Goal-directed treatment of osteoporosis†
For a Commentary on this article, please see McCloskey and Leslie (J Bone Miner Res. 2013;28:439-441. DOI: 10.1002/jbmr.1859).
Drug treatment for osteoporosis typically begins with an oral bisphosphonate, regardless of initial bone mineral density (BMD) or fracture risk, and decisions to continue or change treatment are often based on evidence of response to treatment based on changes in BMD, bone turnover markers, and occurrence of fractures. This pattern differs from preventive therapy for other conditions, such as hypertension, where treatment is based on achieving a goal. We propose that a goal be established to guide treatments to reduce fracture risk. The goal could be a certain risk of fracture or level of BMD. Goal-directed treatment would individualize the initial choice of treatment based on the probability that alternatives would achieve the patient's goal. In contrast to changing treatments based on years of use or failure to respond, the patient's BMD and risk would be reassessed periodically and decisions to stop or change treatment would be based on achieving or maximizing the chance of reaching an acceptable level of fracture risk or BMD. The acceptance of goal-directed treatment and application to practice will require a consensus on a number of issues about goals along with models of fracture risk while on treatment and probabilities of achieving goals. The result could be more rational and effective use of the expanding array of treatments for osteoporosis. © 2013 American Society for Bone and Mineral Research.
Drug therapy for osteoporosis is usually initiated based on bone mineral density (BMD) and fracture risk, considering clinical risk factors for fracture, especially age and prior fracture. Treatment typically begins with an oral bisphosphonate. Many physicians then monitor for response to treatment by periodically measuring BMD and sometimes a bone turnover marker (BTM). Patients with a significant loss of BMD, no expected change in BTM level, or having a fracture during treatment are often considered treatment failures, nonresponders, or inadequate responders and may be changed to a different treatment.1 When bisphosphonate therapy is deemed “successful” (ie, stability or improvement in BMD, the expected reduction in BTM level, and no fracture), therapy may continue indefinitely, although recent concerns about limited effectiveness and risk of adverse effects from long-term treatment has led many to stop therapy after about 5 years. Other treatments with rapid reversible effects are generally continued indefinitely.
This pattern of treatment differs from preventive therapy for other conditions, such as hypertension and hypercholesterolemia, where treatment is based on achieving a goal. For example, the goal of treating hypertension in patients who do not have risk factors for cardiovascular disease is a blood pressure less than 140/90 mm Hg.2, 3 Goals of treatment for these conditions are feasible because a wide variety of treatments allow achievement of goals for most patients. Continued treatment is required to maintain the goal, unless it can be maintained by changes in weight, diet, or exercise.
In contrast, treatment of osteoporosis has no clear consensus goals for BMD, BTMs, or fracture risk. Some clinicians aim to keep markers of bone resorption in the lower half of the reference range for premenopausal women to maximize the effectiveness of antiresorptive therapy.4, 5 The lack of a BMD goal may reflect modest effect of most current antiresorptives on BMD. As the options for treatment for osteoporosis expand, it is likely that more powerful agents will result in substantially greater increases in BMD, perhaps with greater reduction of fracture risk.
A Goal for Osteoporosis Treatment
We propose that a goal be established to guide treatments to reduce fracture risk. The goal could be a certain risk of fracture or level of BMD. A goal of treatment would guide selection of an initial treatment. A goal would also guide decisions to change or stop treatment. In both instances, decisions about selection or change in drug therapy would be based on the probability that the treatment would bring the patient to her/his goal.
For example, imagine that there is a consensus that the goal of treatment is to reduce the 5-year risk of all clinical fractures to less than 10%. Consider a hypothetical example of a 60-year-old woman who has a 15% 5-year risk of clinical fractures and femoral neck T-score of −2.6. Drug treatment would be recommended. The choice of treatment may depend on the probability that the treatment would achieve her goal and the time it would take. A bisphosphonate might be chosen because it would afford her, say, more than a 50% probability of reaching either goal of risk less than 10% within 3 years of treatment. Follow-up would focus on reassessing the patient's risk of fracture and the probability she would reach the goal by continuing current treatment. Suppose that after 3 years of bisphosphonate, she had reached the fracture risk goal with an estimated 8% 5-year risk of clinical fractures. Goal-directed guidelines would recommend that she stop drug treatment because bisphosphonates are likely to have a residual antifracture effect, at least for a while, after discontinuation. Follow-up would include periodic reassessment of her risk relative to her goal and restarting treatment if her risk exceeded the goal. Alternatively, the goal could be based on BMD, say, a femoral neck T-score higher than −2.0. At her starting value of −2.6, it could be estimated that starting a bisphosphonate would give her a 30% chance of reaching that goal within 3 years. Achieving that level might lead to a recommendation to stop.
In contrast, consider the hypothetical case of a 75-year-old woman with a recent vertebral fracture, a history of a humerus fracture, a femoral neck T-score of −3.2, and a 30% 5-year risk of fracture. Currently, a bisphosphate would usually be considered “first-line” treatment. In contrast, goal-directed treatment would base the selection of initial treatment on the probability of reaching goals: a T-score above −2.0 or a 5-year fracture risk less than 10%. She has very low probability of achieving either goal, even with 10 years of bisphosphonate therapy. A more potent treatment, such as a bone-forming agent, would be preferred because it may substantially increase her chance of reaching her goal of treatment.
Nevertheless, assume that she starts taking a bisphosphonate. Reassessment after 3 years of bisphosphonate therapy finds that she has had no fractures or changes in risk factors and her femoral neck T-score has improved to −2.8. This would currently be considered an adequate response to therapy by most clinicians. However, from the perspective of goal-directed treatment, her 5-year fracture risk may have improved to 20% but remains well above the 10% goal, and her T-score falls well short of −2.0. The probability of achieving the goals by continuing bisphosphonate therapy could be estimated and would certainly be very low. In the paradigm of goal-directed treatment, this low probability would be an indication for switching to or adding a more potent agent.
Issues With Establishing Goals for Treatment
Setting treatment goals raises several issues. For example, should a goal be a certain BMD or T-score value? If so, what should that value be? Or should it be an acceptable level of fracture risk? If so, what should that be? Should the goal of therapy vary according to baseline age, BMD/T-score, or fracture risk? Should BTMs be included as a treatment goal?
Fracture Risk as a Treatment Goal
Treating to achieve the goal for lower fracture risk reflects the fundamental purpose of therapy: to reduce fracture risk. This approach assumes that the decision to start treatment is based on the patient's risk of fracture. The choice of treatment would be based at least in part on the probability that the patient would reach his/her goal with that therapy. Once the patient begins treatment, the patient's fracture risk would be re-estimated periodically. This reassessment would be based on changes in BMD, the occurrence of fractures during therapy, and changes in important risk factors, such as a new chronic disease, weight loss, or a new medication. Reassessment of risk may also involve vertebral imaging, as with vertebral fracture assessment (VFA) by dual-energy X-ray absorptiometry (DXA) to detect asymptomatic vertebral deformities. These would be used together to recalibrate the patient's risk.
Clinical decisions about continuing or changing therapy would be based on whether the patient has achieved the goal and—if not—the probability of reaching the risk goal with continued or alternative treatments within a reasonable amount of time. If the probability of reaching the goal is low, the choice of alternative treatments could be considered in part on the basis of the probability that they would achieve the patient's goal. Bisphosphonate therapy may be stopped when a goal is reached because its effect on BMD and risk may last for several more years. However, if a goal is reached as a result of treatments with drugs that have reversible effects, such as teriparatide or denosumab, sustaining that goal would require continued treatment or switching to a course of a bisphosphonate.
Goal-directed treatment based on risk of fracture will require establishing a consensus risk goal, models for assessing the risk during treatment, and methods for estimating the probability of achieving risk goals by continuing or changing treatments.
Requirements for Establishing Fracture Risk as a Treatment Goal
A fracture risk goal requires a decision of which types of fractures to include. It is attractive to include all types of fractures, including radiographic vertebral fractures, because this captures the whole burden of fractures and all types of fractures appear to respond to most antiresorptive treatments.6 It may be convenient to use the four major osteoporotic fractures included in FRAX (clinical spine, hip, wrist, and proximal humerus) because that risk has become familiar to many clinicians and the initial assessment of risk and subsequent reassessments would be based on the same fractures.
Risk requires a time interval. Risk on treatment may use shorter intervals than 10 years used in FRAX because physicians and patients may be most concerned about the effects of treatment over the next few years, and because there are few data about risk of fractures during long-term treatment.
Goal-directed treatment would need methods for assessing risk of fracture during treatment. Risk models, such as FRAX, may retain their value for predicting risk if a reduction in risk of hip fracture resulting from the specific treatment is considered.7 However, some risk factors that develop during treatment may have different meaning for the risk of future fractures on treatment. For example, we speculate that the occurrence of a fracture while on treatment might indicate that the patient has a higher risk compared with the risk of a history of fracture before initiation of treatment. Fractures during follow-up will generally be recent fractures that indicate a greater risk than a more remote history of fracture included in general risk models.8, 9 Loss of BMD despite treatment might suggest a greater increased risk of future fracture during treatment than indicated by the resulting BMD value alone.10, 11
Estimates of risk during treatment will also depend on the effects of the specific treatment. The risk of various types of fracture, such as vertebral or hip fracture, may be reduced soon after treatment begins. For some agents (eg, alendronate, risedronate, and raloxifene), the reduction in risk may be greater than would be estimated from the BMD on treatment. For denosumab and zoledronate, the change in femoral neck BMD appears to accurately estimate the reduction in risk of nonvertebral fracture.
In goal-directed treatment, selection of initial treatment and subsequent changes in treatment would be guided by probabilities of achieving a goal risk. These probabilities could be derived from the databases of large clinical trials and their extensions. The probabilities may depend on a patient's characteristics such as age, weight, initial BMD, or measurements of BTMs.
The frequency and content of monitoring risk would need to be defined. Occurrence of events, such as a fracture or stroke, may trigger reassessment of risk. In the absence of such events, optimum intervals for finding clinically meaningful changes in risk from aging and changes in BMD could be developed from the databases from large clinical trials.
BMD as a Treatment Goal
A goal to achieve a level of BMD is attractive because BMD is easily measured and resembles the familiar treatment goal like blood pressure or cholesterol level. BMD predicts fracture risk, although its value for predicting risk during treatment has not been adequately studied. Reaching a T-score greater than −2.5 is appealing because it might be said that the patient “no longer has osteoporosis.” There is evidence that patients who have achieved that level of BMD in the femoral neck may have risk of subsequent fracture that is low enough that little benefit could be expected from continued treatment, therefore stopping treatment could be considered.12, 13
The probability that a patient would achieve the BMD goal could be assessed at baseline or at later points during treatment with algorithms derived from analyses from various clinical trials. Reasonable intervals between BMD measurements could be estimated; for example, BMD would be measured sooner in patients close to goal than in those who are far from the goal.14, 15
However, a BMD goal raises several limitations and issues. If the purpose of a goal is to achieve an acceptable fracture risk, the value of BMD for predicting fracture during treatment is poor for alendronate, risedronate, and raloxifene16–18 but better for denosumab and zoledronate.19, 20
In addition, if BMD is the goal, then a consensus would be needed about which skeletal site(s) of BMD measurement should define the goal. Almost all trials and fracture risk models have femoral neck BMD. On the other hand, because treatment is often started based on the lowest T-score of several skeletal sites, it may be appealing for some to have goals that match that practice. However, the various goals and probabilities of reaching goals would be more complex and supported by less evidence.
Goals that involve BMD will require that measurements be done with excellent quality control, reproducibility, and consistency from one BMD measurement to the next.21–24
Goals That Combine Risk and BMD
BMD and fracture risk might be combined into simple algorithms. In the treatment of hypercholesterolemia, the goal for low-density lipoprotein cholesterol is <100 mg/dL if the patient has coronary heart disease, <130 mg/dL if the patient has two or more risk factors, and <160 mg/dL if one or none.25 Analogously, two or three different BMD goals could be set based on a patient's risk of fracture. For example, a BMD goal would be higher, such as T-score of −1.5, if a patient has a vertebral fracture or 2 strong risk factors, and −2.0 for others.
Perhaps the goal may be achievement of either a fracture risk or BMD goal. This flexibility is appealing. However, it would lead to instances where the patient had achieved one but was far from the other goal with contradictory guidelines about whether to continue or change treatment.
Other Issues About Setting a Goal
The most fundamental issue is how to establish a goal. Eliminating risk is impossible. The goal might best be described as an “acceptably low risk” of fracture. It would be simple for the treatment goal to be the same as the indication for treatment. However, that would lead to awkward situations where patients just below a goal BMD could receive only brief and intermittent treatments as they crossed back and forth across the line. The same rapid shift may result because trials suggest that risk decreases soon after treatment is initiated. Therefore, once treatment is initiated, it is appealing to have a goal to improve BMD and reduce risk more substantially, if for no other reason than to provide a cushion against the rising risk with aging. This suggests that treatment goals should represent lower risks and higher BMD values than used as indications for treatment.
Ideally, randomized trials with fracture endpoints will compare the effectiveness of the assumptions that underpin goal-directed treatment. Specifically, it would be valuable to show that in patients who have a very high fracture risk or very low BMD, a potent agent reduces fracture risk to a substantially greater degree than does an oral bisphosphonate. Similarly, we need evidence in patients who remain at very high risk or have very low BMD despite years of bisphosphonate therapy that switching to a more potent agent reduces fracture risk to a greater degree than continuing. Such trials have been done with BMD endpoints. Comparative efficacy trials with fracture endpoints are planned or underway. The process of setting goals of treatment will point out additional research that is needed to define specific goals.
Because the costs, efficacy, and safety of treatment varies, goals might differ by specific treatment or class of drug. For example, very inexpensive treatments would generally allow setting goals at lower levels of risk than for expensive drugs. This could lead to several goals for each of several treatments. However, goal-directed treatment would be more acceptable if there are one or very few goals based on one or two general classes of treatment. If cost-effectiveness analyses are used to define goals, these results would need to be combined with judgments about the clinical and practical implications of goals based on economic analyses. As with all guidelines, individual patients may feel differently about risk of fracture and the effects of treatments, so goals need flexibility to incorporate those considerations, and the patient's risks from treatment may influence the choice.
A goal may not be feasible for some patients. Their initial risk may be so high or their BMD so low that the probability of reaching a goal may be nearly zero. This might be recognized and the patient treated with the most potent treatments and combinations and for longer durations. Alternatively, it would be possible and perhaps less frustrating to allow goals to be individualized. However, in this situation, achievement of a goal would not change the effort to reach better levels of BMD and risk.
Should goals depend on the patient's age? On one hand, it seems logical that an acceptable level of risk should be acceptable regardless of age. On the other hand, risk goals may be much easier for 60-year-olds to achieve and impossible at age 85 years. Similarly, a certain BMD value or T-score, such as T > −2.0, will generally be more difficult for elderly than younger patients to achieve and having goals that are not specific to age would lead to more intensive treatment for elderly than for younger patients.
A goal will need to be simple and compelling to patients and clinicians. In the paradigm of goal-based treatment, the probability of reaching a goal would be the basis for initial treatment selection and subsequent decisions about changing treatment. The data supporting these probabilities will need to be strong. The estimates would need to be simple and very easily made and incorporated into medical practice and electronic medical record systems. The probability of reaching the goal is a novel concept that would need discussion, education, and simple and intuitively appealing presentation to patients and physicians.
Putting goal-directed treatment into practice may require changes in payments for the patient care. Instead of first-line treatment regardless of risk or BMD, payments for initial treatment must allow prescription of appropriate agents for high-risk patients as indicated by goal-directed guidelines. Such changes may require evidence about the comparative efficacy of treatments and the value of personalizing the initial selection of treatments. In addition, as for care of patients with hypertension who have reached their goal blood pressure, payments should cover ongoing follow-up and reassessment of patients who have achieved a goal.
Achieving a Consensus About Goal-Directed Treatment
Achieving guidelines based on goals will require a consensus of experts and organizations that issue clinical policies. The process will start with discussions of and agreement about the numerous issues outlined above. Formulation of acceptable goals and models for estimating the probabilities of achieving goals will need analyses of data from large trials. Setting goals also involve analyses of the cost-effectiveness of stopping, continuing, or changing treatment at various levels of risk or BMD. Implementing goals will require agreement by groups that generate guidelines and reimbursement policies. Ideally, but ambitiously, a consensus could be reached in time to guide the use of potent agents that could achieve goals even in patients with very high fracture risk or very low BMD.
Goal-directed treatment with clinical decisions based on the probability of reaching a goal would change the initial selection of treatment, the way the success of treatment is assessed, and decisions about stopping, continuing, or changing treatment. Compared with the approach of simply designating several agents as first line and others as second line, goal-directed treatment would individualize the choice of treatment on the patient's risk and/or BMD. In contrast to changing treatments based on years of use or failure to respond, this paradigm would guide decisions to stop or change treatment based on maximizing a patient's chance of reaching an acceptable level of BMD or fracture risk. The acceptance of goal-directed treatment and application to practice will require a consensus about a number of issues about goals and new models of risk on treatment and probabilities of achieving goals. This will require effort by a broad collaboration among clinicians, scientists, patients, and their organizations. The result could be more rational and effective use of the expanding array of treatments for osteoporosis.
SRC has received consulting/advisory board fees from Eli Lilly, Merck, and Amgen. FC has received consulting/advisory board fees from Eli Lilly, Novartis, Merck, Unigen, Tarsa, and Amgen; lecture fees from Eli Lilly, Novartis, and Amgen; grant support from Eli Lilly and Novartis; and grant support for clinical trials from Amgen, Eli Lilly, and Novartis. RE has received grant and honoraria for speaking and consultancy from Novartis, Amgen, Eli Lilly, Merck, Johnson and Johnson, Warner-Chilcott, and Ono Pharma. IRR has received honoraria and research support from Novartis, Merck, Amgen, and honoraria from Genzyme. MM has no conflicts of interest. EML has received consulting/advisory board fees from Eli Lilly, Novartis, Merck, Warner Chilcott, GSK, and Genentech; and grant/research support from Amgen, Merck, Eli Lilly, Novartis, Warner Chilcott, GSK, and Genentech.
Authors' roles: SRC wrote the first draft. All authors reviewed and commented on subsequent drafts and approved submission of the manuscript.