The value of routine BMD monitoring after starting bisphosphonate treatment

Our paper in the British Medical Journal on the value of monitoring bone density after starting bisphosphonates1 has been the topic of a highly critical Perspective by Watts and colleagues in this Journal.2 Our analysis of the FIT trial found that when yearly bone mineral density (BMD) testing is performed, the between-person variation in treatment effects was small, and treatment was beneficial for most patients, and the within-person variation was large so that the actual treatment effects often were masked. We now address the points raised by Watts and colleagues' Perspective.

We agree with Watts and colleagues that the goal of therapy is to reduce the risk of fractures. They cite evidence that the relationship between BMD and fracture risk is not as strong as is often assumed, but this actually lends support against monitoring BMD. Although Watts and colleagues agree that monitoring is not needed if a patient's BMD is stable or has increased, they say that monitoring is needed to detect if there has been a decrease in the patient's BMD. The paper cited by Watts and colleagues to support this notion3 actually shows similar vertebral fracture risk reduction for women on alendronate whether their one-year change in spine BMD was a 0% to 4% loss [odds ratio (OR) = 0.40, 95% confidence interval (CI) 0.16–0.99], a 0% to 4% gain (OR = 0.49, 95%CI 0.30–0.78), or a greater than 4% gain (OR = 0.46, 95%CI 0.32–0.66). Women who lost more than 4% also appeared to benefit from therapy (OR = 0.15, 95%CI 0.02–1.29); the fact that this result was not statistically significant may be due to the very small numbers included in this category (only 55 women) rather than a lack of effect of treatment.

Watts and colleagues' second point is that women in FIT do not represent women in the “real world.” While it is theoretically possible that some adherent patients in clinical practice may not respond in the same way as patients in randomized, controlled trials (RCTs), empirical evidence would argue against this with a recent analysis suggesting similar treatment efficacy in adherent community populations compared with trial populations.4 Similarly, we believe that our findings may be applied to women in clinical practice who adhere to therapy.

We agree that adherence in the general population is likely to be lower than that in a clinical trial. However, because of the large background within-person variation, monitoring BMD is likely to be a poor method of detecting non-adherence and should not be used for that purpose. In any case, most problems with adherence to osteoporosis treatments occur within 3 months,5 much earlier than the first monitoring for response at one year. A better method for detecting nonadherence is to simply ask patients if they are having problems taking therapy.6

Watts and colleagues state that our findings should not be extrapolated to other potent oral bisphosphonates, suggesting that there is a greater chance of a “significant decline in BMD” with risedronate and generic alendronate than there is with branded alendronate. Evidence presented to government pharmaceutical subsidy bodies demonstrates that the effects of these medications in post-menopausal women may be considered equivalent.7, 8 However, if there is real concern that other potent oral bisphosphonates may have larger variations in treatment effect between patients than branded alendronate, then we need studies to estimate BMD variation for these drugs using either mixed models of individual patient data9 or a simpler method using summary data.10 In this way we would have the evidence needed to decide whether to monitor individuals' BMD on these drugs rather than just opting to monitor when there is no evidence to support this decision and substantial evidence against it.

Our findings do not in any way negate the importance of BMD measurements to assess fracture risk or to determine when therapy is indicated. As is the case for other conditions where the abnormality worsens with time (such as age-related loss of renal function and lung function11), random within-person variation in BMD is likely to be more of a problem for monitoring than it is for initial diagnosis. Notwithstanding Watts and colleagues' statement that BMD is more precise than many other clinical variables, there is considerable noise in BMD (the amount of random within-person variation is especially apparent when the variation is considered as SD of g/cm², a more statistically appropriate measure than CV% for BMD, where measurement error is not proportional to the mean12). However, when BMD is used to decide whether to start a woman on treatment, the “noise” in measurement is many times smaller than the “signal”: loss in BMD compared with peak bone mass some 30 years before. This means that overall there is a large signal (loss in BMD) to noise (random within-person variation) ratio.

In contrast, for monitoring BMD in the first few years after treatment, the signal is much smaller: the effect of treatment in altering the decline in bone density. And the noise component is much larger: Random within-person variation is doubled because we are using change between two observed measurements rather than one, and there is also the between-person variation in longer-term BMD changes in the absence of treatment. This means that it is very difficult for a clinician to know what effect therapy has had for his or her patient on the basis of an observed change in BMD. Because of the poor signal-to-noise ratio, BMD is unlikely to be useful for monitoring in the first few years of therapy.

In the absence of large, clinically relevant RCTs comparing different monitoring strategies (including no monitoring), we believe that our analysis of the FIT data is the best evidence available to clinicians deciding whether to monitor their patients' BMD during the first 3 years after starting therapy. There is certainly no scientific evidence to support the conclusion by Watts and colleagues that “clinicians should consider a follow-up BMD test 1 year after starting pharmacologic therapy for osteoporosis and thereafter at intervals determined by individual patient circumstances.” Our study suggests that routine bone density monitoring 1 to 3 years after starting treatment is unnecessary, potentially harmful, and a waste of scarce health care resources.