There is a notable lack of clarity regarding the conceptual definition of `delayed onset'. Merely because a disorder is recognised many years after the aetiological event is not evidence that onset of the disorder was `delayed'. It has been noted that PTSD diagnosed more than 6 months after a traumatic event may indicate delayed treatment or seeking of disability benefits, delayed onset of any symptoms of PTSD (identified as `definition 1' by Andrews et al)<sup>1</sup>, or delayed onset of the full disorder such that a change in one or two symptoms alters PTSD diagnostic status (identified as `definition 2' by Andrews et al). Another issue is the actual time interval from traumatic exposure to onset, with `delayed onset' counting as any PTSD onset that occurs from 7 months to 50 or more years post-trauma. Thus, there is definitional and conceptual ambiguity in DSM–IV<sup>20</sup> that affects our understanding of delayed-onset PTSD. In fact, Spitzer et al<sup>3</sup> have proposed revised PTSD diagnostic criteria for DSM–V, changing the onset criterion (criterion E) to read as either `the symptoms develop within a week of the event' or `if delayed onset, the onset of symptoms is associated with an event that is thematically related to the trauma itself (e.g., onset of symptoms in a rape survivor when initiating a sexual relationship)'.

The finding that PTSD is more common in younger veterans suggests a possible cohort effect, which raises the question of whether there will be higher rates of delayed onset in future generations of military veterans. A previous finding from this sample¹⁷ was that veterans 65 years of age report lower PTSD and better mental health than those under 65. Other studies have also found evidence of cohort effects, with lower rates of PTSD among Second World War veterans³⁰ relative to studies of Vietnam veterans. Several possible explanations may account for this finding. First, people may become more psychologically healthy as they age (e.g. a maturational ageing process). Second, older veterans may be less likely to acknowledge psychiatric symptoms that exist (e.g. a sociocultural cohort effect related to `self-reliance' or perceptions regarding stigma for mental illness). Third, veterans with psychiatric problems may be less likely to survive to advanced age (e.g. a mortality effect). Last, younger veterans may be more sensitive to and more likely to report psychiatric symptoms based on changing social expectations (e.g. evolving interpretations and perceptions of psychiatric illnesses or a social learning effect).<sup>31</sup> This fits with general international trends towards higher levels of psychiatric disability among younger generations, as well as disorder-specific nuances such as the finding that `flashbacks', which are a common symptom among recent combat veterans, are conspicuously absent among veterans' symptom reports prior to the Vietnam War.³² There is little basis from the current data to expect significant rates of future delayed-onset PTSD in those younger veterans currently without PTSD symptoms in this sample. However, one might wonder whether delayed-onset PTSD, like `flashbacks', is possibly a culturally bound expression likely to become more prevalent in the future.

This study has important limitations inherent in the cross-sectional, retrospective nature of its design. Certainly, there is reason to be concerned about the potential for instability of recall and various memory biases.^1,33–35 Thus, recall or report bias cannot be excluded in the ascertainment of combat experiences, PTSD symptom severity or the time of onset for PTSD symptoms. Data show that military veterans' reports of combat exposure and other military hazards can change over time³⁶ and may even be subject to exaggeration.³⁷ Further, recall for symptom onset that may have occurred as long as 40 or 50 years ago is likely to involve a certain degree of imprecision. We acknowledge this as an important limitation of this study, as it is for most studies reporting on past trauma exposures or delayed-onset PTSD. However, as Andrews et al¹ note, there are also disadvantages to prospective studies, given that `individuals may have had onsets of PTSD after one assessment that then remitted before the next'. In other words, many existing longitudinal PTSD studies have significant flaws with regard to precision of onset estimates. For example, in a 20-year longitudinal study, veterans (n = 214) were assessed at four time points (1 year, 2 years, 3 years and 20 years post-combat) with a 17-year gap between time point three and time point four.¹⁴ Thus, it is not clear when PTSD onsets after time point three occurred, and the sample was so heavily pathological at the study outset (61% of the sample had a history of combat stress reactions, 45% met criteria for PTSD at time point one) that generalisability is a concern. Thus, there remains a large gap in the knowledge base regarding delayed-onset PTSD, especially among veterans and existing studies may even contribute to potential misconceptions about the prevalence of delayed-onset PTSD.

The noted limitations of our cross-sectional design are balanced by important study strengths, including a large, representative sample of veterans from Veterans Affairs primary care clinics, inclusion of a long time frame post-trauma (typically 25–40 years in this sample) for potential PTSD onset to occur, and use of widely accepted, psychometrically robust structured clinical interviews for evaluating traumatic event exposure and PTSD assessment. This study design would have met Andrews et al's¹ suggested criteria for inclusion in their systematic review of prevalence studies on delayed-onset PTSD and therefore represents a meaningful addition to an important, yet understudied and likely misunderstood, conceptual aspect of PTSD. Even if we stipulate that current estimates of PTSD symptom onset in our sample are unlikely to be precisely accurate, it seems plausible that estimates are at least in the `ballpark'. Notably, there is no recency effect for reported PTSD onset, as one might expect given data that people may forget past episodes of other psychiatric disorders.<sup>34</sup> In fact, the converse appears to be true here, where people with current PTSD recall their symptom onset as occurring many years before, shortly after the index trauma, and persisting chronically to the present time. Thus, if we collapse onsets into three general time periods we are left with the essence of an important finding: for those with current PTSD, symptom onset occurs primarily within the first 6 months after the index traumatic event (91.6%, 77/84), with no onset occurring after 6 years post-trauma. All told, the vast majority (96.4%, 81/84) of current PTSD onsets occurred within the first year post-trauma, and a similar pattern holds for subthreshold PTSD and lifetime PTSD only. Thus, PTSD symptom onset is remarkably consistent across our three groups (current, subthreshold, lifetime only). Moreover, the overall pattern of onset and the fact that we did not find a single incident of `late' delayed onset is consistent with the findings of the most methodologically rigorous large-scale prospective studies with civilian samples^5,6 and with historical perspectives regarding post-combat psychiatric reactions gained from experience with veterans of the First World War.³⁸