Age and Ageing Advance Access originally published online on January 25, 2007
Age and Ageing 2007 36(2):235; doi:10.1093/ageing/afl170
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Reply
SIR—We are delighted to be able to clarify a couple of points which Dr Capellà and Dr Vidal have raised in their comments. First, we agree that multivariate Cox proportional-hazards methods are usually the standard procedures employed when the effect of several risk factors on survival is analysed. However, several studies have shown that, when model assumptions are held, similar results are obtained for the logistic and Cox models in shorter periods of follow-up of 5 years or less, but not in longer periods of 10 years or more [1–4]. Thus, although the Cox model was our first-line option, our study design and data required the use of logistic regression. Cox regression yields the relative hazard of death, which is assumed to be constant over time, unless time-dependent covariates are included in the model, but because of the high percentage of censored data (up to 85%), the use of Cox proportional-hazards methods could induce an estimation bias [5]. Thus, we decided to use the logistic regression and to estimate the log odds ratio of death at a specified time point, including as a covariate of the follow-up time for each patient.Second, the death rate was estimated taking into account the deaths occurring among patients who continued the AChEI as well as those who abandoned the AChEI. This group was included because patients were exposed to the drug and received treatment for a mean period of 13.6 months (95% CI = 0.2–32.2). Thus, we were not mimicking an ITT analysis. We included patients in our analysis who had been treated with an AChEI for a clinically relevant period of time. Moreover, in order to control the differential effect of these two groups of patients, the number of months being treated with AChEI was included as a covariate in the multivariate model. And last but not the least, we have reviewed the data and the statistical analysis; the total number of patients treated with donepezil was 202. The discrepancy between Table 1 and the text and Figure 1 is an unfortunate mistake.
1 Unitat de recerca, Parc Hospitalari Martí i Julià, C/Dr Castany s/n, 17190 Salt (Girona), Spain
2 Unitat de valoració de la Memòria i les Demències, Spain
* To whom correspondence should be addressed Email: uvamid{at}ias.scs.es
References
- Annesi I, Moreau T, Lellouch J. (1989) Efficiency of the logistic regression and Cox proportional hazards models in longitudinal studies. Stat Med 8 1515–21.[Web of Science][Medline]
- Ingram DD and Kleinman JC. (1989) Empirical comparisons of proportional hazards and logistic regression models. Stat Med 8 525–38.[Web of Science][Medline]
- Moriguchi S, Hayashi Y, Nose Y, Maehara Y, Korenaga D, Sugimachi K. (1993) A comparison of the logistic regression and the Cox proportional hazard models in retrospective studies on the prognosis of patients with gastric cancer. J Surg Oncol 52 9–13.[Web of Science][Medline]
- Callas PW, Pastides H, Hosmer DW. (1998) Empirical comparisons of proportional hazards, poisson, and logistic regression modeling of occupational cohort data. Am J Ind Med 33 33–47.[CrossRef][Web of Science][Medline]
- Kivela T and Grambsch PM. (2003) Evaluation of sampling strategies for modeling survival of uvela malignant melanoma. Invest Ophthalmol Vis Sci 44 3288–93.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||