Colorectal cancer treatment in people with severe mental illness: a systematic review and meta-analysis

Abstract

Aims

People with severe mental illness (SMI) have a greater risk of dying from colorectal cancer (CRC), even though the incidence is lower or similar to that of the general population This pattern is unlikely to be solely explained by lifestyle factors, while the role of differences in cancer healthcare access or treatment is uncertain

Methods

We undertook a systematic review and meta-analysis on access to guideline-appropriate care following CRC diagnosis in people with SMI including the receipt of surgery, chemo- or radiotherapy. We searched for full-text articles indexed by PubMed, EMBASE, PsychInfo and CINAHL that compared CRC treatment in those with and without pre-existing SMI (schizophrenia, schizoaffective, bipolar and major affective disorders). Designs included cohort or population-based case–control designs.

Results

There were ten studies (sample size = 3501–591 561). People with SMI had a reduced likelihood of surgery (RR = 0.90, 95% CI 0.92–0.97; p = 0.005; k = 4). Meta-analyses were not possible for the other outcomes but in results from individual studies, people with SMI were less likely to receive radiotherapy, chemotherapy or sphincter-sparing procedures. The disparity in care was greatest for those who had been psychiatric inpatients.

Conclusions

People with SMI, including both psychotic and affective disorders, receive less CRC care than the general population. This might contribute to higher case-fatality rates for an illness where the incidence is no higher than that of the general population. The reasons for this require further investigation, as does the extent to which differences in treatment access or quality contribute to excess CRC mortality in people with SMI.

Introduction

Cancer is a leading cause of mortality in people with a range of mental illnesses including severe mental illness (SMI) (Lawrence et al., 2000; Kisely et al., 2008, 2013a, 2016; Lawrence et al., 2013). For instance, they are 60% more likely to die from colorectal cancer (CRC) than the general population with CRC being second only to lung cancer as a cause of cancer death in this group (Kisely et al., 2008). The disparity is greater for people with SMI such as schizophrenia, major depressive disorder and bipolar disorder (Kisely et al., 2008, 2013a). This is despite the incidence of CRC in people with mental illness being the same or even lower than that of the general population (Lawrence et al., 2000; Kisely et al., 2008, 2013a, 2016). It is unlikely that this pattern can solely be explained by lifestyle factors following diagnosis such as diet or alcohol use.

Differences in cancer healthcare access and treatment may be another factor mediating the relationship between SMI and the increased risk of CRC mortality (Grassi and Riba, 2021). A recent systematic review found that women with SMI were less likely to be screened for breast and cervical cancer, although data on CRC were more limited (Solmi et al., 2020). An Australian study found that people with SMI had lower rates of screening in primary care for colorectal, prostate and cervical cancer after adjustment for age, gender and clinic visits (Tuesley et al., 2019). Australia has a National Bowel Cancer Screening Programme and there is ongoing research into whether there are similar disparities in participation for people with SMI (Protani et al., 2021). Reduced access to screening may therefore be one explanation for the finding that people with pre-existing mental illness are more likely to have advanced cancer stage at diagnosis, particularly those with SMI (Davis et al., 2020).

There is less information on care following diagnosis and it is possible that another contributor to higher mortality is differential access to guideline appropriate treatment such as resection and adjuvant radio- or chemotherapy (Brown et al., 2019; Grassi and Riba, 2021). We therefore undertook a systematic review of CRC treatment rates and modalities in those with and without SMI. A further aim was to assess if any differences in treatment between those with and without SMI were reflected in differences in subsequent mortality between the two groups.

Methods

Search strategy

This systematic review was registered with PROSPERO (ID CRD42021224360) and conducted according to PRISMA guidelines (Page et al., 2021) and recommendations for the reporting of meta-analyses of observational studies in epidemiology (Stroup et al., 2000). PubMed, EMBASE, PsychInfo and CINAHL were searched from inception to December 2021 to identify studies comparing CRC treatment in groups with and without pre-existing SMI. The search strategy included key terms for CRC, SMI and cancer treatments (surgical, systemic and radiation therapies) (see online Supplementary Table 1 for full list of search terms). There were no language restrictions. The reference lists of all eligible papers and related reviews were also scanned to identify any additional relevant studies.

Study selection

Studies were eligible for inclusion if they were cohort or population-based case–control studies of adults that reported original data on cancer treatment, stratified by pre-existing SMI status, in those with CRC. Studies that did not establish that the SMI diagnosis preceded the cancer diagnosis were excluded, as were those that did not include a representative comparison group of CRC patients without mental illness. Search results were imported into EndNote software, which was then used to eliminate duplicates. Articles were initially screened by title and abstract and then full text for their eligibility for inclusion in the review by pairs of reviewers working independently.

Data extraction and quality assessment

Data were extracted into an Excel spreadsheet and included study characteristics (country, sample size, age at diagnosis, years of cancer diagnosis, cancer type and staging), type and definition of psychiatric disorders, the cancer treatment of interest, effect estimates of the relationship between SMI and cancer treatment, and confounders adjusted for. Although some studies looked at predictors of subsequent mortality, none directly compared mortality in those with and without SMI as a result of any differences in treatment between the two groups. Study quality was assessed using the Newcastle–Ottawa Scale for cohort studies (Wells et al., 2011). Study selection, data extraction and quality assessment were independently conducted by three co-authors working in pairs with disagreements settled by consensus with or without the assistance of a fourth reviewer. Consensus was achieved in all cases.

Statistical analysis

Outcomes were the receipt of surgery, radio- or chemotherapy. Where data were available for three or more studies, they were combined in a meta-analysis using RevMan and Win-Pepi (Abramson, 2011). Odds ratios were converted to risk ratios (Zhang and Yu, 1998; Schünemann et al., 2019; ClinCalc.com). Where studies reported both crude and adjusted risk ratios, adjusted ratios were included in analysis. If there were at least 10 studies in a meta-analysis, we planned to assess for publication bias using funnel plots. We used an I² statistic value of greater than 50% as an indicator of significant heterogeneity. We explored any heterogeneity further through sensitivity analyses of the effect of omitting each study in turn. A random effects model was used for all analyses because of variation in studies between settings and methods.

Results

The search identified 13 153 citations, of which nine met the criteria for inclusion. We also included re-analysed data from a further published study by two of the present review's authors (SK and DL) (Kisely et al., 2013a). Although this had compared rates of colorectal surgery between those with any psychiatric disorder and the general population, separate data for schizophrenia/psychosis were available.

The three main reasons for exclusion were that studies did not examine treatment for CRC in people with SMI, did not evaluate the exposure (SMI) prior to cancer diagnosis, or did not report original outcome data (Fig. 1).

Fig. 1. Study selection process.

Notes: Exclusion criteria A: studies that did not examine treatment in SMI patients with colorectal/colon/rectal cancer. Exclusion criteria B: studies that did not examine SMI (schizophrenia/psychotic disorders/bipolar/major depression) that occurred prior to cancer diagnosis. Exclusion criteria C: studies which did not contain a comparison population (i.e. CRC patients without history of SMI). Exclusion criteria D: studies which contained a repeated analysis for the same outcome in a single population. Exclusion criteria E: Studies not reporting original data (e.g. reviews/text book chapters). Exclusion criteria F: Full text could not be retrieved/conference abstract contained insufficient information.

Characteristics of the ten studies are included in Table 1. These were all retrospective cohorts of patients diagnosed with CRC between 1988 and 2013 (colon = 1, rectal = 1; colorectal = 8). Three were conducted in the United States (Baillargeon et al., 2011; Wieghard et al., 2015; Ho et al., 2018), two in Canada (Kisely et al., 2012; Mahar et al., 2020), and one each from Australia (Kisely et al., 2013a), Denmark (Kaerlev et al., 2018), Finland (Manderbacka et al., 2018), Taiwan (Huang et al., 2018) and Japan (Ishikawa et al., 2016). The sample size ranged from 3501 to 591 561, with a median of 24 507. The median number of people with SMI was 1106 (range = 136–11 837).

Table 1. Characteristics of studies included in the systematic review (n = 10)

CRC, colorectal cancer; dx, diagnosis; NS, not stated; SMI, severe mental illness.

^a Excludes substance use disorder.

Four studies reported cancer treatment by psychotic disorders and mood disorders separately, three looked at combined SMIs and three studies considered only schizophrenia (Table 1). Pre-existing SMI diagnoses were identified using databases from insurance records, hospital admissions or outpatient/psychiatrist visits. One study also used prescriptions of antipsychotic medication or selective serotonin/norepinephrine reuptake inhibitors (SSRI/SNRIs) as indicators of schizophrenia, schizoaffective, bipolar and major affective disorders (Kisely et al., 2012).

Quality assessment revealed that all studies, except for one (Manderbacka et al., 2018), reported adjusted estimates controlling for at least age and cancer stage (Table 2). Other important confounders adjusted for in most studies included sociodemographic factors (e.g., race/ethnicity, income, rurality), comorbidities and year of diagnosis. In the one study that did not give adjusted estimates for treatment access, largely because this was not the primary outcome, the authors presented raw numbers on the presence of metastases stratified by sex. Using these it was possible to calculate that there were no significant differences for either males or females in the proportion of people presenting with metastases in the psychosis, mood or control groups (Table 3). Two studies did not explicitly demonstrate that cancer treatment outcomes occurred after, rather than before, the psychiatric diagnosis (Wieghard et al., 2015; Ho et al., 2018).

Table 2. Quality assessment of studies included in the systematic review using the Newcastle–Ottawa Scale

Notes: * indicates that the study met the criterion; N/S: not stated; †There were no significant differences between cases and controls in the presence of metastases at presentation.

Table 3. Presence of metastases on presentation from Manderbacka et al

Receipt of surgery

Table 4 presents results for the seven studies that examined surgical outcomes by SMI status. Outcomes including any operation, sphincter preserving or emergency surgery. Four studies compared the likelihood of any surgery in people with schizophrenia/psychosis to controls (Kisely et al., 2012, 2013a; Ishikawa et al., 2016; Manderbacka et al., 2018), although in one study this was combined with endoscopy (Ishikawa et al., 2016). Two studies presented results for mood disorders (Kisely et al., 2012; Manderbacka et al., 2018). In almost all comparisons, people in either diagnostic group were significantly less likely to receive surgery (Table 4). The one exception was the study by Manderbacka et al. (2018) that found non-significant results for females although they were still significant for males (Table 4). However, in the other study that also presented results by sex, there were no differences between males and females (Table 4) (Kisely et al., 2013a). There was little difference between the two diagnostic groups given overlapping 95% confidence intervals (Table 4).

Table 4. Results of studies examining receipt of surgery for CRCs, by SMI type

dx, diagnosis.

^a Restricted to stage I-III patients.

^b Odds ratios were presented, rather than relative risks.

^c Study included both gastric and CRC included (~50% CRC).

^d Restricted to stage II/III patients.

We were only able to meta-analyse results for schizophrenia/psychotic disorders and this confirmed the findings from the majority of individual studies that people with SMI were less likely to receive any surgery (RR = 0.90; 95% CI 0.84–0.98; I ² = 76%; p = 0.003) (Fig. 2). Omitting the study that combined colorectal surgery with endoscopy did not alter the findings (RR = 0.85; 95% CI 0.74–0.97; I ² = 81%; p = 0.02).

Fig. 2. Receipt of surgery in people with schizophrenia/ psychotic disorders.

In terms of other surgical outcomes that could not be meta-analysed, people with either schizophrenia/psychotic or mood disorders were significantly less likely than people without SMI to receive sphincter preserving as opposed to non-sphincter preserving rectal surgery on adjusted analyses in one study (Table 4) (Wieghard et al., 2015). In another study, people with all forms of SMI were also significantly more likely to require emergency colorectal surgery, even after adjustment for important confounders (RR = 1.25; 95% CI 1.04–1.50) (Ho et al., 2018).

A final study reported on the likelihood of not receiving CRC surgery in people with SMI who had an inpatient psychiatric history and those who had only been outpatients (Table 4). Only those with an inpatient history were less likely than non-psychiatric controls to have had surgery on adjusted analyses (RR = 2.15; 95% CI 1.07–4.33).

Receipt of adjuvant therapy

Four studies examined the receipt of adjuvant therapy such as chemo- and radiotherapy therapy, all but one presenting adjusted results (Table 5). However, in the latter case, there were no significant differences between cases and controls in the presence of metastases (Table 5). In two studies, the outcome was any adjuvant therapy (Kaerlev et al., 2018; Mahar et al., 2020), in another, the non-receipt of chemotherapy (Baillargeon et al., 2011), and in the fourth, the receipt of chemotherapy and radiation therapy separately (Manderbacka et al., 2018).

Table 5. Results of studies examining receipt of adjuvant therapies (chemotherapy and/or radiation therapy), by SMI type

SMI, severe mental illness; RR, relative risk; CI, confidence intervals; dx, diagnosis; SEER, Surveillance; Epidemiology and End Results database.

^a Restricted to stage III patients.

^b Odds ratios were presented, rather than relative risks.

^c Restricted to stage II/III patients.

Three studies found that those with SMI were less likely to receive adjuvant therapies (Baillargeon et al., 2011; Kaerlev et al., 2018; Mahar et al., 2020). One of the three studies stratified by SMI severity and reported that participants who had previously received inpatient psychiatric care were significantly less likely to receive adjuvant therapy (RR = 2.07; 95% CI 1.72–2.50) than those with SMI only receiving outpatient care (RR = 1.22; 95% CI 1.00–1.49) (Mahar et al., 2020). However, in another that presented results separately by diagnostic group, the likelihood of not receiving chemotherapy was similar for people with psychotic illness (RR = 1.56; 95% CI 1.21–2.03) and mood disorders (RR = 1.27; 95% CI 1.10–1.46) as shown by overlapping 95% confidence intervals (Baillargeon et al., 2011).

There were mixed findings in the fourth study which presented separate results for males and females. In the case of radiotherapy, both males and females with either psychotic illnesses or mood disorders had the same likelihood of treatment as the controls (Manderbacka et al., 2018). This was also true for the receipt of chemotherapy in females. However, males were significantly less likely to receive this treatment irrespective of diagnostic group.

Other outcomes

A study from Taiwan examined CRC palliative care outcomes in patients with and without schizophrenia between 2000 and 2012 (Huang et al., 2018). This included palliative care consultation services (OR = 0.59; 95% CI 0.43–0.82) and chemotherapy (OR = 0.60, 95% CI 0.55–0.66). By contrast, they were more likely to receive intensive care treatment (OR = 1.21, 95% CI 1.07–1.36) or invasive interventions, such as cardiopulmonary resuscitation (OR = 1.34, 95% CI 1.15–1.57) (Huang et al., 2018). There were no significant differences in the use of hospice ward or home care (Huang et al., 2018). As noted previously, there were no studies that assessed if any differences in treatment between those with and without SMI were reflected in differences in subsequent mortality between the two groups.

Non-receipt of any treatment

Two studies reported on the non-receipt of any CRC treatment in people with psychotic disorders or mood disorders (Baillargeon et al., 2011; Manderbacka et al., 2018). The first study reported an increased risk of non-receipt of treatment in people with psychotic illness (RR = 1.42; 95% CI 1.13–1.78) and mood disorders (RR = 1.28; 95% CI 1.08–1.52) compared to those without mental illness (Baillargeon et al., 2011). This study adjusted for age, stage, race, ethnicity, sex, marital status, region, income, comorbidity, and year of diagnosis. The second study also observed an increased risk of non-receipt of treatment in those with psychotic illness, but inconsistent results for severe mood disorders (Manderbacka et al., 2018). Although unadjusted, these results were stratified by sex and there were no significant differences between cases and controls in the presence of metastases at presentation.

Heterogeneity and publication bias

The results for the receipt of surgery in SMI showed significant heterogeneity (Fig. 2). We therefore explored this by excluding each study in turn in every analysis. The omission of the unpublished re-analysed data that came from one of the studies (Kisely et al., 2013a) resulted in an I² of less than 50% (RR = 0.93; 95%CI 0.89–0.97; p = 0.0008; I ² = 46%). We were unable to analyse for the effects of publication bias as none of the analyses had 10 or more studies.

Discussion

This systematic review identified a small number of studies (n = 10) examining CRC treatment in those with and without SMI. Despite significant inter-study heterogeneity, those with SMI appeared to be generally less likely to receive CRC treatment (any treatment, surgery or adjuvant therapy) compared to those without SMI. These differences persisted after adjustment for socio-demographic variables and cancer stage at presentation. The latter is an important potential covariate given that people with pre-existing mental illness are more likely to have advanced cancer stage at diagnosis.

These overall findings are consistent with studies examining treatment for other cancer sites such as breast and cervix. For instance, people with SMI were less likely to receive guideline recommended treatment for breast cancer (Mahabaleshwarkar et al., 2015) (Dalton et al., 2018) and encountered greater delays before initiation of therapy than those without SMI (Iglay et al., 2017; Haskins et al., 2019).

This mirrors findings for other chronic physical illness such as cardiovascular disease and diabetes in studies in people with SMI from the United States, Canada, Australia, and Great Britain (Druss et al., 2001; Hippisley-Cox et al., 2007; Kisely et al., 2007, 2009; Kilbourne et al., 2008; Mitchell et al., 2009; Lawrence and Kisely, 2010). For instance, psychiatric patients are less likely to have their weight or blood pressure measured in primary care or be assessed or treated for hyperlipidaemia despite physician consultation rates being generally high in people with SMI (Jablensky et al., 2000; Hippisley-Cox et al., 2007; Kilbourne et al., 2008). In secondary care, psychiatric patients are less likely to receive specialist procedures such as cardiac catheterisations and coronary artery bypass grafting than the general population, even though their mortality rates for the same conditions are significantly higher (Kisely et al., 2007, 2009). On discharge from hospital following myocardial infarction, they are also less likely to be prescribed beta-blockers and statins (Kisely et al., 2009).

In terms of variations within the SMI group, males were significantly less likely to receive chemo- or radiotherapy than the general population while rates for females were no different. Although not the focus of the present study, there were mixed findings on sex as a predictor of CRC treatment in overall study samples. For instance, females were less likely to require emergency resection but more likely to have sphincter-sparing surgery in adjusted analyses from two studies (Wieghard et al., 2015; Ho et al., 2018). By contrast, there were no differences between males and females in the overall receipt of surgery ± endoscopy in two further studies (Kisely et al., 2013a; Ishikawa et al., 2016). The reasons for these conflicting results are unclear but are reflected elsewhere. On one hand, a care pathways study in Great Britain found that while CRC incidence was higher in males, subsequent access to services was generally the same for both sexes (White et al., 2018). On the other hand, qualitative work found that males and females with CRC had different treatment experiences (Brewer et al., 2020). We did not find larger disparities in participants with psychotic illnesses compared with severe mood disorders although those who had previously received inpatient psychiatric care were significantly less likely to receive adjuvant therapy than people with SMI only receiving outpatient care (Mahar et al., 2020).

Possible mechanisms

There are several possible explanations as to why people with SMI who are diagnosed with CRC are less likely to receive guideline recommended cancer treatment. Firstly, those with SMI are more likely to have higher comorbidity burdens (e.g., cardiovascular disease, chronic obstructive pulmonary disease, obesity, diabetes) compared to those without SMI (Viron and Stern, 2010; Janssen et al., 2015; Onyeka et al., 2019). This may influence clinician/patient decision making around the provision of treatment such as chemotherapy (Gross et al., 2007; Boakye et al., 2021). However, most studies in this review adjusted for differences in comorbidities and still identified treatment disparities, so this is unlikely to be the primary mediating factor. Clinical decisions on chemotherapy may also be influenced by concerns over potential interactions between particular anti-neoplastic agents and some psychotropic medications such as clozapine (Yap et al., 2011). Another mechanism for reduced surgery could be the perception of poor post-operative outcomes in those with SMI (Irwin et al., 2014; McBride et al., 2018). For instance, people with schizophrenia have higher rates of complications and mortality following surgery. These include respiratory failure, sepsis, deep venous thrombosis, pulmonary embolism, paralytic ileus, stroke, and delirium (Irwin et al., 2014).

Other explanations for our findings could include health service access and/or patient treatment adherence. For instance, three of the studies were from the United States where people with mental illness may face barriers to private health cover. It is also possible that people with SMI are treated differently by medical professional with negative attitudes or stigma leading to disparities in care (Thornicroft, 2008; Ostrow et al., 2014). Finally, ‘overshadowing’ may contribute to delays in diagnosis or treatment (Jopp and Keys, 2001; Giddings, 2013; Jones et al., 2008). This is the tendency to regard somatic symptoms such as decreases in energy, appetite or weight as being due to an underlying psychiatric disorder (Giddings, 2013), or that the presence of psychiatric co-morbidity adversely affects the quality of care (Jopp and Keys, 2001). This might include an unwillingness to address possible barriers to appropriate treatment.

Limitations

There are several limitations to these findings. Firstly, there was a large variation in the definitions of SMI and the reference categories used. While all studies used medically diagnosed SMI from hospital/insurance records, definitions included any psychotic illness, only schizophrenia, the prescription of anti-psychotic medication, any mood disorder, severe mood disorder and any SMI. In addition, the reference categories ranged from the absence of SMI to that of any mental disorder, making direct comparisons between studies difficult. Most studies also did not incorporate markers of severity within their definition of psychiatric illness. In the one that did, people with SMI who were treated as inpatients were more likely to have greater treatment disparities than those managed as outpatients (Mahar et al., 2020).

From an oncology perspective, the majority of studies used simple binary measures of receipt/non-receipt of CRC treatment. However, there are several other indicators of the quality of cancer care that have, thus far, received limited attention in those with SMI and cancer. Outcomes such as lower chemotherapy relative dose intensity (which considers both chemotherapy dose reductions and delays between cycles), early cessation of chemotherapy/radiation and longer time between diagnosis and initiation of cancer treatment have all been shown to be associated with poorer long-term outcomes such as increased recurrence and poorer cancer survival (Lyman, 2009; Cone et al., 2020). There was also no standard definition of guideline-appropriate CRC care.

None of the included studies examined the reasons for the differences in treatment rates in those with and without SMI. We are, therefore, presently unable to distinguish whether treatment disparities are due to lack of patient adherence (being offered treatment with subsequent refusal) or inequitable access to treatment (not being offered/having access to relevant treatment options). Identifying the cause of treatment disparities will help to determine the best avenues for intervention, e.g., clinician-based education, enhanced multidisciplinary team meetings or better patient support and education around the processes and benefits of cancer treatment. In addition, more information is required about the effect of treatment disparities on subsequent outcomes such as mortality.

Other limitations to this research are that we only undertook backward citation searching in retrieved articles and did not search forwards. We also did not calculate agreement between reviewers on the inclusion of studies but resolved any disagreements through consensus with the assistance of a third author if required. We were only able to meta-analyse the results for one outcome, the receipt of surgery. Despite being statistically significant, the likelihood of surgery was reduced by less than 10%. The results also showed heterogeneity. Although we used a random effects model to incorporate heterogeneity into our analyses and the I² value was no longer significant with the removal of one study, our findings should still be viewed with caution. We were also unable to test for publication bias.

Possible interventions

In terms of possible interventions, a small study from Japan reported that case management including education and patient navigation for CRC screening in people with schizophrenia resulted in greater participation than treatment as usual (Fujiwara et al., 2021). This approach might also be applied to CRC treatment following diagnosis, including the use of navigators, possibly in combination with collaborative care between general practitioners, oncology and mental health services (Irwin et al., 2014). One particular focus might be people with SMI who have been lost to psychiatric follow-up. As an example, re-engagement in psychiatric care was associated with a six-fold reduction in mortality, including that due to cancer (Bowersox et al., 2012; Irwin et al., 2014). In another, improved community or outpatient follow-up by mental health teams led to reductions in all-cause mortality, the vast majority of which was due to medical illness (Kisely et al., 2013b). These interventions should be combined with efforts to address stigma, patient factors (e.g., lack of trust), clinician factors (e.g., inadequate training), and healthcare fragmentation (e.g., between psychiatry and oncology) (Grassi and Riba, 2021). Further research is indicated into where along the CRC care pathway, from screening to end of life care, barriers to intervention occur and the reasons for these (Protani et al., 2021). In particular, people with experience of SMI and CRC, or their carers, should be asked about their experience of barriers and enablers to treatment (Protani et al., 2021).

Conclusion

In conclusion, this review identified a small number of studies on the receipt of CRC treatment in those with and without SMI. There was consistent evidence that those with SMI received less CRC treatment than those without SMI, and that the disparity may be greater in those who have required inpatient treatment. Despite this, there is limited understanding as to why those with SMI experience CRC treatment disparities, and to what extent these treatment differences may mediate the excess CRC mortality observed in those with SMI.