Neurology -- Correspondence for Haley, 61 (6) 750-756

Correspondence published:

Excess incidence of ALS in young Gulf War veterans: Carmel Armon, Springfield, MA 01199 (19 June 2004)

Reply to Dr. Armon: Robert W. Haley (19 June 2004)

Excess incidence of ALS in young Gulf War veterans 19 June 2004

Carmel Armon,
Division of Neurology, Baystate Medical Center
759 Chestnut Street, S4648,
Springfield, MA 01199
Send Correspondence to journal:
Re: Excess incidence of ALS in young Gulf War veterans

carmel.armon{at}bhs.org Carmel Armon, et al.

I agree with Dr. Haley [1], that the fundamental question is whether the incidence in the young Gulf War veteran population has exceeded the expected incidence for that age group, based on age-specific incidence in the general population. Comparison [2] to age-matched data from a contemporaneous US population with excellent case finding [3] shows that was not the case.
How did Dr. Haley find otherwise? New, elaborate, and unvalidated methodology was used for the first time in this article to calculate the number of cases to be expected in the deployed 1990 Gulf War cohort [1]. In my opinion, an unvalidated method cannot be relied upon to draw conclusions, just as one would not allow an unvalidated cholesterol assay to be used to determine cholesterol levels in patients, and then base treatment decisions on the results.
In my opinion, this method introduced two types of systematic errors: (a) errors that resulted in a “low-ball” estimate of the expected number and (b) errors that resulted in over-confidence in the accuracy of the results.
The method to derive incidence from mortality data resulted in underestimating disease occurrence in this population [4]. Inspection of Figure 1A at the end of the statistical appendix shows that age-specific US mortality rates for the age groups 50 years and under are approximately 50% of those of exponential-linear regression-fitted age-specific incidence rate estimates derived from studies published worldwide as far back as 1983. Furthermore, the older incidence studies likely underestimate the rates that might be obtained using contemporary methods including more complete case finding.
Applying average disease duration from onset to diagnosis and from diagnosis to death to infer incidence from mortality data in the age group of young, deployed, 1990 Gulf War veterans with longer than average disease duration compounded the underestimation of the expected incidence.
A wide scatter of incidence rates was noted in the below 40 age group in the studies considered by the author, resulting in inability to fit regression curves to incidence data. This was one of the author’s reasons for using mortality data rather than incidence data. Consequently, in my opinion, the method had the appearance of greater precision than the reality it was modeling.
I would like to pay tribute to the bravery and courage of all veterans and all patients with ALS.
References
1. Haley RW. Excess incidence of ALS in young Gulf War veterans. Neurology. 2003;61:750-6.
2. Armon C. Occurrence of amyotrophic lateral sclerosis among Gulf War veterans. (Letter to the editor) Neurology 2004;62:1027-1029.
3. McGuire V, Longstreth WTJ, Koepsell TD, van Belle G. Incidence of amyotrophic lateral sclerosis in three counties in western Washington state. Neurology 1996; 47: 571–573.
4. Chio A, Magnani C, Oddenino E, Tolardo G, Schiffer D. Accuracy of death certificate diagnosis of amyotrophic lateral sclerosis. J Epidemiol Community Health. 1992;46:517-8.
5. Horner RD, Kamins KG, Feussner JR, et al. Occurrence of amyotrophic lateral sclerosis among Gulf War veterans. Neurology 2003; 61: 742–749.

Reply to Dr. Armon 19 June 2004

Robert W. Haley,
University of Texas Southwestern Medical Center
5323 Harry Hines Blvd., Dallas, TX 75390-8874
Send Correspondence to journal:
Re: Reply to Dr. Armon

Robert.Haley{at}UTSouthwestern.edu Robert W. Haley

Dr. Armon, a leader in ALS research, poses questions probing the validity of my study that first supported the significant cluster of ALS in young Gulf War veterans. [1] Given the rarity of real ALS clusters and thus their research importance, such dialogue is useful.
The method I used to test for an increase in ALS above the expected rate was the standardized morbidity ratio (SMR). [6] Far from an “elaborate,” “unvalidated” methodology, the SMR is a time-honored epidemiologic method, used in over 2,000 papers listed in MEDLINE since the early 1970s. What may have appeared as novel in my on-line appendix [7] was but the detailed mathematical documentation of how I performed the usual steps of an SMR calculation with the Gulf War data.
In an earlier draft of my paper, I included SMRs comparing observed to expected ALS incidence calculated from both published studies of age- specific incidence and nationwide mortality data. Contrary to Dr. Armon’s prediction, SMRs from the two approaches agreed closely. The journal’s peer reviewers argued, however, that the published incidence studies contained too much sampling variation in younger age groups to justify regression-derived estimates of mean age-specific incidence rates (see Figure A1-D [7]), and so I deleted the SMRs calculated from the published incidence studies. Dr. Armon’s suggestion to compare only to the one published incidence study with the highest ALS rates at young ages, [3] while ignoring others with lower rates, [7] fails to consider the extreme sampling variation in those studies. One cannot “cherry-pick” just the study with the highest rates.
Dr. Armon’s attribution of low incidence rates to less complete case ascertainment in older studies is not supported by the data. In the 13 incidence studies I cited, [7] there is no correlation between year of publication and ALS rates in men <50 years old (p = 0.34). For example, compare recent U.S. studies [3] and [8].
To make population mortality rates better approximate incidence rates, I adjusted age-specific mortality rates by average duration from diagnosis to death, a firmly established interval. [9,10] This reduced the SMRs. How long Gulf veterans with ALS will survive on ventilatory support is irrelevant to this adjustment.
Rate estimates in young age groups from published incidence studies are imprecise due to small numbers, but nationwide mortality data, which contain large numbers at young ages, provide highly precise estimates. [7]
Most importantly, the significant ALS cluster identified by my study was confirmed by Horner et al. using fundamentally different methodology. [5]
References
6. Breslow NE, Day NE. Statistical methods in cancer research. Vol 2. the design and analysis of cohort studies. Lyon, France: International Agency for Research on Cancer. 1987: 65-72, 91-103.
7. Haley RW. Statistical appendix: estimation of expected incidence of ALS. Neurology. 2003;61:750-6. Available on-line at: http://www.neurology.org/content/vol61/issue6/images/data/750/DC1/Haley_supplemental_data.doc.
8. Annegers JF, Appel S, Lee JR, Perkins P. Incidence and prevalence of amyotrophic lateral sclerosis in Harris County, Texas, 1985-1988. Arch Neurol 1991; 48:589-593.
9. Magnus T, Beck M, Giess R, Puls I, Naumann M, Toyka KV. Disease progression in amyotrophic lateral sclerosis: predictors of survival. Muscle Nerve 2002; 25:709-714.
10. Chio A, Mora G, Leone M et al. Early symptom progression rate is related to ALS outcome: a prospective population-based study. Neurology 2002; 59:99-103.