Using prospective data we observed graded relationships between the ACE score and the risk of lung cancer. Moreover, relationships between a high ACE score and lung cancer were particularly strong for those who died from lung cancer at younger ages. The increase in risk of lung cancer was only partly due to relationships between ACEs and an intermediate causal factor, smoking. The occurrence of ACE-related lung cancer not attributable to conventional risk factors suggests other mechanisms by which childhood traumatic stressors negatively affect health.
The observed associations between ACEs and lung cancer may be conservative. Case fatality for lung cancer is high. The overall 5-year relative survival rate for 1996-2004 from 17 Surveillance Epidemiology and End Results (SEER) geographic areas in the United States was 15% (age <65 years, 18%; age >/=65 years, 13%) with a survival rate for small cell lung cancer of about 6% and for non-small cell of only 17% [37
]. Thus, given relationships between ACEs and smoking behaviours (particularly associations with early smoking initiation) which would increase the probability of developing smoking-related disease, it is possible that some Kaiser members with higher ACE scores were less likely to survive and to be included in the baseline data collection because they had already died from lung cancer or another smoking-related disease.
Some degree of selection bias is inevitable in observational research simply because not all persons who are born will survive to the observation period of interest and because the population that does survive often differs from the population that does not. In the case of ACEs, which are associated with numerous adverse health behaviours and health outcomes (perhaps most importantly premature death), it is reasonable to postulate that persons who are exposed to ACEs (particularly multiple ACEs) are more likely than those who are not exposed to die during childhood or young adulthood, be institutionalized, or otherwise lost prior to the initiation of the ACE Study and baseline survey resulting in a downward bias for the association between ACEs and lung cancer. Some caution must be exercised in making such an assertion with regard to the direction of the bias since this does not always hold for non-dichotomous exposures.
A strength of this study lies in the use of two prospective data sources to identify cases of lung cancer. The prospective data from hospital and mortality records are not subject to recall bias and are reported by physicians who were unaware of the patient ACE score. Also, the ACE Study incorporates multiple forms of childhood traumatic stressors. Studies that examine only one or at most two types of stressors may 1
) underestimate the burden of exposure, 2
) fail to recognize the interrelationships between different types of traumatic stressors during childhood, and/or 3
) incorrectly attribute long-term consequences to single types of childhood traumatic stress [38
] despite convincing evidence suggesting that exposure to multiple forms of abuse and traumatic stressors appear to influence health behaviors and outcomes through a cumulative process.
The results of this study are subject to several limitations. The frequency of ACEs may represent an underreporting of their actual occurrence given the sensitive nature of the questions. However, our estimates of the prevalence of childhood exposures are similar to estimates from nationally representative surveys [39
] indicating that the experiences of our participants are comparable to those of the larger population of adults. For example, in our study we found that 16% of the men and 25% of the women met the case definition for contact sexual abuse; a national telephone survey of adults in US conducted by Finkelhor and colleagues [41
] using similar criteria for sexual abuse estimated that 16% of men and 27% of women had been sexually abused. Of the men from our study, 30% had been physically abused as boys, which closely parallels the percentage (31%) found in a recent population-based study of Ontario men in Canada that used questions from the same scales [42
]. The similarity in estimates of the prevalence of these childhood exposures between the ACE Study and other population-based studies suggests that our findings are likely to be applicable in other settings.
The adverse effects of smoking are in part a function of the amount smoked, duration of smoking, inhalation, and tobacco product smoked. While we were able to incorporate the amount smoked, this analysis did not have data on duration and therefore was not able to compute the number of pack-years smoked. Thus, associations between ACEs and the occurrence of lung cancer that remained after the addition of smoking into the model may be the result of our inability to capture pathway effects of smoking duration. Also, smoking status was based on a single measure at baseline; therefore, we do not have data on initiation or cessation during follow-up. Similarly, exposure to second hand smoke may have changed over time. While we included variables in the final model for baseline prevalent asthma, COPD, and tuberculosis - conditions associated with the occurrence of lung cancer - we did not have information on occupational or other environmental exposures (e.g., asbestos, radon).
ACEs are associated with risk factors for chronic disease conditions such as ischemic heart disease [43
], liver disease [44
], COPD [45
] and mental disorders [46
] that may result in an increased risk of exacerbating underlying lung disease and/or negatively affect general health, leading to disease progression or perhaps increasing the likelihood of undiagnosed lung cancer being identified [45
]. Although mortality follow-up was available for a maximum of 10 years, statistical power was somewhat limited owing to relatively few deaths during follow-up among persons exposed to multiple ACEs. We plan to continue repeating the NDI search and related analyses in the coming years. As is the case in many observational studies, there may have been unknown or unmeasured confounding factors for which adjustment was not possible. Moreover, measurement error in the assessment or estimation of covariates and their severity may have resulted in incomplete adjustment and residual confounding. We feel these data are compatible with a moderate association between ACEs and risk of lung cancer; however, this assumes that there is no bias in the data collected and that our statistical models are correct [48
Finally, we examined competing risks as a potential explanation for observed results using mortality data. If competing causes of loss to follow-up act independent of the outcome (e.g., lung cancer), then consistent estimates of the survival function are possible. Alternatively, if the independence assumption does not hold, a bias can be introduced because the number of failures from the competing risk may influence the number of subjects at risk for the outcome of interest. After identifying deaths during follow-up from smoking-related diseases [49
] (other than lung cancer) (n = 707 deaths) and removing these observations from the censored group, we repeated the models shown in Table and observed similar results for risk of lung cancer death at any age as well as premature death from lung cancer.
In summary, exposure to adverse childhood experiences is common. Insofar as stressful and traumatic childhood experiences contribute to the adoption of adverse health behaviours, such as smoking, and subsequent development of poor health outcomes, such as death from lung cancer, these childhood exposures should be recognized as underlying causes of premature mortality [50
]. Reducing the burden of adverse childhood experiences should be considered in health and social programs as a means of primary prevention of lung cancer as well as other smoking-related diseases [43
]. In addition, because smoking did not completely explain observed relationships between ACEs and the occurrence of lung cancer, other pathophysiologic pathways by which childhood stressors may influence the risk of lung cancer should be explored.