Tumour growth rate of carcinoma of the colon and rectum: retrospective cohort study

Background The growth pattern of colorectal cancer is seldom investigated. This cohort study aimed to explore tumour growth rate in colorectal cancers managed non‐surgically or deemed not resectable, and to determine its implication for prognosis. Methods Consecutive patients with colonic or rectal adenocarcinoma were identified through the colorectal multidisciplinary team database at Leeds Teaching Hospitals NHS Trust over a 2‐year interval. Patients who received no treatment (surgery, stenting, colonic defunctioning procedures, chemotherapy, radiotherapy) and who underwent CT twice more than 5 weeks apart were included. Multidetector CT/three‐dimensional image analysis was performed independently by three experienced radiologists. Results Of 804 patients reviewed, 43 colorectal cancers were included in the final analysis. Median age at first CT was 80 (73–85) years and the median interval between scans was 150 (i.q.r. 72–471) days. An increase in T category was demonstrated in 31 of 43 tumours, with a median doubling time of 211 (112–404) days. The median percentage increase in tumour volume was 34·1 (13·3–53·9) per cent per 62 days. The all‐cause 3‐year mortality rate was 81 per cent (35 of 43) with a median survival time of 1·1 (0·4–2·2) years after the initial diagnostic scan. In those obstructed, the relative risk of death from subsequent perforation was 1·26 (95 per cent c.i. 1·07 to 1·49; P = 0·005). Conclusion This study documented a median doubling time of 211 days, with a concerning suggestion of tumour progression, which has implications for the current management standard.


Introduction
Colorectal cancer is the third most common cancer diagnosed in men and second most common in women internationally, with an increasing incidence in those aged less than 50 years 1,2 . Screening guidelines vary worldwide 3 ; paired faecal occult blood testing and endoscopic investigation have been shown to increase the detection rate of asymptomatic cancers. As a result, mortality has decreased by 30 per cent in participants, but with an increase in cancer numbers requiring operation and patients therefore waiting longer for surgery [4][5][6][7][8][9][10][11] . There is, however, a paucity of evidence surrounding the rate at which colorectal tumours grow once they are established.
Tumour growth and invasion is paramount to oncological outcomes as cancer advances through multiple distinct stages in its transition from indolent to invasive disease 12,13 . In the UK, guidelines to reduce the time that patients wait for cancer care and specific time standards from referral to first definitive treatment were introduced in 2009, and are now enshrined in the National Health Service (NHS) constitution. Although introduced with a laudable aim, the current 62-day standard does not have a scientific basis 14 .
To date, evaluation of colorectal cancer growth patterns is limited, predominantly because of the inherent ethical issues associated with a prospective longitudinal study of leaving diagnosed colorectal cancer untreated without clinical justification.
Current treatment strategies are determined by the stage of disease, associated co-morbidities, likely prognosis and patient preference. Staging for colorectal cancer is dependent on cross-sectional imaging, which is pivotal in the assessment of recurrence and metastatic disease in patients who have received treatment [26][27][28][29] . Use of imaging to help stratify patients is vitally important as neoadjuvant and adjuvant treatment options continue to advance. There is a growing demand for accurate quantification of tumour growth to optimize the timing of surgery or to assess the potential benefits of non-surgical therapy [30][31][32][33][34][35] . Precise quantification of tumour growth would have a significant impact for patients who choose non-surgical therapy in terms of both prognosis and informed consent. However, this also has medicolegal implications in terms of missed cancers, and in the consequences of a delay to standard investigations and surgical care.
The aim of this study was to measure tumour growth rates in a subgroup of untreated colorectal cancers, to provide prognostic value in patients with tumours managed non-surgically or deemed not resectable, and to determine the implications of delay to diagnosis.

Methods
Consecutive patients with colonic or rectal adenocarcinoma treated between 1 January 2016 and 31 December 2017 were identified through the institutional colorectal multidisciplinary team (MDT) database at Leeds Teaching Hospitals NHS Trust, and through a prospectively maintained radiology discrepancy and educational database at the same institution.
Patients were included if they had undergone CT twice at least 5 weeks apart (at least 1 within the tertiary referral centre) between April 2009 and September 2018 (any indication for repeat CT was considered), and during the same interval had received no tumour treatment: surgery (including stent insertion or colonic defunctioning procedures), chemotherapy, radiotherapy or any combination. Exclusion criteria were: patients with synchronous malignancies (those with altered tumour biology such as patients receiving systemic treatments); inability to identify the tumour or tumour margins (very small tumours or where artefact rendered CT image interpretation impossible); and non-adenocarcinoma subtype.

Data collection
Electronic clinical and radiological databases were used to obtain patient demographic details, clinical history, treatment data, clinical outcome and follow-up duration. Electronic records included the institutional radiology information system (Computerised Radiology Information System; Healthcare Software Systems, Mansfield, UK) and the oncology electronic patient record system (Patient Pathway Manager; EHR Development Team, Leeds Teaching Hospitals NHS Trust, Leeds, UK). Mortality was determined through hospital electronic records, which are paired with community and bereavement systems.
Prospective consent was obtained from all patients at the time of imaging for use of anonymized CT imaging data in research and service development projects. Formal ethics committee approval was waived for this study, which was considered by the institutional review board to represent evaluation of a routine clinical service.

Imaging acquisition and reconstruction
All patient examinations were re-examined retrospectively by three consultant radiologists. Multidetector CT (Siemens, Munich, Germany; GE Healthcare, Waukesha, Wisconsin, USA) was used with full abdominal and pelvic acquisition in a single breath hold. The scan was acquired at 120 kV, 80 mA, tube rotation time 0⋅5 s per rotation, and pitch 6. Images had a collimation and slice thickness of at least 5 mm (median 3 mm) but were often acquired at 1 mm slice thickness. All images were acquired using iterative reconstruction. Image analysis was performed on the thinnest slice thickness available. Where iodinated intravenous contrast material was administered, the image was acquired in the portal venous phase after administration of 100 ml contrast.

Image segmentation
Postprocessing and image analysis included three-dimensional lesion measurement using Advanced Workstation software (AW 3.2; GE Healthcare). The primary tumour was delineated using a semiautomated technique based on thresholding of the tumour using the lesion density (seeding around a set Hounsfield unit) within the tumour. This was then adjusted manually to outline the peripheral contours of the tumour on each image. This was performed in the axial plane with correlation on coronal and sagittal plane imaging. The tumour volume was calculated automatically by the software by multiplication of the cross-sectional area by the slice thickness. This process was repeated independently on the baseline and
% increase in tumour length per 62 days = (tumour length growth rate % of baseline size/(time follow-up -time baseline )) × 62.

Statistical analysis
All data were tabulated in Microsoft Excel ® (Microsoft, Redmond, Washington, USA). Continuous data are presented as median (i.q.r.). Subgroups were compared by means of Kruskal-Wallis ANOVA. All statistical analysis comparing tumour sizes, stages and growth rates was completed using SPSS ® version 23 (IBM, Armonk, New York, USA).

Results
During the study interval, 804 patients were referred to the colorectal MDT, of whom 67 met the inclusion criteria

Mortality
The all-cause 3-year mortality rate was 81 per cent (35 of 43), with a median life span of 1⋅1 (0⋅4-2⋅2) years after the initial diagnostic CT; the cause of death was a direct result of bowel obstruction and subsequent perforation in five patients. Eight patients were alive at the study conclusion, with a median follow-up time of 3⋅4 (2⋅1-6⋅1) years after the initial diagnostic CT. Seven patients had the initial scan during an unscheduled admission. Eighteen patients had the second CT as an emergency, with a median of 237 (93-462) days between scans. For patients who died compared with the sample as a whole, the relative risk of undergoing a second scan as an emergency within the study period was 1⋅34 (95 per cent c.i. 1⋅01 to 1⋅77; P = 0⋅040) and in those who were obstructed the relative risk of dying from bowel perforation was 1⋅26 (1⋅07 to 1⋅49; P = 0⋅005).

Discussion
Delays to cancer investigation and management are of concern both to patients and clinicians. The present results suggest that tumour volume can increase by a median of 34⋅1 per cent within the current NHS constitutional 62-day standard, which brings into question this standard. Furthermore, this standard could be compromised when there is a shortage of resources for delivery of elective surgical care in patients with colorectal cancer, such as during the current COVID pandemic 37 .
To date, seven studies [19][20][21][22][23][24][25] have determined tumour growth rate including a total of 177 colorectal cancers, with tumour doubling times ranging from 18 to 2593 days (Table S1, supporting information). These studies demonstrated no significant association between tumour growth rate and any change in tumour stage, nor its potential impact on treatment planning and prognosis. However, the majority of studies used barium enemas to assess tumour diameter and volume through measurement of filling defects, and were therefore limited in terms of accurate measurement of volume. Findings were also limited in terms of reproducibility owing to observed inaccuracies in tumour dimension measurements in altered views used during initial and follow-up investigations. These studies did, however, highlight that tumour growth rate may be dependent on primary tumour location and that tumour growth rate appears to be linear 20,22,23 . Colorectal cancer cell type heterogeneity, and differences in genetic mutations, epigenetic regulation and the microenvironment in which tumours reside mean that predicting tumour behaviour, independent of sample size, is difficult 38 . Considering tumour growth specifically, tumour hypoxia 39 , expression of growth factors 40,41 and necrosis 42 may be limiting factors.
An association between progression in TNM stage and tumour growth rate was demonstrated here, as expected. This study also evaluated the relationship between tumour growth rate and location (which determines non-luminal diameter), and the findings have implications for symptom onset, time to obstruction and therefore planning of operative management. The results suggest that the more distal the tumour, the greater the median volume growth per 62 days, with a greater median absolute growth per 62 days observed in tumours that progressed from T3 to T4 compared with other stage changes. A median tumour doubling time of 211 days for the whole cohort is similar to that in the only other comparable study using CT 20 .
Eighteen of 43 patients in this cohort underwent the second CT as an emergency after the decision had been made to proceed with a non-surgical strategy. Accepting the small sample size in this study, an all-cause mortality rate of 81 per cent around 1 year after diagnosis is a worrying finding in this underinvestigated group. These patients are at significantly increased risk of bowel obstruction with subsequent perforation 43 . This has significant implications for the communication of prognosis with the patient, consent discussion when considering management strategy and emergency presentations 44 .
Mucinous adenocarcinomas have distinct genetic and clinicopathological features compared with non-mucinous tumours. These tumours have previously been shown to be more frequently located in the proximal colon 45 , but there are no current data on rate of progression compared with non-mucinous tumours. The ten mucinous tumours in the present cohort (6 right-sided and 4 left-sided) showed double the growth rate of non-mucinous tumours over 62 days. If tumour growth rate could be correlated further with specific anatomical location, stage and histology, it would be possible to stratify patients in terms of risk, further prioritize management and provide a more informed consent process.
The main limitations of this study are the inherent selection bias of the study population and small sample size. The study is biased towards the inclusion of older subjects who may have more indolent tumours than a younger cohort 46 . Patient socioeconomic status and race may also affect tumour location and behaviour 46 . Inferring tumour growth rates from the observation of tumour volumes at two time points has been documented previously, but there is currently no consensus regarding the growth patterns exhibited by solid tumours or how they can be measured. The calculations applied here are based on the assumption that colorectal tumours follow a linear growth pattern, a characteristic that has yet to be established but is in line with current limited evidence (Table S1, supporting information) 47,48 .
The inclusion and exclusion criteria potentially biased selection towards slow-growing tumours in frail patients who underwent no treatment following diagnosis. This is because tumours that are large and obstructing may be subject to prompt treatment or result in morbidity. In addition, patients with smaller slow-growing tumours that are causing minimal obstructive symptoms may be more likely to select a non-operative treatment plan. This bias is difficult to avoid owing to the ethical considerations associated with a prospective, observational cancer growth study.
Tumour growth rate remains an important underevaluated variable in colorectal cancer, and is essential for screening, choice of management (operative versus non operative) and prognosis. To date, growth rate and doubling time have shown no true correlation with cancer stage or location, with limited studies pairing these parameters with histological analysis. The present results highlight particular concern regarding mucinous tumours, lesions that arise in the right colon, and tumours diagnosed as T3 that progress.