Response to: Comment on: "Denosumab Rapidly Increases Cortical Bone in Key Locations of the Femur: A 3D Bone Mapping Study in Women With Osteoporosis".

Abstract

Dr. Hinshaw is concerned about our use of the expression “increasing cortical thickness” in the abstract of Poole and colleagues1: “Overall, treatment with denosumab increased femoral cortical mass surface density by 5.4% over 3 years. One-third of the increase came from increasing cortical density, and two-thirds from increasing cortical thickness, relative to placebo.” Although there is no universally accepted definition of cortical thickness,2 cortical bone mapping (CBM) “thickness” measurements are arguably more deserving of the name than those based on thresholding or full-width half-maximum estimation. Such techniques are known to be highly biased when the cortex is less than a few millimeters thick,3, 4 but nevertheless lie behind hitherto uncontroversial references to “cortical thickness” in many scientific abstracts. In the body of Poole and colleagues,1 we did explain in considerable detail how CBM thickness estimates may exclude pores, and consequently how increases in CBM thickness might reflect infilling of cortical pores and not necessarily modeling-based formation on outer and inner surfaces. The latter is not a “hitherto unobserved mechanism of action,” having been observed in preclinical denosumab trials and presented at scientific meetings before the publication of Poole and colleagues1 (and now in Ominsky and colleagues5). However, this is a novel concept and further research is required to confirm or refute the findings in humans. In any case, it was not our intention to favor one mechanism over the other in the abstract, and it is assumed that either may have advantages for bone strength and the prevention of hip fractures, the goal of treatment. Dr. Hinshaw is also concerned that CBM “was not applied to measuring changes over time or after therapy.” Here, he is presumably referring to our validation studies,6-8 because CBM has in fact been used in several longitudinal studies9-12 in addition to the present one. Any technique capable of measuring a quantity t, with accuracy and precision a ± p, can be expected to measure small, discrete changes in t with approximate accuracy and precision 0 ± √2p. This assumes that the accuracy a does not change between scans, an assumption that we have verified13 subject to reasonable precautions (same scanner and settings). There is also the added reassurance that comes from measuring changes relative to placebo-treated controls. Statistical parametric mapping, which we employ in all our longitudinal studies, accounts for the measurement precision when assessing whether the changes are statistically significant or not. We do not agree that Zebaze and Seeman2 “quite effectively” dispute the details of our analysis, nor do we believe that it was their intention to do so. On the contrary, there is considerable overlap between the thesis of Zebaze and Seeman2 and the caveats spelled out in Poole and colleagues.1 Dr. Hinshaw correctly infers that we are enthusiastic about our imaging technique, but only inasmuch as the evidence supports its efficacy. CBM has now been scrutinized by three sets of expert reviewers for a leading medical image analysis journal.(6–8) All the software required to replicate our workflow, including cortical thickness measurement,14 surface alignment,15 and statistical parametric mapping,16 is available for free download on the Web. We urge interested researchers to try it for themselves.