Smoking may be considered an established risk factor for sporadic ALS

Armon's dismissal of the two largest cohort studies in his review on smoking and ALS is surprising. [1] Both studies provide important information and the Cancer Prevention Study-II (CPSII) [4] is similar in design to the EPIC study, [3] which Armon classified as rigorous.

The advantages of EPIC over CPSII according to Armon are: use of validated tools to collect precise exposure data; the fact that smoking was the first ALS risk factor reported; no gaps in case finding; and population representativeness. We disagree.

In CPSII, detailed smoking behavior information was collected at recruitment similarly to EPIC. The number of CPSII men/women missing smoking information was 3.0%/1.9%-- not 31.8%/11.8% as Armon states--because cigar and pipe smokers were considered separately. Armon [1] implies that the strength of evidence from each study depends on the order in which the analyses were conducted or the investigators' intention, whereas both are irrelevant.

Considering gaps in case finding, this is the only CPSII limitation not shared by EPIC yet this gap is unlikely to bias results. [4] We have now identified male ALS deaths in CPSII during the 1982-88 gap (n=75). When these were included in the analyses, the relative risk (95% CI) of ALS comparing current to never smokers was virtually unchanged from 0.69 (0.49-0.99) to 0.73 (0.53- 1.01).

Finally, EPIC investigators reported that participants were primarily recruited from the general population. [3] As in CPSII, participants were volunteers and may not be representative of the underlying population. Nevertheless, this is unlikely to bias either investigation. If smoking increased ALS risk, then both cohorts of smokers should develop ALS at a higher rate than non-smokers.

Armon appears to apply double standards in assigning class of evidence scores. This may be due to bias towards the null from exposure misclassification, but the degree of smoking misclassification is modest as demonstrated by strong associations with several smoking-related diseases in CPSII. [5] Thus, this is unlikely to obfuscate an adverse smoking-ALS effect, particularly given the much larger CPSII size.

We suggest using established tools for evaluating evidence when carrying out this type of reviews, [7,8] to obtain comparable results using standard methods. We believe that the smoking and ALS risk relationship deserves further investigation.[6] Behind the heterogeneity of results across studies and the lack of dose-response in most investigations, there may be valuable information for the prevention and treatment of ALS.

References

5. Ezzati M, Henley SJ, Thun MJ, Lopez AD. Role of smoking in global and regional cardiovascular mortality. Circulation 2005;112:489-497.

6. Weisskopf MG, Ascherio A. Cigarettes and amyotrophic lateral sclerosis: only smoke or also fire? Ann Neurol 2009;65:361-362.

7. IARC. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Preamble. Lyon, France: WHO, 2006.

8. Greenland S. Invited Commentary: A Critical Look at Some Popular Meta-Analytic Methods. Am J Epidemiol 1994;140:290-296.

Disclosure: Dr. Weisskopf received an honorarium from the ALS Society of Canada for a speaking engagement and received funding from NIEHS [#K01 ES012653 (PI)] and the Department of Defense [# W81XWH-08-1-0499 (PI)]. Drs. Gallo and O'Reilly report no conflicts of interest. Prof. Vineis serves on the scientific advisory boards of Istituto Toscano dei Tumori, CREAL and IMIM scientific advisory boards; servves as a member of the Editorial Boards of Biomarkers, Mutation Research-Review in Mutation Research, International Journal of Cancer, Cancer Epidemiology, Biomarkers, and Prevention, Carcinogenesis, European Journal of Cancer, Journal of Cancer Epidemiology; serves as editorial consultant for the Journal of Clinical Epidemiology; and serves as Senior Editor of Mutagenesis and the European Journal of Clinical Investigation; and receives royalties for "I due dogmi" (Feltrinelli, 2009). Dr. Ascherio serves on the editorial boards of Neurology and Annals of Neurology; received funding for ALS research from the Department of Defense [# W81XWH-05-1-0117 (PI)] and NIH/NINDS [#R01 NS045893 (PI)]; receives research funding from the National Insitutes of Health, the Department of Defense, the Michael J. Fox Foundation, and the National Multiple Sclerosis Society, Autism Speaks; received honoraria from Merck Serono for two scientific presentations; served on the editorial board of the American Journal of Epidemiology; and served on a scientific advisory board for the Michael J. Fox Foundation.

Armon's review article on smoking and the risk of amyotrophic lateral sclerosis (ALS) [1] referenced our cohort study based on the Swedish Construction Workers Cohort. [2] The authors classified our study as Class IV evidence for a conclusion regarding the association between smoking and ALS. We do not agree with several of their criticisms.

Firstly, the mean age of the Construction Workers Cohort was 35.5 years at the beginning of follow-up and 55.5 years at the end. Armon contends that this violated the validity of the study results. We do not believe that the population characteristics could influence the validity of a study when only internal comparisons were made.

Considering the generalizability of the results, we were not trying to assume that these results apply to other general populations given the specific features of construction workers. Although mainly young men, this cohort had a wide age range (i.e., age at entry to the cohort was 15-81 years). Age at enrollment is not relevant. However, age at ALS diagnosis could be relevant if smoking had differing effects on risk of ALS dependent on age. Regardless, the majority of our ALS cases were older. The mean age at ALS diagnosis was 64.4 years (range: 31-85 years) with 15.6% at age <_55 years="years" _15.6="_15.6" _55-59="_55-59" _18.1="_18.1" _60-64="_60-64" _18.8="_18.8" _65-69="_65-69" _14.4="_14.4" _70-74="_70-74" and="and" _17.5="_17.5" _880575="_880575" years.="years." p="p"/>Secondly, the smoking information in our baseline questionnaire was obtained via the construction workers' personal interviews with a trained nurse. Presumably, this method is more reliable than self-report. Although the smoking status is not validated, as Armon mentions, we believe it is common in other large-scale follow-up studies. [3,4] Furthermore, this was the method employed by another study cited by Armon, which they classified as Class I evidence. [3]

References

1.Armon C. Smoking may be considered an established risk factor for sporadic ALS. Neurology 2009;73:1693-1698.

2.Fang F, Bellocco R, Hernan MA, Ye W. Smoking, snuff dipping and the risk of amyotrophic lateral sclerosis--a prospective cohort study. Neuroepidemiology 2006;27:217-221.

3. Gallo V, Bueno-De-Mesquita HB, Vermeulen R, et al. Smoking and risk for amyotrophic lateral sclerosis: analysis of the EPIC cohort. Ann Neurol 2009;65:378-385.

4. Weisskopf MG, McCullough ML, Calle EE, et al. Prospective study of cigarette smoking and amyotrophic lateral sclerosis. Am J Epidemiol. 2004;160:26-33.

Disclosure: The authors report no disclosures.

I appreciate the interest in my article [1] by Weisskopf et al. and by Fang and Ye. However, it is concerning that they are unwilling to consider that their studies may have methodological limitations. Their results conflict with three studies where the quality is not questioned.

Weisskopf et al. cite a paper [5] that relied on the CPSII smoking data to support the value of that database for evaluating the role of smoking in ALS. However, Ezzati et al. used lung cancer mortality "as an indirect marker for accumulated smoking hazard (signal)." Since the risk of lung cancer in persistent smokers (up to 16-32 times of that in non-smokers [9]) is much greater than that of ALS (approximately double that of non-smokers), the detection of the lung cancer risk ("large signal") even when there is some "background noise" (methodological limitations) cannot be used to imply that an increased risk of ALS ("small signal") might not be obscured by those limitations.

The same paper [5] analyzed 1982-1988 mortality data "…while minimizing misclassification of exposure resulting from cessation of smoking during follow-up." This statement supports my reservations about Weisskopf et al.'s original report that analyzed ALS mortality only in later years [4]. However, they contend that including 75 male ALS deaths identified during 1982-88 did not change their conclusions. The authors do not provide the point estimate for the mortality ratio for those years alone. I suspect that it is close to 1.0 with a 95% confidence interval with the upper limit perhaps reaching 2.0. Their findings, based on the first 7 years of observation, would not refute those of EPIC. [3] Furthermore, I believe that 75 cases is too low a number, and may reflect between 56%-67% under-ascertainment. More complete case findings for those years may change results even further. I calculated the percent of non-responding (or missing data) directly from Table 1 and the methods section of Weisskopf et al. [4]

Even if Weisskopf et al. propose that the representativeness of a sample, investigator intent and the number of associations researched are irrelevant – these factors are at the core of statistical inference and assignment of class of evidence. Readers may be more familiar with these concepts when applied to the analysis of clinical trials, where the primary outcome variable is the one that counts. The distinction between "confirmatory" and "exploratory" analyses is explicit in the classification system used. [10] These concepts are also reflected within the AAN Classification system for inferring causality [11]. I stand by the application of these criteria. [10]

I would like to refer Drs. Fang and Ye to the original paper [1], including table e-1 and a standard text regarding use of stratified analysis to look for confounders. [12] Their statement: "we were not trying to assume that these results apply to other general populations given the specific features of construction workers," further supports the classification of their report [2] as class IV evidence.

The original conclusions are strengthened by this opportunity for investigator response. I thank Weisskopf et al. and Fang and Ye for their remarks, and encourage the readers to rely on the results of the better-quality studies when drawing their own conclusions about smoking and ALS.

References

9. Huq S, Maghfoor I, Perry M. Lung Cancer, Non-Small Cell. http://emedicine.medscape.com/article/279960-overview. Accessed February 10, 2010.

10. Armon C. An evidence-based medicine approach to the evaluation of the role of exogenous risk factors in sporadic amyotrophic lateral sclerosis. Neuroepidemiology 2003;22:217–228.

11. Harden CL, Meador KJ, Pennell PB, et al. Practice Parameter update: Management issues for women with epilepsy—Focus on pregnancy (an evidence-based review): Teratogenesis and perinatal outcomes: Report of the Quality Standards Subcommittee and Therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology and American Epilepsy Society. Neurology 2009;73:133–141.

12. Greenland S, Rothman KJ. Introduction to Stratified Analysis. In: Rothman KJ, Greenland S, and Lash TL, editors. Modern Epidemiology, 3rd ed. Philadelphia: Lippincott, Williams and Wilkins;2008:258-282. Disclosures: See original article for full disclosure list.