Post-hoc analysis of HORIZON on novel question with potential clinical implications. Well written manuscript by highly experienced investigators. These data might improve patient adherence if clinicians can provide reassurance to patients that their ZOL infusion reactions are “worth it” by indicating greater treatment efficacy. Since the biological mechanism of this effect isn't obvious, my main question is whether there might be fundamental differences in the APR+ group that might be influencing the improved fracture outcomes.
1. Table 1 reflects the randomized groups from the original HORIZON-PFT study, but doesn’t really reflect the analyses (which involves stratification by APR status where randomization is broken). It would be helpful to instead see subcategorizations of the ZOL/PBO baseline characteristics by those who developed APR vs those who did not, since these are the categorizations that forms the basis of much of the subsequent analyses (e.g. ZOL APR+ vs PBO APR+; ZOL APR- vs PBO APR-; ZOL APR+ vs ZOL APR-, etc)
2. As the authors note in the Discussion, it is challenging to explain why BMD improvements do not show a difference by APR status, but there are nevertheless fewer fractures among those with APR+. Could this be due to residual confounding that are incompletely accounted for in statistical adjustments? Revised baseline table would be helpful to gauge how APR+ group differs from comparator groups, e.g. younger age etc. 3. Later in the Results, it’s also mentioned that there were other significant baseline differences between APR+/APR- groups that favor a lower fracture rate among APR+ (e.g. higher femoral BMD, lower prior vert fx). These additional factors are not in the MV model presented in Table 4, but authors state additional adjustment for them didn’t change the results -- however these data aren’t shown. I would respectfully suggest that these data from the fully adjusted model be shown in Table 4.
4. What were rates of prior oral bisphosphonate use in APR+/- groups? How might differential rates of pre-treatment affect results?
5. What was dropout rate among those with and without APR reaction? What influence might variable dropout rate have on the observed results?
6. What are effects on clinical vertebral fractures?
7. Please clarify: Figure 1 lists vertebral fracture rate of 4.3% for APR- group; Results text lists 4.4%
8. Do Figure 2 and Supplemental Table 1 reflect models with adjustment for covariates? Can this be clarified in the legends and the Results text.
9. Reference 12 needs a publication year