Post-traumatic reactions have been reported by parents of children with cancer in numerous studies during the past decades. However, research has not revealed which aspects of the cancer experience that are challenging enough to cause these reactions. Previous studies have rarely identified any associations between disease-related factors and long-term PTSS among parents. Reasons could be that the cancer diagnosis per se is the traumatic event, and that the specific experiences attached to the individual case are of less importance, or that previous studies did not cover the key potentially traumatic events. In the present study we aimed at extending the knowledge about which factors that predict long-term PTSS among parents of children struck by cancer.
A medically more troublesome disease trajectory is related to parental stress 
. However, in the multivariate model a finally fatal disease was the only objective disease-related factor that demonstrated predictive power for PTSS.
Not surprisingly, the death of a child was one of the two strongest predictors of parental PTSS, also when analyzed together with other potentially traumatic aspects of childhood cancer. The death of a loved one is a truly traumatic event. However, a resemblance between PTSS and grief may contribute to an overestimation of PTSS reported by bereaved persons 
. Nonetheless, we conclude that parents who meet this tragic ending are vulnerable to post-traumatic stress and/or prolonged grief 
For parents whose child had survived their disease at the time when PTSS were assessed, none of the objective medical events targeted in the present study seemed to be traumatic enough to produce lingering post-traumatic stress symptoms. Since fear of relapse is known to be a prominent stressor 
, one could assume that an actual relapse should be such an event. Yet, non-fatal relapse did not predict PTSS. This is in line with results from a previous Swedish cross-sectional study 
, although findings by others have been ambiguous about this. Jurbergs et al. 
have reported that a child's cancer-related relapse predict parental traumatic stress symptoms. However, their cross-sectional sample most likely includes parents in acute crisis, and is not comparable with the present sample. Medical complications associated with the child's treatment, such as serious infections, ICU treatment et cetera are definitely stressful when occurring 
, but do not seem to produce enduring PTSS.
Poor prognosis and intense treatment evidently correspond with a fatal disease development. Accordingly, the death of the child may be the underlying factor explaining later post-traumatic stress symptoms in those parents, while general factors accompanying a more problematic disease and treatment do not automatically produce lingering PTSS.
Every parent can certify that experiencing that one's child suffers arouses parental distress 
and the findings indicate that parents' perceptions of their children's suffering predict long-term PTSS. In the multivariate model, the perceived number of child symptoms and child psychological distress demonstrated a significant predictive power for PTSS. However parental perceptions of physical symptoms in the child seemed to be protective, when its impact on PTSS was analyzed together with other predictors. This result is difficult to interpret, and therefore we refrain from speculations. The implication of this needs to be further explored.
It is well documented that unemployment is a risk factor for PTSS as well as for other mental health problems 
. In our categorization, non-employed includes being job-seeking and unemployed and/or on long-term sick-leave at the time of the child's diagnosis. Tentatively we suggest that the normalizing milieu including support from the social network of a workplace may buffer against mental health problems.
Supporting previous findings 
, being an immigrant was shown to substantially predict PTSS. A reasonable assumption could be that immigrants more often than non-immigrants are troubled by consequences of previous trauma, making them more vulnerable to an additional trauma. However, when analyzed in the multivariate model, previous trauma did not predict PTSS. Instead, the explanation may be sought in insufficient social support, and cultural differences in the connotations of illness and the communication with the health care staff 
. Noteworthy is that gender did not predict PTSS when analyzed with other variables, indicating that mothers and fathers alike may develop long-term post-traumatic stress in the face of childhood cancer.
Certain limitations are attached to the present study. Firstly, we addressed only some of several possible predictors of PTSS in parents of children with cancer. Through the selection or the means of assessing those factors we may have failed to spot important issues, regarding for example the individual child's reactions to specific treatment procedures. In addition, any previous traumata and their consequences should be analyzed in more detail in future studies. Moreover, assessing PTSS through self-report questionnaires is a cost-effective approach in large samples like this, but a face-to-face clinical interview would most certainly capture the concept in a more correct way. Strengths of the study include its population based longitudinal design and relatively large sample.
It may seem inconsistent to include factors in the multivariate model, which had not shown association with PTSS in the univariate analyses. However, the factors satisfaction with care and non-fatal relapse were included for an explorative purpose for the following reasons: Satisfaction with care was considered to potentially indicate a feeling of security and a safe environment, which could be a protective factor against post-traumatic stress, and non-fatal relapse was considered to potentially indicate a re-traumatization, which could be a risk factor for post-traumatic stress (repeated trauma has been shown to be a risk factor for PTS). Although these factors were not related to PTSS in the univariate analyses, there was a hypothetical possibility for impact in the hierarchical clustered model.
In conclusion, parental traumatic stressors in childhood cancer seem not to be found in treatment complications, but in parents' subjective perceptions of their child's suffering. Relapse may be a severe stressor, but for those whose child survives, the fear evoked by a relapse typically subsides and does not leave post-traumatic stress symptoms. Moreover, although the chronic stress of present problems and feared future difficulties bring about exhaustion in parents 
, the death of a child remains the ultimate trauma in the childhood cancer experience. In addition, certain demographic factors previously recognized as risk factors for mental health problems point to a vulnerability to PTSS in parents of children with cancer: being an immigrant and being unemployed. We may well assume that these more vulnerable parents are less apt to ask for support from the paediatric medical service.
There is reason to emphasize the clinical implications of the present findings. Parents' perceptions of their child's situation should always be considered since these may have a significant impact on long-term parental mental health. Parents' mental health may in turn have an impact on their child's mental health as well as on the communication with the health care professionals. We cannot presuppose that parents of children with a medically unproblematic journey through disease and treatment do not run a risk of lasting post-traumatic stress. In addition, we should be as attentive to fathers' long-term distress as we are to mothers'. Moreover, when a parent loses his or her child, we should keep in mind that signs of PTSS or prolonged grief may indicate a condition that requires professional psychological treatment.