Outcomes of patients with non-lower gastrointestinal cancer diagnosed via the lower gastrointestinal two-week-wait service

Abstract

Introduction

Patients referred via lower gastrointestinal two-week-wait (LGI-2WW) services deemed at ‘low risk’ of LGI cancer may have delayed or no investigation. However, 20% of patients diagnosed with cancer via the LGI-2WW have non-LGI cancer. This study investigates the outcomes in this under-reported group.

Methods

A retrospective analysis of patients referred to a UK LGI-2WW service with a non-LGI cancer diagnosis (International Classification of Diseases 10th revision) between 1 January 2015 and 31 December 2018. The primary outcome was all-cause mortality. Statistical analysis was undertaken using Kaplan–Meier survival curves with generalised log-rank test and binomial logistic regression of pre-investigation information to predict death. A p-value of < 0.05 represented statistical significance.

Results

In total, 140 patients were diagnosed with non-LGI cancer (of 10,155 patients referred via the LGI-2WW pathway). Median follow-up was 13 months (range 0–65 months). Thirty-five patients (35/140; 25%) did not meet LGI-2WW referral criteria. Median survival varied by cancer site (upper gastrointestinal: 43/140 patients, median survival 4 months; gynaecological: 25/140, 25 months; haematological: 22/140, < 50% died; urological: 21/140, 47 months; unknown primary: 12/140, 3 months; lung: 10/140, 17 months; breast: 3/140, 5 months; retroperitoneal: 2/140, 9 months; brain: 1/140, 4 months; adrenal: 1/140, 52 months). Binomial logistic regression was statistically significant (chi-squared goodness-of-fit test = 12.334; p = 0.03); but the presence of comorbidity was the only significant predictive factor of death (p = 0.03).

Conclusions

Patients with non-LGI cancers cannot be easily predicted and have a poor prognosis. Pathways to streamline investigations for LGI cancer should include computed tomography scans for patients at ‘low risk’ of LGI cancer to ensure that non-LGI cancers are diagnosed.

Introduction

Colorectal cancer (CRC) is a leading cause of death in the UK and an emphasis of the National Health Service (NHS) long-term plan is to prioritise early cancer diagnosis.¹ The two-week-wait (2WW) service was established to ensure that patients with symptoms indicative of malignancy were referred rapidly by their general practitioner (GP) to a specialist service for investigation. 2WW referral criteria recommend that non-specific symptoms should be referred to the most likely cancer service.² Therefore, the broadening of symptom criteria and the reduction in the positive predictive value of indicative symptoms requiring referral to 3% in 2015 has resulted in an increase in referrals via the 2WW pathway, but without a corresponding increase in CRC detection,³ or survival.⁴

The gold-standard investigation for CRC is colonoscopy,⁵ which is invasive, expensive and has rare but significant complications. The increase in referrals to lower gastrointestinal (LGI) 2WW services has led to intense service pressures on diagnostic services. Currently, over 50% of patients are waiting more than 6 weeks for endoscopic procedures,⁶ delaying cancer diagnosis for those patients with CRC.⁷ To decrease these waiting times, straight-to-test pathways have been instituted; however, they are commonly endoscopic investigations.^8–10 A concern with efforts to streamline the LGI-2WW pathway is that those deemed at ‘low risk’ of LGI cancer may be discharged without any investigation.^11,12 Computed tomography (CT) scans have a shorter waiting time, suggesting less service pressure,⁶ and may be a more appropriate method of investigating patients with a lower chance of having CRC (due to approximately 80% accuracy in detecting early CRC¹³), and presenting with non-LGI-specific symptoms.^12,14

The authors reported previously that 20% of patients with cancer diagnosed via the LGI-2WW service had a non-LGI cancer.¹⁵ This suggests that either symptom referral criteria are not specific to cancer site location or commonly experienced LGI symptoms are incidental and unrelated to the cancers that are subsequently diagnosed. It also highlights that endoscopic examination alone does not exclude non-LGI cancers, which are prevalent in this population. These patients are often excluded from analysis when looking at LGI-2WW pathways;^4,9,16 however, their outcomes are equally important.

This study investigates the outcomes of patients referred to an LGI-2WW service who were subsequently diagnosed with a non-LGI cancer to better understand how the pathway could be improved to optimise investigation and diagnosis for those with LGI and non-LGI cancers.

Methods

All patients referred from primary care to the secondary care provider (Royal Devon and Exeter Hospital, UK) on the LGI-2WW pathway between 1 January 2015 and 31 December 2018 were identified from a prospectively gathered database of cancer referrals.

Hospital databases and clinical information systems were searched and mortality data collected via NHS Spine in July 2020. The study was approved by the local audit and service provision department (approval number 19-4481). During the study period, referrals from primary care were triaged by a colorectal consultant; patients with LGI-specific symptoms and no contraindications were booked directly to colonoscopy. All other patients were reviewed in clinic by a medical professional for assessment and onwards investigation.

Inclusion criteria were patients aged over 18 years, with a recorded International Classification of Diseases 10th revision (ICD-10) cancer diagnosis. Patients with a cancer diagnosis code of LGI malignancy (anal and CRC; ICD 18–21) were excluded.

Primary outcome was all-cause mortality; censored data was the time interval (in months) between date of referral and date of death or July 2020 (where data are censored). Secondary outcomes of treatment intention (either curative or palliative) were recorded as per multidisciplinary team or clinical documentation. Audit of symptoms against National Institute of Health and Care Excellence (NICE) LGI-2WW referral criteria were as per the 2015 referral criteria for colorectal and anal cancers.² The reporting of this study conforms to the Strengthening The Reporting of OBservational Studies in Epidemiology (STROBE) guidelines (Appendix 1 – available online).

Statistical analysis

Statistical analysis was performed with SPSS v.26 software (IBM). Continuous data are given as median (range), and demographic data as percentages. Survival time analysis was performed by Kaplan–Meier survival curves and compared with generalised log-rank test, with data given as median survival time and 95% upper and lower confidence intervals. Cancer sites with ten or more patients were plotted on a graph and subjected to significance testing, because smaller patient groups did not allow meaningful analysis.

Binomial logistic regression was performed of pre-investigation information (meeting LGI-2WW criteria guidelines, primary presenting symptom, primary comorbidity, gender and age at referral) to see if this could predict the primary outcome of death. Statistical significance was assessed by chi-squared goodness-of-fit test, and variance described by Nagelkerke R-squared, with the cut-off value set to 0.5. Statistical significance was set as p < 0.05.

Results

Patient demographics

In total, 10,155 patients were referred to the LGI-2WW service during the study period; 140 of whom were subsequently diagnosed with a non-LGI cancer and were included in the study. Follow-up (either to the study end point or death) was a median of 13 months (range 0–65 months).

The baseline demographic details of included patients are listed in Table 1. Median patient age was 77, 58% of patients were male and 76% had a pre-existing comorbidity at referral.

Table 1 .

Patient demographics

Patient characteristic	N = 140	Percentage
Age, median (range)	77 (35–96)
Gender, M:F	81:59	58:42
Comorbidity
Cardiac	48	34
Respiratory	11	8
Renal	3	2
Diabetes	16	11
Neurological	6	4
Bowel disease	4	3
Previous cancer	19	13
Any	107	76

Symptom analysis

The primary and, if present, secondary presenting symptom prompting referral to the LGI-2WW service are presented in Table 2, which demonstrates that the most common primary presenting symptoms were change in bowel habit (CIBH; 35%) and abdominal pain (24%). These presenting symptoms, along with the patient’s age were audited against the NICE LGI-2WW to see if they met the referral criteria. Table 3 demonstrates that 25% of patients did not meet the referral criteria. The primary investigations and, if performed, subsequent (secondary) investigations are given in Table 4, which demonstrates that 85% of patients were investigated primarily by CT scan.

Table 2 .

Primary symptom and secondary symptoms prompting referral from general practitioner to lower gastrointestinal two-week-wait service

	No. of patients (%) N = 140	Secondary symptom
		CIBH	Bleeding	Abdominal pain	Abdominal mass	Weight loss	Tenesmus	Non-specific
Primary symptom
CIBH	49 (35)	0	3	8	1	10	2	3
Bleeding	7 (5)	2	0	1	0	1	0	0
Abdominal pain	34 (24)	12	3	0	2	6	0	3
Abdominal mass	8 (6)	1	0	1	0	0	0	0
Weight loss	29 (21)	6	1	4	1	0	0	10
Non-specific	13 (9)	1	0	0	0	2	0	0

CIBH = change in bowel habit

Table 3 .

Outcome of audit of symptoms prompting referral to the lower gastrointestinal two-week-wait criteria

Symptom	Did not meet criteria	Met criteria
CIBH	0	49
Bleeding	0	7
Abdominal pain	9	25
Abdominal mass	0	8
Weight loss	16	13
Non-specific	10	3
Total	35 (25%)	105 (75%)

CIBH = change in bowel habit

Table 4 .

Primary and secondary investigations performed

	Number (%)	Second investigation
		Any second investigation	CT	US	OGD	Colonoscopy	Invasive biopsy	Staging scans
First investigation
CT	120 (85)	72	0	5	10	11	35	11
US	4 (3)	3	3	0	0	0	0	0
OGD	9 (6.4)	9	6	0	0	2	1	0
Colonoscopy	6 (4.3)	5	4	0	0	0	1	0
Invasive biopsy	1 (0.7)	1	1	0	0	0	0	0

CT = computed tomography; OGC = oesophagogastroscopy; US = ultrasound

Diagnosis and outcome

Cancer site diagnosis, treatment intent outcome and median survival in months are shown in Figure 1 and Table 5. Upper gastrointestinal (UGI) malignancy was the most diagnosed non-LGI cancer accounting for 31% of patients (43/140). Some 72% (101/140) of patients were treated with palliative intent, and 75% (103/140) died within the study period with generally poor median survivals that varied according to cancer site diagnosis (median survival was 13 months overall; range 4–52 months).

Figure 1 .

Kaplan–Meier survival graph according to cancer site. Cumulative survival of patients according to cancer site from Ref2event (time of referral to event in months; either end of study or censored = death). Cancer sites with ten or more patients are plotted only. Upper gastrointestinal (UGI) vs urological, haematological, gynaecological and lung, statistically significant (p < 0.001). UGI vs unknown primary, non-significant (p = 0.976). Statistical analysis log-rank test, p < 0.05.

Table 5 .

Outcome according to cancer site : treatment intent, mortality and median suvival

Cancer site	Treatment intent		Total	Percentage	Number died	Median survival (months)	CI
	Curative	Palliative
UGI	4	39	43	31	39	4	(2.2–5.8)
Urological	11	10	21	15	11	47	(19–75)
Haematological	16	6	22	16	9	–
Gynaecological	3	22	25	18	20	25	(18–31.5)
Unknown Primary	1	11	12	8	11	3	(1.3–4.7)
Lung	2	8	10	7	7	17	(2.7–31.3)
Retroperitoneal	1	1	2	1	1	9	n/a
Breast	0	3	3	2	3	5	(1.6–8.2)
Brain	0	1	1	0.7	1	4	n/a
Adrenal	1	0	1	0.7	1	52	n/a

CI = confidence intervals; n/a = not applicable; UGI = upper gastrointestinal

Prediction of outcome from symptoms

Binomial logistic regression of pre-investigation factors to predict the primary outcome of death was performed. It was statistically significant (chi-squared goodness-of-fit test 12.334, p = 0.030), with presence of comorbidity being the only factor that reached statistical significance in the prediction model (p = 0.030); no other factor reached statistical significance (adherence to LGI-2WW criteria, primary symptom, gender, age at referral; p > 0.05). The model explained 16.8% of the variance (Nagelkerke R-squared test) and achieved an overall accuracy of 76.6%, with a sensitivity of 95%, but poor specificity of 9% (Table 6).

Table 6 .

Classification of binomial logistic regression prediction model using pre-investigation factors to predict outcome of death

		Predicted
		Death	Survival
Observed	Death	80	4
	Survival	21	2

Cut-off value = 0.5

Discussion

This study describes the outcomes of patients diagnosed with non-LGI cancer via the LGI-2WW service. These patients generally have a poor outcome, with 72% being treated with palliative intent, and a poor median survival time of 13 months. LGI-2WW referral guidelines and configuration of services need to recognise that a significant number of poor prognosis non-LGI cancers are diagnosed via this pathway.

An audit of symptoms demonstrated that 25% of patients did not meet the criteria for being referred on the LGI-2WW pathway. It is likely that although these patients did not meet the referral criteria, the GP and triaging clinician had sufficient concern of a cancer that investigations were performed. The large proportion of patients not meeting the referral criteria contributes to the extreme service pressure, in keeping with previous studies.³ Conversely, it is interesting that the majority (75%) of patients who had a non-LGI cancer met the LGI referral criteria. This suggests that either some referral symptom criteria are not cancer site specific (eg, weight loss), bowel symptoms are not specific to the LGI tract (eg, CIBH in pancreatic cancer¹⁷) or are so prevalent within the population that LGI symptoms are incidental to the subsequently diagnosed cancers.¹⁸

The majority of patients in this study were investigated by CT scan due to the high prevalence of non-LGI-specific symptoms (eg, weight loss). The survival analysis demonstrates that many of these patients have a poor outcome with a survival time from diagnosis of as little as 3–4 months. The potential long wait of over 6 weeks for an endoscopy investigation is highly significant and if patients had endoscopy only, many of the non-LGI cancers may not have been diagnosed. This suggests that patients with non-LGI-specific symptoms should be investigated with a CT scan to investigate intra- and extraluminal pathology, and enable prioritisation of endoscopy for those with LGI-specific symptoms, similar to previously described pathways.^11,14

Most patients with non-LGI cancers were treated with palliative intent and had a poor prognosis. This population of patients referred via the LGI-2WW service have much worse outcomes than the population with LGI cancer described previously.¹⁵ The breakdown by cancer site demonstrates that some sites were associated with reasonable survival outcomes, for example haematological and urological cancers; however, the majority of patients had poor survival of less than a year. The logistic regression model was based on the pre-investigation data that would be available to a GP or secondary care practitioner; however, only pre-existing comorbidity was a significant predictor of death, which is to be expected. Neither primary symptom nor adherence to the LGI-2WW criteria were prognostic indicators, making it difficult to predict these poor prognosis cancers.

Since the end of this study period, NICE have recommended the use of quantitative faecal immunochemical testing (qFIT) to guide GP referrals for patients.² A number of studies investigating the use of faecal immunochemical testing (FIT) in primary care have eloquently demonstrated the efficacy of qFIT with a sensitivity of 97% for excluding CRC with a threshold of 2μg/g for patients referred via the 2WW pathway (similar sensitivity to colonoscopy).^19–21 A concern would be that those with a negative FIT result may not reach the threshold for investigation and have a non-LGI cancer. The findings of our study suggest streamlining of referral LGI pathways must consider non-site-specific cancer pathways to ensure non-LGI cancers remain diagnosed. The authors would recommend the use of qFIT as a triaging tool in primary care; qFIT-negative patients with concerning symptoms can be referred to the non-specific cancer site pathway and should have cross-sectional imaging with CT. Patients reaching symptom referral criteria and a positive qFIT result should go straight-to-test colonoscopy. This would not lead to an increase in the numbers of investigations, but a prioritisation of resources based on the likely cancer site.

Study limitations

There are limitations to this study: it is retrospective and data collection is limited by what information was collected at the time. The sample size is modest (140); however, this is a subset of patients from a large number of LGI-2WW referrals of more than 10,000 patients collected over a period of 4 years.

Another limitation is not having symptom information for all of the more than 10,000 patients referred. This denominator data would allow analysis of which symptoms were associated with different cancer types or a non-cancer diagnosis. However, this requires prospective gathering of information and is the subject of future work.

Conclusions

This study describes a previously under-reported cohort of patients referred to the LGI-2WW clinic and diagnosed with non-LGI cancers, and is the first to describe their outcomes. These patients frequently have cancers treated with palliative intent, with a very poor survival outcome. We propose that patients referred on the LGI-2WW pathway considered to be at low risk of LGI malignancy (eg, with a negative FIT result) be investigated with CT scan to ensure non-LGI cancers remain diagnosed.

Data availability statement

Data available on request from authors.