Determinants of emergency presentation in patients with colorectal cancer: a systematic review and meta-analysis

Colorectal cancer remains a significant cause of morbidity and mortality, even despite curative treatment. A significant proportion of patients present emergently and have poorer outcomes compared to elective presentations, independent of TNM stage. In this systematic review and meta-analysis, differences between elective/emergency presentations of colorectal cancer were examined to determine which factors were associated with emergency presentation. A literature search was carried out from 1990 to 2018 comparing elective and emergency presentations of colon and/or rectal cancer. All reported clinicopathological variables were extracted from identified studies. Variables were analysed through either systematic review or, if appropriate, meta-analysis. This study identified multiple differences between elective and emergency presentations of colorectal cancer. On meta-analysis, emergency presentations were associated with more advanced tumour stage, both overall (OR 2.05) and T/N/M/ subclassification (OR 2.56/1.59/1.75), more: lymphovascular invasion (OR 1.76), vascular invasion (OR 1.92), perineural invasion (OR 1.89), and ASA (OR 1.83). Emergencies were more likely to be of ethnic minority (OR 1.58). There are multiple tumour/host factors that differ between elective and emergency presentations of colorectal cancer. Further work is required to determine which of these factors are independently associated with emergency presentation and subsequently which factors have the most significant effect on outcomes.


Scientific Reports
| (2022) 12:4366 | https://doi.org/10.1038/s41598-022-08447-y www.nature.com/scientificreports/ with mode of presentation may yet to be studied. To the best of our knowledge, to date, the existing literature comparing mode of presentation and clinicopathological factors has yet to be comprehensively summarised. The present systematic review and meta-analysis aims to comprehensively review thirty years of literature analysing the association between clinicopathological factors and mode of presentation of colorectal cancer to identify those factors that differ between elective and emergency presentations of colorectal cancer.

Methods
This systematic review and meta-analysis of published literature was carried out according to a pre-defined protocol. The primary outcome was to compare the differences between tumour factors and host factors and mode of presentation of colorectal cancer.
Studies published between January 1990 and August 2018 were identified through an electronic search of the US National Library of Medicine (MEDLINE) and the Cochrane Database of Systematic Reviews. Selected other studies were identified through a manual bibliography search. The following search strategy was used: (colon OR rectum OR rectal OR colorectal) AND (cancer OR carcinoma OR adenocarcinoma OR neoplasm OR malign OR tumour) AND (emergency OR acute OR urgent OR non-elective) AND (surgery OR surgical OR operation OR resection OR procedure).
On completion of the online search, the title and abstract of each identified study was examined for relevance with full text being obtained for all potentially relevant studies. This was undertaken by an individual researcher with discussion with a senior author if required. Studies were included regardless of design, with both trials and observational studies being eligible for inclusion. Studies that were not in English, studies where the full text was not available, studies that included patients undergoing colorectal resection for pathology other than cancer or patients undergoing colonic stenting were excluded. The present study involved a wide literature search to capture as much of the pre-existing literature as possible however small studies (deemed those with less than 50 patients within the emergency group) were excluded to reduce the risk of bias. In those instances where multiple studies were available using the same patient population only the most recent study was included. If populations varied the most inclusive study was used. Those studies that did not provide comparison between elective and emergency patients were excluded from this review. This is shown in our PRISMA flow diagram (Fig. 1). Provided there were 3 or more studies for a particular factor, a meta-analysis of tumour/host factors was performed. Papers included either reported the numbers of emergency and elective patients and the number of patients with the factor of interest analysed or reported percentages in a way that allowed these numbers to be calculated. The Cochrane Handbook for Systematic Reviews 12 has been used to guide the reporting of results within the present study.
Statistical analysis. All statistical analyses were performed using Review Manager (RevMan) Version 5.3, The Cochrane Collaboration. For all comparisons an unadjusted odds ratio was used. Where possible, total sample sizes and events were taken from the raw data presented in each study. If events were reported as a percentage of total sample size, the event size was calculated from this percentage. 95% confidence intervals were used throughout and a p value of < 0.05 was considered to be significant. Forest plots were used for graphical display

Results
Literature search. Studies were selected as demonstrated in the PRISMA diagram ( Fig. 1). The initial search strategy identified 7,609 studies whose titles and abstracts were reviewed. Studies were excluded that were published prior to 1990 (n = 600), not in English (n = 1,035), primarily compared colonic stenting (n = 141), did not have an available full paper (n = 648) or were either not relevant to this topic or included pathologies other than colorectal cancer (n = 5,034). This led to the review of 151 full papers. Of these a further 97 were excluded as they included less than 50 patients (n = 23), did not provide a comparison between elective and emergency patients (n = 30), included pathologies other than colorectal cancer (n = 13), were articles (n = 1), duplicate studies (n = 4) or were not relevant (n = 26). The remaining 54 studies were included in this review.

Tumour factors.
Tumour location. 20 studies examined the association between tumour location and mode of presentation in 97,788 patients (Supplementary Table S1). Within this review, tumours of the right colon, hepatic flexure and transverse colon were considered right sided. Tumours of the splenic flexure, left colon and sigmoid colon have been considered left sided. Rectosigmoid and rectal tumours have been considered rectal. 11 studies 7,13-23 examined the association between colonic/rectal location and mode of presentation in 62,867 patients. On meta-analysis including all of these studies (Fig. 2) there was an association between emergency presentation and colonic location (OR 2.45, 95% CI 2.33-2.57, P < 0.001, I 2 = 94%). 19 studies 7,13-19,21-31 examined the association between colonic location (left/right) and mode of presentation in 95,911 patients. On meta-analysis including 15 studies of 61,738 patients ( Fig. 3) no significant association was reported between emergency presentation and colonic location (OR 0.98, 95% CI 0.94-1.01, P = 0.22, I 2 = 77%).
Tumour size. 1 study 15 examined the association between tumour size and mode of presentation in 1,672 patients (Supplementary Table S2) and reported an association between emergency presentation and larger tumour diameter (p = 0.011).
Metastatic disease (M stage). 7 studies 15,19,22,24,25,35,43 examined the association between M Stage and mode of presentation in 8,703 patients (Supplementary Table S6). On meta-analysis including all of these studies (Fig. 7)   Table S7) and reported an association between emergency presentation and tumour circumference of greater than two thirds of the luminal circumference (p = 0.009).
Tumour type. 4 studies 13,15,18,45 examined the association between tumour type and mode of presentation in 84,791 patients (Supplementary Table S8). One study 45 of 81,825 patients found an inverse association between emergency presentation and simple adenocarcinomas (83% vs 85%) and an association between emergency     Table S10). On meta-analysis including all of these studies ( Fig. 9) there was an association between emergency presentation and vascular invasion (OR 1.92, 95% CI 1.62-2.27, P < 0.001, I 2 = 70%).
Tumour perforation. 1 study 36 examined the association between tumour perforation and the mode of presentation in 707 patients (Supplementary Table S11) and reported an association between emergency presentation and microscopic perforation (P = 0.010).
Perineural invasion. 3 studies 13,30,43 examined the association between perineural invasion and mode of presentation in 3210 patients (Supplementary Table S12). On meta-analysis including all of these studies (Fig. 10) there was an association between emergency presentation and perineural invasion (OR      Table S17). On meta-analysis that included all of these studies (Fig. 12) there was an association between emergency presentation and female sex (OR 1.08, 95% CI 1.07-1.09, P < 0.001, I 2 = 98%).    Table S18). Due to heterogeneity of data it was not possible to perform a meta-analysis of this factor. 11 studies of 514,205 patients did not find a significant association between emergency presentation and age. This included a large study 48 from the USA of 507,750 patients that compared the proportion of patients aged over 65 who presented either electively or as an emergency. 18 studies of 394,926 patients found an association between emergency presentation and older age. This included a study 51 Table S19). Three of these studies were from the USA and one was from the UK. Two studies compared white vs African-American individuals, one study classified patients as either White, Black or Asian and the final study classified patients as ethnic minority (yes/no) however did not provide further description of ethnic minority status. On meta-analysis including all of these studies (Fig. 13) there was an association between emergency presentation and ethnic minority status (OR 1.58, 95% CI 1.51-1.65, I 2 = 81%). 33,43,54 examined the association between Body Mass Index (BMI) and mode of presentation in 1,700 patients (Supplementary Table S20). Two studies 43,54 of 1071 patients reported no significant association between emergency presentation and median BMI. One study 33 of 455 patients reported an association between a BMI < 25 or > 40 and emergency presentation (P = 0.001).

Body mass index. 3 studies
Distance to hospital. 1 study 55 examined the association between distance to hospital and mode of presentation in 380 patients (Supplementary Table S21)-no significant association was found.
Comorbid status. ASA grade. 3 studies 29,39,42 examined the association between ASA grade and mode of presentation in 31,359 patients (Supplementary Table S23). On meta-analysis including all of these studies (Fig. 14) there was an association between emergency presentation and ASA ≥ 3 (OR 1.83, 95% CI 1.72-1.94, P < 0.001, I 2 = 48%).     29,48 reported an association between emergency presentation and less co-morbid status. This included a study 48 of 508,032 patients that reported a Charlson Score ≥ 2 in 8.6% of emergency patients and 9.2% of elective patients (p ≤ 0.001). A further study 29 of 30,907 patients reported a Charlson score of ≥ 2 in 24% of emergency patients and 26% of elective patients (level of statistical significance not provided). 7 studies 5,14-16,18,59,60 of 183,286 patients reported an association between emergency presentation and more co-morbid status.
Pre-operative systemic inflammatory response. 2 studies 39,61 examined the association between pre-operative systemic inflammatory response and mode of presentation in 1246 patients (Supplementary Table S25). 1 study reported both the modified Glasgow Prognostic Score 62 (mGPS) and Neutrophil-Lymphocyte ratio 63 (NLR) and 1 study reported preoperative C-reactive protein (CRP). Both studies reported an association between emergency presentation and the preoperative systemic inflammatory response. Seasonal variability. 1 study 25 examined the association between seasonal variability and mode of presentation (Supplementary Table S26) and reported an association between emergency presentation and presentation during the summer months (June-August) in comparison to the winter months (December-February)-36% vs 23% P = 0.05.
Other factors. 1 study 64 examined the association between haemoglobin and weight loss and mode of presentation in 372 patients (Supplementary Table S27). Low haemoglobin levels and weight loss were both associated with emergency presentation (both P ≤ 0.001).
1 study 39 examined the association between CEA, TNF A, IL1 and IL6 and mode of presentation in 106 patients (Supplementary Table S28) and reported a significantly higher CEA, IL1 and IL6 in the emergency cohort. No significant difference was reported in TNF A levels.

Discussion
The present systematic review and meta-analysis confirms multiple differences in tumour, host and other factors between elective and emergency presentations of colorectal cancer. It may therefore be a combination of these factors that are associated with the poorer short-and long-term outcomes reported in emergency presentations of colorectal cancer 10,11 rather than emergency presentation per se.
In particular, tumour location (colon vs rectum), tumour stage, lymphovascular/perineural invasion, tumour differentiation, ethnicity and ASA grade differed significantly on meta-analysis between the elective and emergency cohorts as summarised in Fig. 15. Although not analysed in the meta-analysis due to study heterogeneity/< 3 studies other factors that differed between elective and emergency presentations include age, socioeconomic status and the preoperative systemic inflammatory response. Many of these factors have been reported to be associated with oncological outcomes in colorectal cancer 38,[65][66][67][68] and it therefore cannot be assumed that the negative effect of emergency presentation is solely due to more advanced disease. More recently, factors including body composition 69 and perioperative blood transfusion 70 have been reported to be associated with poorer www.nature.com/scientificreports/ long-term outcomes following curative resection for colorectal cancer and would be of interest for inclusion in future studies comparing elective and emergency presentations. The present review found that, on meta-analysis, ethnic minority status was associated with emergency presentation. However, given that the included studies were either from the USA or UK, non-Caucasian was essentially considered the ethnic minority group. No studies compared the effect of ethnic minority status in a country where Caucasian was the minority group and this would be an interesting area of future research. Emergency presentations of colorectal cancer remain associated with poorer long-term outcomes than elective presentations, even after adjustment for TNM stage. Indeed, within TNM Stage II colorectal cancer, emergency presentation is considered to be a high-risk factor requiring consideration for adjuvant chemotherapy [71][72][73] . Further research would allow for both adjusted analysis of factors associated with emergency presentation and the subsequent effect of these on long-term outcomes both within the overall patient population and within stage-specific disease.
Over the last two decades, colorectal cancer screening programs have become widespread throughout the developed world. While participation in screening programs has resulted in a significant reduction in the proportion of patients presenting emergently 74 many patients continue to present with acute symptoms requiring emergency investigation and treatment. The present review included literature from both a screening and prescreening era. It has been shown that factors including age, sex, socioeconomic status and tumour stage and site 75 differ between unscreened patents and those patients who have either participated in or been diagnosed through screening. No studies have been identified to date comparing emergency presentations between those patients who did/did not participate in screening and this would be of interest in future work.
The present study has several limitations. Due to the nature of this study, a significant degree of heterogeneity was present both in terms of inclusion criteria and reported outcomes within individual studies. Therefore, it was not possible to compare adjusted data hence the use of unadjusted data within the present review. Factors within the present study including age and BMI have not been included within meta-analysis due to data heterogeneity and the continuous nature of these variables. Consideration was given to conducting meta-regression however in keeping with guidance 12 this could not be carried out due to the small number of studies suitable for such analysis. While the present review identified a large number of studies comparing elective and emergency presentations of colorectal cancer, very few studies subclassify emergency presentations into their presenting diagnoses, predominantly obstruction, perforation and bleeding. It therefore remains uncertain how factors and outcomes vary between different emergency presentations. One would hypothesise that patients presenting with perforation may have significantly different characteristics and outcomes than those presenting with an otherwise uncomplicated large bowel obstruction. The optimal management of patients presenting as an emergency with large bowel obstruction remains uncertain. While the majority of patients undergo emergency colonic resection, some clinicians opt for primary colonic stenting in the emergency setting with subsequent elective resectional surgery. This in an important question which remains unanswered however lies outside the scope of the present review [76][77][78] . It is commonplace within Systematic Reviews and Meta-Analyses to present risk of bias and quality of included studies using a variety of measures 12 . However the nature of the present review does not analyse the effect of an intervention on outcomes and therefore such measures are not applicable to the present review. Furthermore, with reference to specific factors, the small number of studies precluded meaningful analysis of the overall quality of studies and risk of bias.
In summary, the present study has identified multiple factors that differ between elective and emergency presentations of colorectal cancer as reported within the past 30 years of literature. This literature review paves the way to determining which tumour and host factors are independently significant with mode of presentation and which have the most significant effects on short-and long-term outcomes therefore explaining the poorer outcomes reported within emergency presentations. Defining these factors would help to determine those patients that have the worst short-term and long-term outcomes and therefore identify strategies within the perioperative and adjuvant settings to improve outcomes for these high-risk patients.