Authors’ reply to Abrahamsen and colleaguesBMJ 2013; 346 doi: http://dx.doi.org/10.1136/bmj.f1518 (Published 12 March 2013) Cite this as: BMJ 2013;346:f1518
- Yana Vinogradova, research fellow in medical statistics1,
- Carol Coupland, associate professor and reader in medical statistics1,
- Julia Hippisley-Cox, professor of clinical epidemiology and general practice1
Pazianas and colleagues’ study did show a 31% reduced risk of incident colorectal cancer in alendronate users compared with non-users (adjusted hazard ratio 0.69, 95% CI 0.60 to 0.79),1 2 as we noted.3 The alendronate group in table 2, however, included all women with one or more prescription for alendronate, and mean follow-up was 3.4 years, so the analysis seemed to look at any duration of alendronate use, not five or more years, as the authors’ response implied. There was also no significant dose-response association in their analysis restricted to alendronate users (0.89, 0.66 to 1.22 when comparing users exposed to more or fewer than 180 defined daily doses).
We used a nested case-control design because this can better quantify time dependent exposures.4 To account for longer survival for bisphosphonate users, we matched cases and controls by age and calendar time, and required all controls to be alive and registered with the practice at the date of the first recorded diagnosis of cancer in their matched case. This ensured that survival up to the point of the case diagnosis of cancer would be equivalent between the two groups, so that the potential for bisphosphonate exposure would be comparable. All patients with initial cancer diagnoses were included in the analysis, whether or not diagnosis was after death. To ensure an unbiased estimate of relative risk, our study was based on an underlying cohort structure, where matched controls are randomly selected from all remaining subjects at risk, including potential future cancer cases (incidence density sampling).5
Cite this as: BMJ 2013;346:f1518
Competing interests: None declared.