varying degrees of efficacy in adherent patients, with between 30% and 70% reductions in risk for vertebral fracture, 15% and 20% for non-vertebral fracture, and up to 40% for hip fracture . However, differences between adherence of patients in an RCT setting and that of individuals in the real world mean that the benefits provided by these agents in RCTs may not translate into equivalent benefits in daily practice.
There is abundant evidence that adherence is not optimal in osteoporosis [20, 25–28]. In a large survey of a US managed care database, including more than 38,000 women receiving bisphosphonates or hormone replacement therapy, Huybrechts et al.  found that about a quarter of patients (26%) had an MPR ≥80% over the average follow-up of 1.7 years. Compliance declined rapidly in the first year, with the mean MPR of 66% (range, 39% to 94%) at 12 months. Huybrechts et al.  also reported very low rates of persistence: 22% of the sample was non-persistent at 1 year, and 45% was non-persistent at 5 years. Similar ranges were found in a systematic review of 14 studies by Cramer et al. , with MPRs for bisphosphonates between 59% and 81%, and rates of non-persistence at 1 year of 18% to 78% (with marginally better rates for weekly dosing than daily dosing). A smaller survey of patients receiving raloxifene reported similar rates with a mean MPR of 53% over 19 months and 19% non-persistence at 1 year rising to 23% at 3 years . These data have been recently reconfirmed with data on patients receiving bisphosphonates, strontium ranelate, or raloxifene from a large Swedish registry  and from the UK General Practice Research Database . A preliminary report of observational data with strontium ranelate suggests rates of persistence of 80% at 12 months and 70% at 2 years . To summarize these data, about 50% of patients fail to comply or persist with osteoporosis treatment within 1 year, and <60% has adequate compliance at 12 months (MPR ≥ 80%) .
In terms of outcomes, partial adherence would be expected to be associated with smaller gains in BMD, reduced impact on bone turnover, and therefore increased rates of fragility fracture, which is supported by data from RCTs and surveys in claims databases . One large survey by Siris et al.  found that 43% of more than 35,000 patients were compliant with their bisphosphonate treatment for osteoporosis (i.e. MPR ≥ 80%). By comparing fracture rates in the compliant and non-compliant patients, Siris et al. found that the risk of fractures was significantly lower in the compliant patients, with relative risk reductions of 21% for total fracture, 37% for vertebral fracture, 20% for non-vertebral fracture, and 37% for hip fracture (all p < 0.001) . Elsewhere, it has been estimated that every 1% decrease in MPR for bisphosphonates is associated with a 0.4% increase in the risk for hip fracture (p < 0.001) . Poor persistence with bisphosphonates (gap length, >1 month) was associated with risk reductions of 29% for total fracture, 40% for vertebral fracture, 29% for non-vertebral fracture, and 45% for hip fracture (all p < 0.001) . Another study reported a relative risk reduction of 60% for hip fracture in persistent versus non-persistent women prescribed bisphosphonates (hazard ratio, 0.40; 95% confidence interval 0.36–0.46, p < 0.0001) .
These results emphasize the necessity of addressing treatment adherence in order to maximize the benefits on these agents in terms of outcomes and thereby reduce the burden that osteoporosis and associated fractures place on individuals and health care systems. This should all be taken with the caveat that patients with good adherence may simply be at intrinsically lower basal risk (and vice versa). This so-called “healthy adherer” effect has been demonstrated in other fields , and there is evidence for the notion that non-adherent osteoporosis patients are high-risk takers and therefore at greater risk of fracture [36, 37]. Identifying and targeting such patients would be useful since they would have the most to gain from programmes designed to improve adherence and persistence.
Modelling adherence in health economic evaluations in osteoporosis
Given that adherence affects both treatment effects and costs, cost-effectiveness models in osteoporosis need to incorporate them in order to accurately evaluate cost-effectiveness of treatment alternatives. This has been the subject of several studies.
In a recent study, Ström et al.  addressed the impact of adherence on cost-effectiveness and explored the important drivers of cost-effectiveness. This involved the construction of an individual state transition model to compare two hypothetical osteoporosis treatments. One treatment was assumed to optimize adherence but to be 50% more costly than the other. The model assumed that partial adherence (i.e. patients who discontinued treatment and those who were non-compliant) was associated with a 20% loss of fracture efficacy compared with full adherence, i.e. the fraction of benefit of the treatment was 80% with partial adherence. If, for example, the agent reduced fracture risk by 50% in adherent patients, then the expected reduction in fracture in partially adherent patients would be 40% (0.5 × 0.8 = 0.4). The offset time was fixed to be equal to the time on treatment up to a maximum of 5 years. If the patients stopped taking their treatment in the first 6 months, then they were considered not to have benefited from treatment at all (i.e. onset time of 6 months), but to have incurred 3 months of drug cost. Other data incorporated in the model, such as health care costs, fracture risk in the population, rates of mortality, cost of intervention, and quality of life measures, were estimated according to data for the general and osteoporotic population in Sweden .
In this model, full adherence was associated not only with fewer osteoporotic fractures (−0.027) and QALYs gained (0.0379) than partial adherence, but also with considerably higher treatment costs (+€2,333), which were offset to some extent by lower fracture-related costs (−€1,621). Fewer patients needed to be treated to avoid a hip fracture in the full adherence arm (37 versus 107), indicating the potential value of programmes to improve adherence. The results showed that full adherence is more cost-effective (i.e. a lower ICER) in high-risk patients with a prior fracture . For example, the ICER for full versus partial adherence was €14,000 in patients with a T-score of −2.5 SD and prevalent vertebral fracture at age 60 years, versus €52,000 in the same patients without prevalent fracture. This was linked to the larger number of fractures avoided, for the same cost of treatment. Full adherence became cost saving above the age of 70 years. The analysis was also found to be highly sensitive to fraction of benefit, since if a full fraction of benefit (100%) was assumed in the partially adherent patients, then the fully adherent alternative became much less cost-effective.
Whilst improved adherence avoids more fractures, the model demonstrated that it also decreases the cost of treatment with a very small net impact on the ICER. However, the assumptions made regarding the offset time had a marked effect on this. For example, if the offset time was fixed at zero (i.e. no residual treatment effect), then poor adherence was not associated with a large change in ICER, since the loss of effectiveness was always accompanied by decreasing treatment costs. On the other hand, when the offset time was fixed at 5 years, there was a greater impact on ICER especially in patients at increased risk for fracture. The authors linked this to the ageing of the sample : since increasing age is associated with a higher risk for fracture, then a higher risk sample was benefiting from the residual effect of treatment over 5 years.
Ström et al. concluded that adherence was indeed an important driver of cost-effectiveness, together with offset time, fracture risk, fracture efficacy of treatment, fracture-related costs, and drug cost . They showed that the health benefits of full adherence to treatment were at least partially offset by the increased cost of treatment. They also underlined the sensitivity of their analysis to poorly defined variables such as fraction of benefit (representing non-compliance with dosing instructions) and offset times, and called for more research to clarify these measures .
More recently, a study estimating the clinical and economic burden of non-adherence with oral bisphosphonates in Belgium  integrated “real-world” estimates for compliance. Patients taking medications (i.e. persistent patients) were classified as compliant (MPR ≥ 80%) or partially compliant (MPR < 80%). The probabilities of being poorly compliant (i.e. MPR < 80%) were derived for any given year in the subgroup of persistent patients. Partially compliant patients were assumed to be associated with an increased risk of fractures [26, 33], and their drug cost was calculated using the mean MPR in the group. The impact of non-adherence was shown to be primarily driven by the issue of persistence.
Similar models have also been used to compare the cost-effectiveness of screening/treatment versus no intervention in osteoporosis. The model incorporated the concepts of adherence, primary non-adherence, and persistence. Primary non-adherence described the case where patients are diagnosed with osteoporosis, in this case due to the screening programme, but fail to take any of their prescribed treatment . Non-adherence and primary non-adherence were found to significantly increase the ICER of screening strategies (case finding) in osteoporosis.
The Markov methodology has been used to evaluate the cost-effectiveness of a variety of treatments for osteoporosis [4, 7, 15–17, 42], though the models tend to oversimplify the contribution of adherence. For example, a Markov cohort model was used to evaluate the cost-effectiveness of alendronate . A value of 50% long-term persistence was incorporated in this model, as used by the National Institute for Health and Clinical Excellence in the UK [7, 43]. Non-persistent patients were assumed to have treatment for 3 months with no health gain. Compliance was modelled using MPRs of 30%, 50%, and 70%. This model was used to conclude that both compliance and persistence had a large impact on the ICER . On the other hand, a study comparing women discontinuing treatment within 6 months versus those at 3 years indicated very little impact on the ICER . This contrasts with the models described above in which the impact on ICER depended on the assumptions made. Health economic evaluations can be expected to perform better only if they employ assumptions closer to the empirical data .
There are a number of limitations to these cost-effectiveness analyses that should be considered. First, cost-effectiveness depends on a range of factors, of which adherence is just one component. Second, the impact of programmes designed to improve compliance and persistence on cost-effectiveness could turn out to be rather small, considering the extra costs incurred implementing such programmes in the general population. This needs to be explored in more depth. Similar considerations apply to the cost of screening.
Themes for the research agenda
• Definition of thresholds for adherence according to the differing dosing intervals in the osteoporosis class
• Definition of gap lengths for non-persistence
• Definition of offset times, onset times, and fraction of benefit, notably with respect to mode of action
• Creation of tools to measure of how well patients follow intake instructions (i.e. missing doses, timing, and specific instructions)
• Exploration of the determinants of adherence and possible interventions to improve them in the population
• Further exploration of the “healthy adherer” effect in osteoporosis
• Exploration of relationship between adherence and fracture risk
• Epidemiological surveys to collect empirical data on the true cost for patients who discontinue therapy early and the impact of patients who restart therapy after discontinuation
• Further epidemiological surveys to collect country-specific data on the impact of non-adherence
In summary, partial adherence is associated with increased risk for osteoporotic fractures and has some impact on cost-effectiveness in osteoporosis. Indeed, partial adherence may well prove to be critical to the management of osteoporosis, perhaps more so than efficacy of treatment . For example, that improving persistence with bisphosphonates by 20% could have the same impact as increasing clinical efficacy by 20% . Whilst adherence should be incorporated in health economic analyses, our review has indicated that this poses a particular challenge.
This paper was derived from a working group meeting on 1 December 2010 supported by the European Society for Clinical and Economic Aspects of Osteoporosis and Osteoarthritis (ESCEO). We would like to thank the following for their valuable input to this paper: P. Alegre, E. Badamgarav, S. Corcaud, W. Dere, and B. Mitlak.
Conflicts of interest
C. Cooper received consulting fees and paid advisory boards for Alliance for Better Bone Health, GlaxoSmithKline, Roche, Merck Sharp and Dohme, Lilly, Amgen, Wyeth, Novartis, Servier, and Nycomed. M. Hiligsmann received research grants and/or lecture fees from Amgen, Servier, and Novartis. J.A. Kanis received consulting fees, paid advisory boards, lecture fees, and/or grant support from the majority of companies concerned with skeletal metabolism. V. Rabenda received research grants from Servier and Nycomed. J-Y. Reginster received consulting fees, paid advisory boards, lecture fees, and/or grant support from Servier, Novartis, Negma, Lilly, Wyeth, Amgen, GlaxoSmithKline, Roche, Merckle, Nycomed, NPS, Theramex, UCB, Merck Sharp and Dohme, Rottapharm, IBSA, Genevrier, Teijin, Teva, Ebewee Pharma, Zodiac, Analis, Novo-Nordisk, and Bristol Myers Squibb. R. Rizzoli received paid advisory boards and lecture fees for Merck Sharp and Dohme, Eli Lilly, Amgen, Wyeth, Novartis, Servier, Nycomed, and Danone.