Our predominantly male and Caucasian study population of US veterans has a higher head injury and smoking prevalence than comparable civilian cohorts and a larger sample size than most previous studies of ALS and environmental risk factors. In this population, we did not detect any evidence for an association between ALS and various measures of cigarette smoking. Since our dataset had 80% power (at α=0.05) to detect an OR of 1.6 for an exposure prevalence of 64.2%, it is unlikely that the absence of a smoking association is due to lack of statistical power. A previous prospective study (5
) identified an association between ALS and cigarette smoking only in women, which is consistent with the lack of association in men observed here. The proportion of current
smokers in our dataset is similar to the US adult population, with reported rates of 22.6% for ages 45–64 years, 12.4% for ages 65–74 years, and 5.7% for ages 75+ years, per 2008 National Health Interview Survey (Table 25, http://www.cdc.gov/nchs/data/series/sr_10/sr10_242.pdf
). Thus, a bias of our sample towards a higher proportion of health-conscious individuals is unlikely.
In contrast to smoking, we identified a significant association between ALS and a shorter interval between the last head injury and the reference date, relative to no reported head injuries. This indicates that head injuries experienced during childhood or young adulthood may not confer an increased ALS risk later in life. While head injuries experienced in later adulthood may have a greater impact on risk, they may more likely be experienced by individuals with a lifetime history of vigorous physical activity. Consistent with this possibility, a previous study reported that ALS patients were approximately twice as likely as controls to have always been slim or to have been varsity athletes (19
). In addition to the observed main effect, our results support the possibility of gene-environment interaction, since the association between ALS and head injuries was stronger in APOE-4 carriers than non-carriers. However, given the limited number of individuals in the genotype-specific categories and the relatively wide confidence intervals, this result requires replication in a larger dataset.
Most previous studies of ALS and head injury did not include a more detailed analysis of the injury-to-onset interval or the age at which head injuries occurred. None of the studies conducted to date examined APOE genotypes jointly with head injuries. Our results are consistent with a recently reported trend for an increasing ALS association with a decreasing number of years between the last head injury and ALS diagnosis, and with an increasing age at the last injury (21
). In that study, the strongest association was observed for the interval “<=10 years between last injury and diagnosis” (OR 3.2, 95% CI 1.0–10.2), with a slightly elevated risk for the interval 11–30 years. The same study also estimated a pooled OR of 1.7 (95% CI 1.3–2.2) for at least one previous head injury, based on a meta-analysis of eight ALS studies. An Italian case-control study, which was not included in this meta-analysis, also reported an increased risk of ALS when the last head injury occurred at an older age and closer to the time of diagnosis (22
). While the risk of head injuries is generally twice as high in men as in women (49
), several specific attributes of the GENEVA study population, beyond its high proportion of men, may contribute to the higher prevalence of head injuries. These include combat-related injuries during deployment to major conflicts, and participation (before, during or after military service) in individual or competitive team sports that carry an increased risk of head injuries. The latter category includes professional or intense recreational soccer playing, which has been associated with ALS in some reports (17
). In our dataset, a descriptive comparison of individuals in the “at risk” group (<=15 years between injury and reference date) and those not included in this group was consistent with both possibilities: a higher proportion (61% vs. 38%) were deployed to at least one major conflict (World War II, Korean War, Vietnam War, Persian Gulf War); a higher proportion (55% vs. 34%) had received imminent danger pay for active duty status in an area that presented an imminent danger of being exposed to hostile fire or explosion of hostile mines; a higher proportion (13% vs. 7%) reported combat-related body injuries during deployment to a major conflict; and a higher proportion (71% vs. 58%) reported lifetime participation in the following list of sports: football, soccer, baseball, softball, basketball, hockey, lacrosse, rugby, water polo, boxing, wrestling, and martial arts.
Head injury has also been evaluated as a risk factor for other neurodegenerative diseases. A meta-analysis of 15 studies yielded an estimated OR for AD of 1.6 (95% CI 1.2–2.1) due to ever having sustained a head injury with loss of consciousness (51
), and supported an earlier hypothesis that this effect may be restricted to males (52
). Synergistic effects of head injury and APOE genotypes in AD were reported by some (30
), but not all studies (53
). In a population-based study of PD, an OR of 4.3 (95% CI 1.2–15.2) was reported for head injuries leading to loss of consciousness or hospitalization, as documented in medical records (55
). A study of 93 twin pairs who were discordant for PD replicated this association with a similarly large effect size (OR 3.8, 95% CI 1.3–11.0) (56
). All twin pairs were male veterans and had served primarily in World War II and the Korean War; the head injury prevalence in this dataset was 22.5%. To the best of our knowledge, the joint effect of APOE genotypes and head injury on PD risk has not yet been examined.
As previously discussed (56
), there are multiple biological mechanisms by which head injuries may trigger the molecular pathways leading to neuronal degeneration. They range from inflammatory and glutamate excitotoxicity pathways (57
), which increase the metabolic demands of neurons and microglia, to oxidative stress pathways that may impact mitochondrial function (59
). While the concept of biological interaction differs from that of statistical interaction (60
), modifying effects of the different apoe
protein isoforms on these pathways are biologically plausible and have been evaluated in experimental model systems (61
). Our observed statistical association supports a continued evaluation of apoe
as a potential “injury-response” protein in experimental studies of motor neuron degeneration (62
). Consistent with previous epidemiologic studies of head injury in PD (55
) and ALS (21
), our data also support the existence of a relatively long period of latency between injury and ALS onset, since head injuries up to 15 years before the reference date contributed to the observed increased odds of ALS. This opens up the possibility of implementing therapeutic measures that may prevent the neurodegenerative cascade triggered by such injuries from fully unfolding. For example, an investigation of non-steroidal anti-inflammatory medications, for which a protective effect has been reported in PD (63
), may be an important area of future research for ALS that has only recently begun to be explored (40
Limitations of our study include the fact that it is not population-based, that enrollment methods of cases and controls were not identical, and that the case enrollment methods may have increased the proportion of more slowly progressing individuals in the analysis sample. However, while the median survival time (22 months from diagnosis based on follow-up efforts to date, ) was higher in the incident cases who participated in both the sample collection and study interview than in those who did not, it is well within the range reported by several population-based studies, which extends from 16 months (65
) to 28 months (67
), with most studies reporting values between 19 and 23 months (68
). Although the control participation rate of 41% is less than ideal, Supplemental Table 1
shows that the primary difference between the controls included in this analysis, those who refused to participate, and those who were non-responsive to study invitations is the age distribution, with a mean age of 61.7 in the participants (22.4% in the 18–54 years range), 61.1 in refusers (27.2%), and 53.7 (48.9%) in non-responders. The apparent greater difficulty of contacting and successfully recruiting younger veterans likely explains other observed differences between these three groups in terms of VA health care, military branch and service period (Supplemental Table 1
). However, it is reassuring that the age difference between participants and refusers is negligible, and that the enrolled controls are very well matched by age to the subset of incident cases included in the present analysis (). An additional study limitation is that we did not query participants about the context of the reported head injuries and do not have access to more detailed medical or military records about the nature and severity of these injuries. Therefore, we cannot comment on whether or not the head injuries were combat-related or due to other characteristics of the study sample (e.g., participation in individual or competitive team sports that carry an increased risk of head injuries, as mentioned above). However, the fact that the nature of the reported injuries was very similar for cases and controls suggests that the result cannot be solely explained by recall bias, which would be expected to lead to under-reporting of milder injuries by controls. Strengths of our study include its relatively large sample size, the availability of some information about the nature and severity of the reported head injuries, and the age at which they occurred, and the ability to evaluate genetic and environmental risk factors simultaneously. While pooled genome-wide association studies (GWAS) of thousands of individuals are just now beginning to yield credible results that can be replicated in independent and similarly large datasets (71
), the success of GWAS in ALS has been much more limited than for other complex human diseases. This may partially be due to an important role of environmental risk factors, which have been completely ignored in all of the ALS GWAS studies published to date.
In summary, our results add to the existing body of evidence suggesting that head injuries may play an important role in multiple neurodegenerative diseases, including ALS. A novel finding of our study is the possibility that APOE genotypes may modify the association between ALS and head injuries experienced during adulthood, as previously proposed for AD. It would be worthwhile to test this hypothesis with future epidemiologic studies of ALS.