Table compares the two regression
Table 4 compares the two regression models for prostate cancer-specific deaths using the following data algorithms from the Death Registry: i) inclusion of all contributing causes of death; and ii) restriction to only coding for the primary causes of death. Similar to our findings in Table 3, we observed that the hazard ratios for both the Cox and Fine and Gray models were again very similar in magnitude and direction, where cancer stage, Gleason score and the Charlson Comorbidity Index score were associated with an increased risk of prostate cancer-specific death.
Discussion Although this was a nationwide study, it AMG-176 was limited by missing data with a limited sample size for chart review. Complete medical records were available in only 20% of the deceased cohort (670/3343), and data from general practitioners for deaths occurring outside of the hospital and ambulance setting was not available. However, the distribution of death categories for the subset with chart review was relatively similar to the distribution of death categories when comparing to data from the Death Registry restricted to primary causes of death (Table 2). Similar to other countries, autopsy rates are low in Denmark. Only 10% of all deaths and 20% of deaths occurring in-hospital in Denmark are autopsied compared to the autopsy rate of 75% in the 1970s . These trends are unfortunate since autopsy is the highest clinical standard for ascertaining the cause of death. Medical records review of the 670 cases was performed by the same physician for a consistent, albeit person-specific, evaluation of the cause of death. Also, particular attention was paid to the assessment of deaths with complicating events such as acute myocardial infarction, urosepsis, and pneumonia to discriminate whether or not prostate cancer was the underlying cause of death. There are several underlying reasons, taken alone or in combination, that may cause incorrect labelling of death from other causes as death from prostate cancer . The general perception of cancer as a severe diagnosis can, in and of itself, lead to registering prostate cancer as a contributing cause to death. Also, there may be a general misunderstanding as to how to code for causes of death by physicians. Since the death certificates allow for one primary cause of death and up to three contributing causes of death, cancer may likely be added as a secondary or tertiary cause of death to fill in the available lines with the best intention of detailed registration. Finally, in cause-specific mortality analyses, the methodological practice of using the primary cause of death or all contributing causes of death when defining cause of death varies in the literature, which complicates comparison across studies. Our findings are consistent with previous results from Great Britain, showing that in countries with low rates of asymptomatic PSA screening, prostate cancer remains a leading cause of death among men with prostate cancer. In 2013, an analysis of data from the Thames Cancer Registry showed that 50% of the deaths occurring in men with prostate cancer were due to prostate cancer itself . A 2016 study investigated 1236 U.K. death certificates in men from a prostate cancer trial cohort and attributed prostate cancer death to 42%; and reported high sensitivity (91%) and specificity (92%) of the U.K. death certificates . Other Scandinavian countries have investigated causes of death and the validity of death registrations among men with prostate cancer [9,19,20]. In a large validation study of 5675 deaths and the reliability of the Swedish Cause of Death Register, investigators reported an 86% agreement between Swedish registry data and medical records review, in contrast to our findings of 73% agreement for prostate cancer-specific deaths in the Danish death registry . They also found that misattribution of prostate cancer death increased with increasing age and comorbidity. The Swedish validation study defined prostate cancer death strictly as the primary cause of death registered in the Swedish Death Registry. A Norwegian study from 2018 of 764 deaths concluded that prostate cancer deaths were over-reported and misattribution of cause of death was particularly associated with increasing age (> 75 years old) . The Finnish Randomized Study of Screening for Prostate Cancer reported a 95% agreement after comparing 442 medical records with the Finnish cause of death registry . All of the above studies classified death as binary categories, i.e. prostate cancer-specific death and non-prostate cancer death. We extended the knowledge by differentiating between five death categories and by performing a comparative competing risks analysis of two regression models.