Several limitations have to be considered. First, in an ideal study, HRQOL should be measured in each subject before and after ICU admission, because the real interest is not “absolute” health but rather the change in perceived health. However, the HRQOL before treatment could only be scored retrospectively in the patients acutely admitted. This is difficult if not impossible without bias, because patients are still influenced by their critical illness. In a UK study, different authors used the SF-36 questionnaire at the time of discharge from the ICU as an instrument to score the premorbid HRQOL. However, their patients (medical and mixed ICU population) appeared to have a much lower pre-ICU score than the general population.13,42- 45 Nevertheless, the study of Watson et al36 showed that the preinjury scores in trauma patients were better compared with a general population. Also, there have been problems reported with the recall bias influencing preadmission HRQOL.36,43- 45 The possibility exists that the worse long-term HRQOL seen in our study could be associated with the higher prevalence of comorbidities and not solely the ICU admission. The systematic follow-up of ICU patients by Kvåle and Flaatten46 agrees that ICU patients in general have more comorbidity than the normal population. Nevertheless, only 30 patients (5%) had the presence of a preadmission disease (comorbidities, as scored in the chronic health points of the Acute Physiology and Chronic Health Evaluation [APACHE II]) in our study population. This could be an underestimation. The use of a different, more detailed scoring system may have yielded more specific information regarding the degree of comorbidities for the study cohort. From a methodological point of view, the nearest approach is to compare the self-reported health status of patients with that of a healthy reference population of corresponding age without ICU care. A second limitation of our study is that the HRQOL assessment was conducted only once. Ideally, assessment of HRQOL is conducted in a longitudinal design with multiple measurements over time.47 However, a design like ours does provide important and relevant findings, since health problems have most often stabilized some years after ICU admission. This hypothesis is also likely from our data; a comparison of the group with 6 years' follow-up and the group with 10 or more years' follow-up after ICU discharge revealed no significant difference in HRQOL (results not shown). Third, HRQOL measurement was done in 2006 at the end of the study follow-up time. Before HRQOL could be measured, 50% of the total study population (all surgical ICU patients, n = 1822) had already died.48 A group of 92 patients who died did survive a follow-up period of more than 6 years but were not alive anymore when HRQOL was measured. We chose to exclude these patients from this study. We assume that HRQOL would decrease even more if these patients were included and that our estimates of reported health problems should even be seen as conservative or optimistic.