Management of patients with incurable colorectal cancer: a retrospective audit

Patient selection

All patients of any age diagnosed with colorectal cancer from a single hospital trust between 2009 and 2016 recorded on the Somerset Cancer Record (SCR) were identified. Of these, any patient determined to be unsuitable for curative surgery as the primary intent, as documented at the multidisciplinary team meeting (MDT), was included. Prior to 2009, almost all MDT records were not stored on the electronic SCR and as such were not readily accessible or accurate so data were only collected from 2009 onwards.

While annual caseloads were not uniform within our unit, over the preceding 5 years the unit managed approximately 150 new diagnoses each year as recorded on the SCR, of which 20–30 were managed without curative intent. We performed a screening of our data prior to commencement to confirm sufficient numbers to permit exploratory analysis and further sub-group analyses.

Information extracted included demographics, site, TNM staging, age at diagnosis, age at death (where applicable), treatment modalities, survival (taken as time from diagnosis to death) and the predominant reason precluding curative surgery as judged by the MDT. Fitness for surgery was a clinical decision based on anaesthetics assessment via MDT; however, the ultimate decision lay with the operating surgeon. Where patients were alive at the end of the study period a fixed end-point (4 April 2017) was used and median survival was determined.

Patients were followed up using the SCR and cross-checked with electronic hospital systems for evidence of readmission and subsequent intervention including electronic patient letters, discharge summaries, pathology reports and imaging systems to ensure follow-up until April 2017.

Outcome measures

The primary outcome was overall survival (1-year and overall) from diagnosis in patients with colorectal cancer deemed unsuitable for curative surgery at colorectal MDT. This was defined as follows: any patient with colorectal cancer who was felt at MDT to have disease that was unresectable due to advanced local disease or metastatic to an extent that precluded curative treatment; patient frailty associated with high risk of operative morbidity and mortality; or patient choice. Frailty was taken to be any patient deemed by the MDT to be unsafe for major surgery owing to comorbidity or poor performance status or a combination thereof. The concept of frailty remains an emerging and evolving surgical concept, and while a number of scoring systems are being increasingly used, these are not part of routine practice. The MDT routinely comprised five consultant colorectal surgeons, two consultant radiologists, a consultant histopathologist, two consultant oncologists, two colorectal nurse practitioners and the MDT coordinator.

Secondary outcomes were rates of subsequent intervention (defunctioning stoma, major resection with stoma, stent, chemotherapy or radiotherapy). These were re-classified for the purposes of analysis as either endoscopic/surgical or non-surgical with a particular focus on the former in this study. Rate of subsequent intervention was assessed against both the original reason for not proceeding with curative surgery and by stage of disease. For simplicity, TNM stages where available were converted to American Joint Commission on Cancer (AJCC) Stages I–IV.

Statistical analysis

Data analyses were conducted using SPSS version 24 (SPSS Inc., Chicago, Illinois, USA). Kaplan–Meier survival analyses with log rank test (and Breslow test when required) were conducted to assess survival differences between groups. Comparison between groups regarding rate of intervention as well as the effect of disease stage were tested using Pearson's chi-squared test or Fisher's exact test (if appropriate). The significance level was taken as a P value < 0.05. Where significance was identified, where appropriate, a post hoc analysis and a Bonferroni adjustment were performed to account for multiple testing.

Ethical approval

Ethical approval was not considered necessary as it was a retrospective cohort study from a prospectively maintained cancer registry (NBOCA) looking at outcomes in established clinical practice. There were no identifiable patient data.

Demographics

Over the study period 976 patients were identified with colorectal cancer within our unit. Of these 184 (18.9%) patients were originally deemed unsuitable for curative surgery by the MDT. The male to female ratio was 1:0.9. Median age at diagnosis was 81 years [interquartile range (IQR) 71–87] with 104 (56.5%) patients aged 80 or over. Tumour location is shown for the non-curative cohort with data in Table 1. The distribution of cancers was as follows: 26.2% (48 cases) in the rectum, 16.4% (30 cases) in the sigmoid colon and 14.8% (27 cases) in the caecum. For comparison, Table 1 also displays the tumour distribution in patients offered curative resection. The spread of right-sided (tumours up to and including the splenic flexure) and left-sided tumours (descending colon to rectum) were comparatively similar across curative and non-curative groups (chi-squared, P = 0.860).

Reasons precluding curative surgery were initially identified as extent of metastatic disease, 85 cases (46.2%); frailty, 78 cases (42.4%); patient choice, 11 cases (6.0%); locally advanced disease, five cases (2.7%); and other reasons, four cases (2.2%). Only two patients (1.1%) received non-curative surgery initially. One patient had palliative surgery at presentation, and one was admitted as an emergency which resulted in diagnostic laparotomy leading to subsequent palliation. The remaining patients were managed with active monitoring, stent, chemotherapy, radiotherapy or palliative care.

Those with frailty or patient choice as the main issue were grouped as patient-related (PR) factors, 91 cases (49.7%), while those with metastatic or unresectable disease were classed as disease-related (DR), with 92 cases (50.3%). Those falling under “other” were re-classified accordingly where possible (one patient was subsequently excluded owing to a concomitant brain tumour that delayed decisions on management of the colorectal cancer), leaving a total final cohort of 183 patients. Tumour stage at presentation was available for 83 cases in the PR group (91%) and 77 in the DR group (84%), as shown in Table 2. Unfortunately only 61 cases (33%) had recorded World Health Organization (WHO) performance statuses: 39 cases (43%) of the PR group and 22 cases (24%) of the DR group. This is also shown in Table 2, broken down by stage. Retrospective performance status assessment was not possible owing to missing information from MDT records. Where records were kept, the majority of patients in the PR group were placed in the more severe categories (Stages 3–4) whilst a larger majority of cases in the DR group were 0–2.

In those managed non-operatively in the first instance, 22 cases (12%) were subsequently admitted for emergency surgery owing either to acute obstruction or uncontrollable diarrhoea (21 patients had a stoma, one stent).

Outcome in patients with incurable disease

A total of 159 patients (86.9%) died within the study period with a 30-day mortality of 12.6% (20 cases) and 1-year mortality of 62.3% (99 cases). Overall median survival from diagnosis to death was 205 days (IQR 60–532). Kaplan–Meier survival analysis for all patients at our colorectal unit who did not undergo curative surgical management is shown in Fig. 1.

Outcomes broken down by sub-group are detailed in Table 3. The median survival in the PR group was 277 days (IQR 70–593) vs 179 days (IQR 51–450) in the DR group (P < 0.02, Fig. 2). However, as the survival curves only start to deviate beyond 250 days, a Breslow test was performed to further characterize this. The Breslow test was not statistically significant (P = 0.11). This is emphasized by the 1-year survival analysis that demonstrated a weak association between the two groups which did not reach statistical significance (log rank test P = 0.08). One-year survival probabilities for the PR and DR groups were 0.407 and 0.283 respectively.

The relationship between tumour stage and survival was also assessed using Kaplan–Meier survival analysis, where staging was available (87.4%, 160 cases), and is illustrated in Fig. 3 (log rank test, P = 0.001). As expected, Stage 1 disease was associated with the longest survival while Stage 4 disease showed the worst outcomes. A comparison of Stage 2 and 3 disease, however, showed no statistical difference in outcomes (log rank test, P = 0.697) but this may be a consequence of incomplete data in these groups.

Effect of sub-group on rates of re-intervention

An analysis of rates of re-intervention (stent, defunctioning stoma, major surgery with stoma, chemotherapy and radiotherapy) demonstrated that in the majority of cases (153 cases, 83.6%) patients did not require a change from their original treatment modality (Fig. 4). Seven per cent of the cohort (13 cases) required a defunctioning stoma, 4.4% (eight cases) were readmitted for major surgery and stoma and 0.5% (one case) required a stent subsequently. In total, therefore, 12% (22 cases) of the cohort required some form of surgical or endoscopic re-intervention.

A comparison between the two groups demonstrated that patients within the DR group experienced a higher rate of surgical/endoscopic intervention vs those within the PR group, 17.4% (16 cases) and 6.6% (six cases) respectively, as shown in Table 4 (Fisher's exact test, P = 0.005). The median time to surgical/endoscopic intervention was 111 days in the DR group (range 25–621, n = 16) vs 125 days in the PR group (range 13–397, n = 6).

Effect of tumour stage on rate of re-intervention

Table 5 outlines the rate of re-intervention and type (surgical/endoscopic vs chemo/radiotherapy) across disease stages. There was no significant difference between the disease stage on rates of subsequent intervention (P = 0.392). Across those with Stages 2–4 disease rates ranged from 11.8% (11 out of 93) to 14.3% (two out of 14). However, it is worth noting that no Stage 1 tumour required subsequent re-intervention.

Comparison of outcomes in the all-ages cohort vs an elderly population

A separate analysis of PR vs DR outcomes was performed in patients aged 80 or over to compare outcomes in an elderly population. Overall 30-day and 1-year mortality in this cohort was 13.5% (14 cases) and 62.5% (65 cases) respectively, with a median overall survival of 178 days (IQR 52–529). Within groups, median survival was 334 days (IQR 74–602) for the PR group and 63 days (IQR 37–205) in the DR group (P = 0.001). Figure 5 demonstrates the Kaplan–Meier survival curve in this cohort.

Counselling and consent

Over half of patients within our cohort were aged over 80. In an ageing population, the incidence and frequency of comorbid conditions increases with age and, with it, the probability of being deemed unsuitable for curative surgery. Informed consent in modern practice requires that patients be given not just their best management option but all options, including the outcomes of conservative approaches. This underlines the importance of understanding the natural history of the disease in those whose fitness for surgery is in doubt. It allows for more informed discussions about the risks and benefits of operating vs non-intervention. This is especially relevant in the light of the 2015 Montgomery ruling in the UK Supreme Court which emphasizes the importance of fully informed consent 2.

The 2017 “Getting It Right First Time–General Surgery” document discussed an increasing prevalence of metastatic colorectal cancer patients not being offered surgical resection even in specific cases which current National Institute for Health and Care Excellence guidelines may suggest are potentially beneficial. The report observed that surgeons, whose performances are solely measured by mortality and national audit data (without taking into context the individual risk profile for a patient or their individual circumstances), might in some cases feel reservations about offering such patients high-risk procedures during counselling. However, as the Montgomery ruling stipulates, patients will need an understanding of the risks and benefits of not acting, including the risk of complications, symptoms and readmission for palliative surgery as an emergency. In such cases a patient may well wish elective surgery even with elevated risk if given the option and accurate risk assessment 2, 17.

Study limitations

This cohort study was a retrospective analysis from a prospectively maintained cancer registry. There was some heterogeneity in the quality of historical data recorded. Incomplete staging data were more prevalent in the DR than PR groups (15 vs 8 cases respectively). Furthermore, WHO performance status was only available for 61 out of 183 patients (33.3%), while incomplete MDT records precluded retrospective calculation for the remainder. This precluded a quantifiable analysis of frailty and introduces a risk of bias in comparison of groups. Additionally, we accept that larger studies may be required to confirm our findings.

The data presented in this single-centre study come from a semi-rural unit with a more advanced age demographic and inevitably some comparative differences in socio-economic circumstances with other units. This may limit the generalizability of our findings to other units.

A robust system was not in place for large scale data acquisition of cause-of-death data within our unit preventing evaluation of overall survival vs cancer-specific survival. Finally, routine cancer genetics is a recent addition within our unit and may influence decision making in the future. Future work in this group of patients needs to focus on quantifying frailty and performance status. Increasing life expectancy will continue to pose significant challenges to the surgeon looking after patients with colorectal malignancy.