Head injury and the risk of AD in the MIRAGE study

Methods.

Results.

Table 1 shows the characteristics of 2,233 probands and their 14,668 first-degree family members (4,465 parents, 7,694 siblings, and 2,509 spouses). A higher frequency of head injury was seen among both probands and affected relatives compared with unaffected relatives; the frequency of head injury was even higher among probands than affected relatives (p < 0.001). Men had a history of head injury more often than did women in any group. Head injury with and without loss of consciousness seemed to be equally distributed both in probands and affected relatives. However, head injury without loss of consciousness occurred more frequently than head injury with loss of consciousness in the unaffected relatives (p = 0.04). Probands were younger than their affected relatives at the time of onset of AD symptoms (p < 0.001). The mean age at onset was greater for women than men among both probands and affected relatives (p < 0.001). Age at onset was not different among those with and without head injury.

Probands had a history of head injury significantly more often than their unaffected spouses (table 2). Logistic regression analysis revealed that both head injury with and without loss of consciousness increased the odds of AD. However, the relation between head injury with loss of consciousness and AD (OR = 9.9) was stronger (p < 0.001) than that between head injury with no loss of consciousness and AD (OR = 3.1). The odds of AD associated with head injury were the same for patients with early- and late-onset AD.

Table 2 also shows the distribution of head injury among probands and their unaffected spouses according to the source of information. Spouses reported a slightly higher rate of head injury among probands and themselves than was reported by other relatives. However, spouses reported a lower rate of head injury with loss of consciousness. Increased odds of AD associated with head injury were consistent regardless of the source of information, although the relationship between head injury and AD was stronger in the group with information from other relatives than in the group with information from spouses.

Survival analysis revealed that the cumulative risk of AD by age 99 years in first-degree relatives (parents and siblings) was 41.4%. Risk of developing AD was significantly higher among relatives with head injury than among relatives without head injury at all ages after 60 years (figure, A). At age 93 years (the maximum age common to groups of relatives with and without head injury), the lifetime risk was 77.2% for those with and 40.1% for those without head injury. A similar increased risk of AD associated with head injury was also observed in spouses, although the overall risk of AD was lower (figure, B).

• Download figure
• Open in new tab
• Download powerpoint

Figure. (A) The cumulative risk of AD in relation to head injury among parents and siblings of patients with AD. Vertical lines represent the 95% confidence limits for the point estimates. (B) The cumulative risk of AD in relation to head injury among spouses of patients with AD. Vertical lines represent the 95% confidence limits for the point estimates.

Table 3 shows that, in the entire sample of relatives, head injury with loss of consciousness increased the risk of dementia four times and head injury with no loss of consciousness increased the risk about two times. The effect of head injury with loss of consciousness was greater but not significantly so in male relatives (RR = 5.6) than in female relatives (RR = 3.2), whereas there was no gender difference in AD risk associated with head injury without loss of consciousness. The risks of AD associated with head injury with loss of consciousness were the same in biologic relatives and spouses. Head injury with no loss of consciousness did not significantly increase the risk of AD among spouses (RR = 1.3). Considering head injury as a whole, the risk associated with head injury among parents and siblings was similar to that among spouses, suggesting a lack of interaction between head injury and a family history.

We evaluated the joint effects of head injury and APOE genotype on risk of AD in a subsample of 942 probands and a comparison group of 162 unaffected spouses and 165 unaffected siblings. The mean onset age of probands with and without APOE genotyping was 69.8 (SD = 9.3) and 69.9 (SD = 8.9) years. The groups were also similar in terms of gender (60.0% men versus 63.0% women) and frequency of head injury (19.4% versus 16.6%). Unaffected spouses and siblings were older (69.6 ± 9.4 versus 65.8 ± 16.5%; p < 0.001), more likely to be women (56.0% versus 46.1%; p < 0.001), and more likely to have a history of head injury (10.1% versus 2.8%; p < 0.001) than unaffected spouses and siblings without APOE genotype. Among AD cases with head injury, the interval from head injury to onset of AD symptoms did not differ by APOE-ε4 status.

Table 4 shows that subjects either heterozygous for APOE-ε4 or with a head injury (but not both) had an approximately threefold risk of AD compared with subjects lacking ε4 and head injury. Subjects homozygous for APOE-ε4 had the highest risk: the odds of AD were 10.3 and 7.9 for subjects with and without head injury compared with those lacking both factors. Comparison of these odds ratios within subjects with and without head injury clearly showed the dose-dependent effect of the ε4 allele on the risk of AD among those without head injury (table 5). Although the effect appears stronger for those without than for those with head injury, these differences were not statistically significant on either an additive or multiplicative scale. Viewed in another way, head injury increased the odds of AD to a greater extent among those lacking ε4 (OR = 3.3 [3.3/1.0]) than among ε4 homozygotes (OR = 1.3 [10.3/7.9]) or ε4 heterozygotes (OR = 1.8 [5.7/3.1]). Analyses that classified head injury with no loss of consciousness as negative for head injury showed a pattern of results similar to table 4 (data not shown).

We were concerned about pooling unaffected siblings and spouses because of APOE-ε4 frequency differences and genetic influences other than ε4 that affect AD risk. However, evaluation of APOE genotype in this sample revealed that the frequency of ε4 alleles in siblings (21.8%) was only marginally higher than in spouses (15.7%; p = 0.05), and the proportions of ε4 carriers in these groups were not substantially different (p = 0.08). Although these results suggest that pooling siblings and spouses is appropriate, analyses excluding spouses from the control group yielded similar results. Specifically, compared with subjects having neither ε4 nor head injury, ε4/x subjects with head injury had odds of AD of 4.7 (1.7 to 13.5), subjects with head injury but not ε4 had odds of 2.7 (1.2 to 5.8), and ε4/x subjects with no head injury had odds of 2.9 (1.9 to 4.6).

Discussion.

We examined the effect of head injury on the risk of AD in 2,233 probands and their 14,668 first-degree family members. Our results showed that head injury with loss of consciousness and, to a lesser extent, head injury without loss of consciousness increased the risk of AD. We also found that the risk of developing AD was significantly higher among relatives with head injury than among relatives without head injury at all ages after 60 years. Head injury with loss of consciousness also increased the risk of AD among spouses of probands. We did not find evidence for an interaction between family history and head injury, especially head injury with loss of consciousness. Our results indicate that head injury exerts a relatively greater effect on the risk of AD among persons lacking the APOE-ε4 allele compared with those having one or two ε4 alleles.

Compared with previous studies of head injury and AD, our investigation had several strengths. First, this sample of AD probands, first-degree relatives (parents and siblings), and spouses is to our knowledge by far the largest group studied in this manner. Although APOE genotype was available only in a subgroup, the sample size was sufficiently large to analyze APOE-ε4 homozygotes and heterozygotes separately. Second, we allowed for the possibility that the relationship between head injury and AD might vary according to severity of head injury. Finally, this was the first family-based study of the joint effects of head injury and APOE genotype on AD risk.

However, these results should be interpreted cautiously in light of several caveats associated with the study design. First, the association between head injury and AD could be simply an effect of recall bias with over-reporting of head injury in cases and under-reporting in unaffected relatives and spouses. Reporting bias is unlikely to have been a factor in the analyses involving APOE as these control subjects reported on themselves. The data used in the analyses that focused on risk in relatives were verified by multiple informants and medical records when available. These reports have good reliability; for example, one validation study from another dataset reported 91% agreement on head injury between subjects and their next-of-kin.38 Moreover, even if there was a reporting bias between AD probands and relatives, our analyses examining the interaction of head injury and APOE genotype (see table 4) were internally consistent (i.e., probands were compared only with relatives who were interviewed directly).

Second, the frequency of head injury among control subjects (unaffected spouses and siblings) with APOE genotyping (10.1%) was significantly higher than among subjects who were not tested (2.8%). The reason for the discrepancy is unclear but is likely related to the fact that head injury data for those with APOE genotype were self-reported, whereas information about head injury for most other siblings was reported by other family members. It is noteworthy that the frequency of head injury in our genotyped control subjects was very similar to that seen in the studies of Katzman et al.39 and Mayeux et al.25 of head injury and APOE genotype, although the definitions for head injury are not the same across studies. The frequency of head injury history in control subjects was about 5% in a recent study.5 A reporting bias, if it exists, is unlikely to influence the detection of an interaction between head injury and APOE genotype.

Third, our strategy of pooling unaffected siblings and spouses in analyses including APOE is unusual because of expected large differences in ε4 frequency between these groups.40 Whereas in this sample the difference was only 6%, siblings are also more likely than spouses to have other genes in common. Nevertheless, we repeated these analyses by assigning spouses and siblings to separate control groups and found very similar patterns in terms of the interaction between ε4 and head injury. Finally, the diagnosis of AD among relatives was made based on information provided by the informant and review of medical records. Nonetheless, misclassification of AD among relatives could have occurred. As we noted previously,32 this could bias the risk estimate of AD among relatives. However, this limitation would be unlikely to bias the relation between head injury and AD among relatives.

The association between head injury and AD was apparent in both men and women, although the relative risk for head injury with loss of consciousness was higher, but not significantly so, in men (RR = 5.6; 95% CL, 3.7 to 8.9) than women (RR = 3.2; 95% CL, 2.2 to 4.8). This trend is consistent with a metaanalysis of seven case-control studies showing a significant relation between head injury and AD among men but not women.41 A recent case-control study found a similar gender effect.5 However, one study reported an opposite effect, with a significant association only among women.6

In agreement with most previous studies,41 we did not observe a difference in the age at onset between patients with and without head injury. However, one study reported that onset of AD occurred 8 years earlier in those with head injury compared to those without head injury.42 A reanalysis of the Rochester Epidemiology Project data found that the observed time from injury to onset among 31 AD patients with head injury during 1935 to 1984 was shorter than the expected time estimated by a life-time method based on 689 AD patents without head injury in the same cohort. Several studies have shown that head injury occurring within 10 years of onset of AD is related to a greater risk than head injury occurring beyond 10 years of onset of AD.2,6,41 We were unable to evaluate this relationship because reliable information on age at head injury was unavailable for most relatives. However, among AD cases with head injury, the interval from head injury to onset of AD symptoms did not differ by APOE-ε4 status.

Our results suggest that severity of head injury is related to the magnitude of AD risk. Most previous studies have focused on head injury with loss of consciousness, primarily because of the possibility that head injuries without loss of consciousness would be susceptible to a greater recall bias. If that were so, one might observe a greater risk for AD among head-injured persons without than those with loss of consciousness. In contrast, we found that the risk associated with head injury with loss of consciousness was approximately double that associated with head injury without loss of consciousness. Nonetheless, even head injury without loss of consciousness significantly increased the risk of AD in parents and siblings of AD patients. Lack of this effect in spouses suggests the possibility of an interaction between head injury without loss of consciousness and family history. Alternatively, family informants may be more likely to recall a minor head injury in persons they perceive as having a greater genetic risk of AD. Notably, an interaction between head injury and a family history was not detected in two community-based studies.3,6 A metaanalysis revealed that head injury was more likely to be related to sporadic AD than to familial AD (defined simply as a positive family history).41

The data in tables 4 and 5⇑ support the idea that the effects of head injury and ε4 are additive (on the OR scale) but not “synergistic.” Although our results suggest different risks of AD due to APOE-ε4 genotype among subjects with as compared with those without head injury, these differences are not significant, perhaps because there was an insufficient number of subjects with head injury. Our finding of a smaller effect of head injury on AD risk among ε4 carriers compared with noncarriers is qualitatively different from the interaction noted in previous studies.5,25,39 Mayeux et al.25 observed a 10-fold increase in the risk of AD among subjects with both ε4 and head injury, compared with a twofold increase in risk among subjects with ε4 alone. Head injury in the absence of ε4 did not increase AD risk in this community sample from northern Manhattan, composed of a substantial number of African Americans and Hispanic subjects as well as whites. In our study, there were too few African American and Hispanic patients and control subjects to examine the relationship between ε4 and head injury in these ethnic groups. One intriguing possibility is that the populations represented in the study by Mayeux et al.25 are enriched for other genetic or environmental factors that promote a synergistic relationship between head injury and ε4. Because our findings did not confirm other studies showing synergistic effects of head injury and ε4 on AD risk, proposed strategies for incorporating APOE genotype information in counseling cognitively healthy individuals about the deleterious effects of head injury need to be carefully reevaluated.

Our results suggesting an interaction between head injury and ε4 are consistent with the hypothesis that head injury enhances production of amyloid plaque by increasing the expression of APOE.42 There is evidence that ε4 is associated with deposition of amyloid β-protein following head injury.43 A recent study of 30 boxers suggested that ε4 may be related to increased severity of chronic neurologic deficits in boxers sustaining traumatic brain injury.44 Lack of evidence for synergy in even large case-control studies, including ours, might be related to a survival bias if those with both head injury and the APOE-ε4 allele would survive differently from others. A recent study reported that APOE-ε4 predicted a poor clinical outcome after brain injury,45 although there are no data showing that AD patients with head injury and ε4 die early. This question can be addressed effectively by prospective studies currently in progress.