Tobacco Documents Online

3 (iv) I do not really understand from their description of the methods whether the age differences between current and never smokers noted in Table 2 have properly been taken account of in the multivariate analysis. (v) There were a very substantial proportion ofthe origina12128 subj ects who refused further interview (349) or who had missing smoking data (129). It is unclear how this would have affected the representativeness of the subjects followed and the results achieved. (vi) The rate of incident AD is 2 years (142/1062 = 13%) seems high to me, in a population involving predominantly men and women aged 65-74 (see Table 2). Generally the study appears to have some weaknesses. However, it is another prospective study that does not find a negative relationship of smoking with AD and, as such, somewhat weakens the overall evidence, which is complex and requires detailed review. P N Lee 5.10.99

1 REVIEW 1069 CONFIDENTIAL Subject ref 18a "The influence of smoking on the risk of Alzheimer's disease" C Merchant et al Neurology (1999), 52, 1408-1412 In 1994, in Neuroepidemiology (1994, 13, 131-144), in a review of the evidence relating _ smoking to Alzheimer's Disease (AD), I noted that there was substantial evidence of a negative ! association. I also cited various observations that seem to make it plausible that nicotine might protect against AD. In 1998, Ott et al (Lancet, 351, 1840-1843), reported results of a prospective study which found that, after adjusting for age, sex, education and alcohol intake, smokers had a significantly increased risk of AD (Relative risk 2.3, 95% CI 1.3-4.1). Smoking was reported to be "a strong risk factor for AD in individuals without the APOE-e4 allele (4.6, 1.5-14.2) but had no effect on participants with this allele (0.6, 0.1-4.8)." Ott et a1 put over the impression that other studies found a negative relationship because they were case-control. In my review of that paper (Review 1013) I noted that there were a number of other prospective studies none of which reported the positive association that Ott et al did, and some of which reported the negative relationship that is more typical of the epidemiological literature. I also noted that the RRs for ~ individuals with and without the APOE-e4 allele did not actually differ significantly on statistical ~ comparison. The current paper describes the results of a prospective study of "elders" residing in Northem Manhattan in New York, which reported in their abstract that: Results: The relative risk (RR) of AD among former smokers was 0.7 (95 % CI, 0.5 to 1.1). The RR among current smokers was 1.9 (95% CI, 1.2 to 3.0). Smokers without an APOE-e4 allele had the highest risk of AD (RR = 2.1;95% CI, 2.1 to 3.7) compared with those with an APOE-e4 allele (RR = 1.4; 95% CI, 0.6 to 3.3).

2 Conclusions: Our results are consistent with the observation that smoking increases the risk of AD. However, we found that among previous smokers who quit smoking, there may be a slight reduction in the risk of AD. The study involved 2128 persons in the baseline phase ofthe study, who completed an in-person interview providing data on general health and function, smoking history, education, ethnic group and risk factors related to dementia and then underwent a standardized clinical assessment including a medical history, physical and neurologic examination and a neurophychological battery. After excluding those with dementia, Parkinson's disease or stroke, who refused fizrther following, who failed to provide smoking data, or who subsequently died, an assessment of incident AD was made 2 years later on the remaining 1062. The APOE genotype of each participant was also determined. The main results of the study are shown in Table 3 and, separated by APOE-e4 status, in Table 4. Points to note are as follows: (i) It is clear that the results in Table 4 show no significant difference in the smoking-AD relationship according to APOE-e4 status. The total number of incident AD cases, 142, is probably too small to detect anything but the most marked interaction. (ii) I do not understand why the authors concluded multivariate analysis and then reported the crude (unadjusted) results from Table 3 in the abstract. (iii) I do not understand how the authors calculated their crude relative risk estimates. Table 3 shows that for never smokers there were 79 AD cases out of 672 subjects at risk and that there were 27 AD cases out of 171 subjects at risk. A crude estimate of the relative risks would appear to be (27/171)/(79/672) = 1.34, not 1.9 as given. Even if one does a case-control type estimate one gets (27/144)/(79/593) = 1.41. Both are substantially less than the figure of 1.9 given, and could apparently be reduced further by adjustment for education and ethnicity.