Osteoporosis is a common disease with serious medical and economic consequences. Despite great efforts to educate healthcare professionals and the public, it remains underdiagnosed and undertreated.
BMD testing by DXA is an extraordinarily useful clinical tool for assessment of fracture risk and to diagnose osteoporosis before the first fracture occurs. DXA is the only technology for measuring BMD that can be used with FRAX, the World Health Organization fracture risk assessment algorithm that is becoming widely used throughout the world. Several organizations recommend serial DXA testing for monitoring pharmacologic therapy of osteoporosis.(1–3)
The utility of BMD testing to monitor therapy was questioned almost a decade ago when the concept of “regression to the mean” was raised.(4) Although this concept has relevance at a population level, it was subsequently refuted as misleading and irrelevant to the clinical management of individual patients.(5,6) A study published recently in BMJ by Bell et al.(7) raises the question anew. Despite the title, “Value of Routine Monitoring of Bone Mineral Density after Starting Bisphosphonate Treatment: Secondary Analysis of Trial Data,” the authors conclude that monitoring BMD “in postmenopausal women in the first three years after starting treatment with a potent bisphosphonate is unnecessary and may be misleading.” The authors go on to state that “routine monitoring should be avoided in this early period.” These conclusions are based on a secondary analysis of pooled data from the two arms of the Fracture Intervention Trial (FIT) in which postmenopausal women with low BMD were randomized to alendronate or placebo.(8,9) They conclude that using BMD to monitor response to treatment with alendronate was of no value because (1) >97% of the patients on treatment ultimately showed an increase in BMD and (2) the within-subject variability was considerable. Several of the same authors used a similar approach in a post hoc analysis of the Perindopril Protection Against Recurrent Stroke Study (PROGRESS)(10) to conclude that monitoring the initial blood pressure response after perindopril (an angiotensin-converting enzyme inhibitor) therapy was unnecessary.(11) The purpose of this commentary is to address issues raised by Bell et al. and to place the need for BMD monitoring into an appropriate clinical context.
Whereas we applaud all efforts to apply the best available medical evidence to clinical decision-making, the validity and applicability of evidence should be closely scrutinized before making recommendations.(12) The conclusion of the recent BMJ article, that monitoring therapy with BMD testing is unnecessary, rests on four assumptions: (1) the goal of monitoring is to document effectiveness by showing an increase in BMD, (2) the increase in BMD in virtually all treated subjects in FIT can be expected to occur in patients in clinical practice, (3) the response to one bisphosphonate in a clinical trial is indicative of the response to all bisphosphonates in clinical practice, and (4) within-person variability obscures detection of the BMD response to treatment. We believe that all of these assumptions are incorrect or unfounded as described below.
The goal of therapy for osteoporosis is to reduce the risk of fractures. Although, in untreated individuals, low BMD is a good predictor of fracture risk,(13) and there is a statistically significant relationship between BMD increase in response to therapy and reduction in fracture risk,(14) the relationship between the increase in BMD from therapy and reduction of fracture risk is not strong and probably not linear.(15)
Patients on treatment for osteoporosis whose BMD remains stable or increases seem to benefit equally, at least in regard to the rates of new vertebral fractures.(16–18) It is therefore not necessary to monitor BMD to determine how much a patient's BMD increased on treatment. However, in these same analyses, patients on treatment whose BMD decreased had higher rates of fractures compared with those on treatment whose BMD was stable or increased. (Bell et al. misinterpreted one of the studies they cited; they suggest that fracture risk reduction was seen in patients who lost BMD on alendronate, but that study actually showed no significant reduction in fracture risk for patients on treatment whose BMD decreased by the least significant change [LSC] or greater.(18)) Thus, there is clearly a need to identify patients whose BMD decreases despite treatment.
The assumption that virtually all patients who are treated with alendronate would show a significant increase in BMD, as seen in this analysis of the FIT data, does not apply to patients in clinical practice. Patients in clinical trials are different from patients that we treat in the “real world.” Dowd et al.(19) matched key characteristics of the patients they were treating in their osteoporosis clinic with the entry criteria for four different osteoporosis clinical trials. In the best case scenario, 80% of patients being treated in their practice would have been excluded from participation in one of the trials, whereas in the worst case scenario, 97% would have been excluded from another (too old, too young, too many medications, etc.). In FIT, patients were excluded if they had dyspepsia requiring ongoing treatment, prior peptic ulcer disease, or “major medical problems that would likely preclude participation for three years.”(8) In the real world, such patients are not denied treatment for osteoporosis.
Compared with patients in the “real world,” subjects who are enrolled in clinical trials are more likely to take their medicine regularly and correctly and to persist with treatment. Trial subjects probably do better with other lifestyle factors such as calcium, vitamin D, and exercise. In FIT, >85% of subjects were still taking their study drug at the end of the trial; 96% of those still taking their study drug took it at least 75% of the time. In contrast, a study of patients in clinical practice showed that <45% were compliant with prescription refills and only 20% were continuing treatment with bisphosphonates after 24 mo.(20) Patients in FIT who had significant bone loss in the lumbar spine or total hip (≥8% over 1 yr, ≥10% over 2 yr, ≥12% over 3 yr) were dropped from the trial(9) and so were not included in the analysis of Bell et al. Patients in clinical trials are supplemented with calcium and vitamin D, which are not always provided in clinical practice; in fact, we have seen patients stop taking their calcium and vitamin D supplements when they are started on prescription medication for osteoporosis. Moreover, patients in clinical trials are screened for secondary causes of osteoporosis and excluded if these are found. In practice, secondary causes of osteoporosis are common,(21,22) yet many patients go untested and untreated. A study of clinical practice patients showed that ∼10% of those treated with oral bisphosphonates had a significant loss of BMD over 1–2 yr, and many of the BMD-losers had a previously unrecognized contributing disease.(23)
Bell et al. extrapolate their analysis of FIT (which used 5 mg of daily alendronate for the first 2 yr—one half the usual prescribed dose) to make generalizations regarding all older women taking all currently clinically relevant doses and dosing intervals of alendronate and other orally administered bisphosphonates (risedronate or ibandronate). This extrapolation is probably not appropriate. Interestingly, some of these same authors made a similar extrapolation from perindopril to all angiotensin-enzyme converting inhibitors in their earlier article on blood pressure monitoring.(11) It is instructive to look at the “best case” response to alendronate or risedronate in the Fosamax-Actonel Comparison Trial (FACT).(24) The “responder” analysis looked at the percentage of these carefully screened and compliant women who had what would in clinical practice likely be called a significant decline in BMD. At 1 yr, 3% of women treated with alendronate and 7% of women treated with risedronate had a significant decline in BMD at the total hip, and, at the femoral neck, significant decreases were seen in 9% of those treated with alendronate and 14% of those treated with risedronate.(25) Because not all patients who are prescribed medications for osteoporosis will maintain or have significant increases in BMD, monitoring with DXA can be used to identify patients who have significant decreases in BMD on therapy, decreases that can be the result of nonresponse to the medication, poor compliance, poor persistence, incorrect dosing, malabsorption or secondary causes of osteoporosis that were either unrecognized before starting treatment or developed after treatment was initiated.
The dangers in extrapolating the results of the analysis of Bell et al. to clinical practice are further evident when examining patients treated with generic alendronate. A recent study has shown significant differences in the in vitro disintegration rates of generic alendronate from eight different manufacturers.(26) Others have suggested lesser gains in BMD (40–50% lower) when generic alendronate was compared with branded alendronate and risedronate.(27)
Although within-subject variability is not an important part of their argument against serial BMD for monitoring treatment, that should still be addressed. Bell et al. stated there was “considerable within-person variation” that was statistically significant between patients in the treatment and in the placebo arms. The actual variance was 0.012 g/cm2 in the treated group and 0.014 g/cm2 in the placebo group. Although this difference may be statistically significant, it is not clinically relevant. With a mean BMD of ∼0.700 g/cm2, this “considerable” variation amounts to 1.7–2.0%, which is less than the variation seen in commonly used clinical measurements such as height, blood pressure, blood sugar, and cholesterol. Very few biological variables can be measured with the same precision as BMD. The International Society for Clinical Densitometry (ISCD) has established, updated, and promoted quality standards for BMD testing, including precision assessment to determine reproducibility. For serial BMD measurements, the ISCD has emphasized the concept of LSC to distinguish between random variation and true biological change.(6,25,28) This important concept has not been embraced in other areas of clinical medicine that rely on quantitative measures for clinical decision-making. The application of precision and LSC to serial monitoring of BMD by expert densitometrists raises bone densitometry to a higher level than most quantitative measures in clinical medicine. It also means that the within-person variation cited by Bell et al. would not confound or confuse the interpretation of serial BMD measurements by those who appreciate the limitations.
Strategies to improve poor adherence with pharmacologic treatment are desperately needed. Persistence was increased by a nurse visit(29) and by positive reinforcement with bone turnover markers.(30) Whether repeating BMD measurement after 1 or 2 yr of treatment might improve adherence is unknown; more study is needed to evaluate to what degree compliance and persistence are affected by repeating BMD after 1 or 2 yr. In the absence of such studies, it is premature to conclude that repeat densitometry is of no value. Currently, bone densitometry is the only technical examination to fill the clinical vacuum of follow-up in patients who embark on a long-term treatment regimen.
In our opinion, the assumptions and conclusions of Bell et al., based on a post hoc analysis of FIT, are seriously flawed. Even with the most generous interpretation, their findings are only applicable to the unique population of subjects studied, treated with the 5-mg daily dose of alendronate used in FIT, and should not be generalized to the same drug taken in different doses or other drugs in the same class in the heterogeneous population of patients seen in clinical practice. Although it is unknown whether repeating BMD improves adherence with treatment, the true purpose of monitoring BMD in patients treated for osteoporosis is to identify the small but substantial number of patients who experience a significant decrease in BMD. These patients should be considered for further evaluation to search for the cause or causes of their decline in BMD, which may include poor compliance with the treatment program, deficient calcium/vitamin D intake, malabsorption, confounding diseases/disorders/medications with adverse skeletal effects, or true nonresponse to the drug. It is clear that BMD response rates in FIT are not representative of those seen in clinical practice. We suggest that until there is good evidence to do otherwise, clinicians should consider a follow-up BMD test ∼1 yr after starting pharmacologic therapy for osteoporosis and thereafter at intervals determined by individual patient circumstances.