Colon cancer patients with a serious psychiatric disorder present with a more advanced cancer stage and receive less adjuvant chemotherapy - A Nationwide Danish Cohort Study

The study is based on the nationwide Danish Colorectal Cancer Group database (DCCG) including incident patients with CC and RC during 1 January 2007 until 31 December 2013. The DCCG does not register recurrences of the cancer after the primary operation. The database is more than 98% complete for all CRC patients in Denmark of all ages (approximately two-third CC, one-third RC) [10]. The database contains detailed data on the priority or urgency of the operation of the patients with CRC, and the variable has the categories: elective operated patients, acute operated patients, and patients with missing information regarding this issue.

The DCCG has a variable “operative aim”, with the three categories: intended curative or palliative aim of the operation, or missing information regarding this issue.

The DCCG also has a categorical variable “Treatment” based on the decisions made by the multidisciplinary cancer team (MDT). We selected for the present study only those patients in the DCCG having had an operation for CC or RC based on this variable (thus excluding those patients without having an operation, or missing information on operation (< 8% of cases), and after this restriction of the study population the variable had no missing values. The variable on treatment for all patients who were operated had the following categories: whether there was operation only, operation followed by adjuvant oncological treatments, neoadjuvant oncological treatment followed by an operation or operation with both neoadjuvant and adjuvant oncological treatments. This information was combined into a new variable: having oncological treatment (yes/no) -in addition to the operation.

The DCCG contains data on cancer type and on the clinical and the pathological cancer stage. The clinical TNM classification, cTNM or just TNM, is based on e.g. clinical, endoscopic and image diagnostic findings at the time of diagnosis and is essential in relation to decision-making on e.g. operation and evaluation of treatments. The pathological TNM classification, pTNM, is based on the histopathological examination of the operative specimen, possibly modified by per- and postoperative findings, and forms the basis for decision-making on the postoperative oncological treatment and prognostic evaluation [24]. In the DCCG, the 5th edition of The Union for International Cancer Control (UICC)'s TNM-cancer staging classification was used until 2016 (when it was replaced by the 8th edition) [25‐27]. The pathological UICC cancer stage variable was grouped into stage I-II versus III-IV.

Persons with a permanent address in Denmark have a unique 10-digit civil registration number, which includes information on birthday and gender, dead/alive and immigration status in The Danish Civil Registration System [28, 29]. We used this number to link the CRC cohorts in the DCCG with each CRC patient’s hospital contacts as recorded by the Danish National Patient Registry [30] for information on the exposure of serious psychiatric disorders and on comorbidity [31‐33], and public registries developed by Statistics Denmark such as The Household and Family Statistics (based on The Danish Civil Registration System) as well as the personal income statistics to calculate educational level, and socio-economic position group [34].

A total of 26,897 patients with an operation for CRC were retrieved from the DCCG during the 7-year period 1 January 2007 until 31 December 2013. Our study period ended just before the Danish screening programme for CRC in the general population aged 50–74 years was implemented 1 March 2014.

A priori in the study design phase we decided to make restrictions to the study population. Observations were excluded if the patients were not operated for the cancer with a registered date of the operation. To avoid that perioperative complications had influence on the choice of treatments, and to ensure that the patients were alive long enough to be offered an oncological treatment, observations were excluded if the patients died within the first 30 days after their operation.

For the analyses of CRC patients having received a palliative vs an intended curative aim of the operative treatment all the CRC patients were included for the analyses (cohort 1) and split by having a pre-existing psychiatric disorder (yes/no).

For the analyses of CRC patients having received at least one oncological treatment in the DCCG before or after the cancer operation we stratified the analyses into CC patients only (cohort 2) and RC patients only (cohort 3).

Each CRC patient was linked to The National Patient Registry to collect information on hospital contact for: ICD-10 DF20–29: Schizophrenia, schizotypal and delusional disorders (primary psychotic disorders), or ICD-10 DF30–39 mood (affective) disorders as the primary diagnosis in the time-period from 10 years to 120 days prior to the date of the CRC operation [30]. If yes, the patients were regarded as exposed, otherwise not-exposed. Exclusion of psychiatric disorders diagnosed up to 120 days before the date of the cancer operation was done to avoid that natural psychological crisis reactions as a result of the cancer diagnosis were misclassified as pre-existing psychiatric disorders.

The CRC patients (with and without a psychiatric history) were examined in the DCCG for the cancer stage at the time of the operation, and for having received palliative vs an intended curative aim of the operative treatment, and for receiving at least one oncological treatment either before or after the cancer operation [10].

Using the unique civil registration number, we obtained the gender and date of birth, and migration and death. The UICC pathological cancer stage variable was taken from the DCCG. Information on co-morbidity was obtained from The National Patient Registry merged with the DCCG, with calculation of a slight modification of the Charlson Co-morbidity Index by excluding all contacts with CRC prior to the date of the CRC operation from the comorbidity index score [31‐33]. Each of the 19 categories of comorbid diseases was assigned a severity score between 0 and 6, and the comorbidity index was calculated as the sum of the scores (possible range between 0 and 37 for each patient). The predictive value of the comorbidity index is high [32]. Cancer diagnoses within 150 days before the date of CRC operation may be CRC misclassified as other cancers and was not included in the comorbidity index. The comorbidity index was re-coded as a score into a categorical variable (0, 1, and above 1 point).

The registries developed by Statistics Denmark was used to calculate socio-economic position, which was included as quartiles in the analyses (for the main analyses yearly income below DKK 143,927, DKK from 143,927–178,573, DKK from 178,573-247,677.4, and above DKK 247,677.4) [34]. Educational level from Statistic Denmark was included in the analyses as a categorical variable: 1) elementary school only, 2) upper secondary education/vocational education and training 3) qualifying educational programmes/vocational bachelor education/bachelor programmes/masters programmes/PhD programmes.

We compared the CRC patients, who have had a psychiatric disorder with the CRC patients without a psychiatric history with respect to demographic characteristics and cancer stage at the time of the operation with Pearson chi-squared tests. Continuous variables were compared between the groups with Students t-test. The date of the operation was the index date in the study.

Specifically for the descriptive analyses for the patients with CRC and the analyses for cancer stage, we used all the patients with CRC and we further stratified the analyses into the acute operated CRC patients only, and CRC patients who had undergone an elective operation only, and excluding those patients with missing information regarding this issue.

The distribution of the cancer treatment by pre-existing psychiatric disorder status was calculated (provided in Table 3). For the analyses of palliative vs an intended curative aim of the operative treatment, we used all the CRC patients. For the analyses of the oncological treatment outcomes all the CRC patients was split by anatomical site of the cancer, CC or RC. We estimated the odds ratio (OR) with 95% confidence intervals (CIs) for the patients with a psychiatric history for having received palliative vs an intended curative aim of the operative treatment, and for having at least one oncological treatment (before or after the operation) by logistic regression.

We adjust for each of the possible important confounders as described in the CRC literature: gender, age and the Charlson comorbidity index score, pathological cancer stage (UICC cancer stage I-II versus III -IV) [24], socio-economic position (four levels), and educational level (three levels) in the adjusted analysis model. After each adjustment step we evaluated the corresponding OR with 95% CI.

We have analyzed the correlation between each of the possible confounders and between having had a psychiatric disorder and the confounding factors.

Finally, sub-analyses were performed with restriction to primary psychotic disorders only, affective disorders only, and to elective operated patients or to acute operated patients only.

The strengths of our study were the large Nationwide Danish study population with close to complete background information and follow-up from highly valid registries with prospectively collected patient data thus minimizing selection bias, comprehensive measures of potential confounders and nearly complete information on the chosen diagnoses and treatments for the entire study population [10, 30].

There may have been a difference in the initial CRC stage and comorbid disease stages in patients with a pre-existing psychiatric disorder compared with patients without a pre-existing psychiatric condition, and especially among those patients who died within the first 30 days. By only selecting patients, alive at least 30 days after surgery for the study population may have removed proportionally more patients with a pre-existing psychiatric disorder with a more advanced disease. The finding of a more advanced disease in the psychiatric group might thus be underestimated, although we have adjusted the analyses for cancer stage as a possible confounding factor.

The “treatment” variable had less than 8% missing values. Still, it is possible with underreporting in this variable of adjuvant oncological treatment. Since this rather few missing values of the treatment variable is not expected to differ between patients with and without having pre-existing psychiatric disorder, it is not expected to bias the results [23].

We have included possible confounders previously mentioned in the CRC literature in the models. Educational level may to some degree be correlated with socio-economic position. We checked for each of the investigated outcomes whether stepwise adjustment for the possible confounding factors age, gender, Charlson Comorbidity Index score, cancer stage at the time of operation, socio-economic position, and educational level had statistically significantly influenced the OR. This was not the case. Furthermore, the coefficients estimates of the possible confounding factor socio-economic position and educational level was only weakly correlated (correlation coefficient approximately 0.05) to the estimate of psychiatric history. Therefore such correlations may not explain the findings. Hospital treatments are free of charge in Denmark and the majority of patients with symptoms of a severe psychiatric disorder will be referred at least once to and diagnoses at a public hospital and thus registered in The National Patient Registry. However, in Denmark and e.g. Finland a reduction in psychiatric hospital beds and an increase in out-patient services have occurred in the recent decades [4] and our study does not cover patients with e.g. depressive disorders only seen by general practitioners, which may lead to an underestimation of the associations seen.

Since the psychiatric medication may constitute a contraindication for oncological treatment for CC, and oncological treatment requires patient compliance which can be difficult for patients with psychiatric disorders to follow, these factors may to some degree explain the findings [10‐12].

Other possible explanations may include the health behaviors of the patient, difficulties with regard to communication, or failures in the primary or secondary healthcare system for various reasons.