Patients from general practice with non-specific cancer symptoms: a retrospective study of symptoms and imaging

Background Patients with non-specific symptoms or signs of cancer (NSSC) present a challenge as they are a heterogeneous population who are not candidates for fast-track work-up in an organ-specific cancer pre-planned pathway (CPP). Denmark has a cancer pre-planned pathway for this population (NSSC-CPP), but several issues remain unclarified, for example, distribution and significance of symptoms and findings, and choice of imaging. Aim To investigate symptoms, cancer diagnoses, and diagnostic yield of computed tomography (CT) and fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG-PET/CT) in patients on NSSC-CPP to improve the overall diagnostic process. Design & setting A retrospective medical chart review in a 1-year consecutive cohort (2020). Method A total of 802 referrals were reviewed for diagnostic imaging in patients with NSSP from general practices, specialist practices, or the local hospital diagnostic centre responsible for NSSC-CPP. Results The study included 248 patients; 21% had cancer, most frequently gastrointestinal cancer (27%). The most frequent symptom was weight loss (56%). CT had a sensitivity of 85%, specificity of 87%, positive predictive value (PPV) of 65%, and negative predictive value (NPV) of 96%. For 18F-FDG-PET/CT, the numbers were sensitivity 82%, specificity 62%, PPV 33%, and NPV 94%. Patients frequently underwent subsequent examinations following initial imaging. Conclusion The findings were in accordance with the literature. Patients with NSSC had a cancer prevalence of 21%, most frequently gastrointestinal. The most frequent symptom was weight loss and, even as the only symptom, it is a potential marker for cancer. CT and 18F-FDG-PET/CT were sensitive with high NPV, whereas PPV was superior in CT. Better stratification by symptoms or findings is an obvious focus point for future studies to further optimise the NSSC-CPP work-up strategy.


Introduction
Correct and timely treatment of cancer requires a fast and accurate diagnostic strategy. Recognizing relevant symptoms is the first step, but only half of cancer patients present with alarm symptoms 1,2 . 15% of the general population experience alarm symptoms during a year 3,4 , whereas annual cancer incidence is much lower, e.g. < 1% in Denmark 5 . The diagnostic process is even more difficult if the cancer presents with non-specific symptoms and signs of cancer (NSSC), e.g. general malaise, fatigue, or weight loss 2,6,7 . These challenges was revisited by a recent analysis paper 8 .
In the late 1990s, cancers were diagnosed later in Denmark than in other European countries, and the mortality rates were higher 7,[9][10][11] . As a result, Danish Health Authority introduced national cancer pre-planned pathways (CPP) to ensure fast work-up 7, 12 for specific cancer suspicions and alarm symptoms, and also for NSSC (NSSC-CPP) 13,14 . Patients with NSSC are investigated in GP or hospital-based diagnostic centres 12 . Initial work-up usually include clinical examination, standard blood tests and diagnostic imaging; usually either CT of chest and abdomen or a positron emission tomography with 18  The NSSC-CPP are organized slightly different in each Danish region 17 and data is sparse on basic demographics, symptoms, cancer prevalence, and imaging in the NSSC populations at our hospital; this is currently debated between referring physicians and radiology departments.
The study objective was to contribute novel insights into NSSC-CPP patients including the distribution and significance of symptoms and the diagnostic yield of CT and [ 18 F]FDG-PET/CT with the aim to improve the overall workup process.

Methods
This was a retrospective review of medical charts. We included all patients with novel NSSC referred from GP, specialist practice, or local diagnostic centre for CT of chest and abdomen or [ 18 F]FDG-PET/CT at our hospital from 1 January-31 December 2020. We excluded patients with known cancers, cancer of unknown primary, referrals to specific CPP, and suspected nonmelanoma skin cancer.
Primary outcome was the proportion of NSSC patients diagnosed with cancer. Any finding on a scan that was considered suspicious of cancer was classified as true or false positive or negative using the final diagnosis from the medical charts as reference standard.
We followed patients for a maximum of one year after the initial scan, but stopped in case they were diagnosed with cancer. We verified cancer diagnoses from biopsy results in The Danish Pathology Register 18 and grouped them according to ICD-10 chapters C00-D49. When no biopsy was performed despite suspicious scans, e.g. due to patient frailty, an experienced oncoradiologist re-analysed the scans to assess whether the findings were consistent with cancer.
Secondary endpoints included the most frequent initial symptoms and the number of subsequent examinations during the diagnostic workup process.
We registered the most frequent cancer symptoms the patients presented with to their GP or diagnostic centre, and seven well-known warning signs of cancer, i.e. long-term dysphagia, weight loss, long-term coughing or hoarseness, changed bowel habits, unexplained bleeding, changed moles or non-healing wounds, changing lumps or swellings 19 . We limited the list to 20 symptoms/signs plus 'no symptoms' and 'doctor's gut feeling'.
In addition, we registered data on subsequent examinations induced directly by the scan, i.e. imaging, endoscopy, or biopsy performed within the follow-up period after initial referral, i.e. for instance control scans of known findings were not included.
Study data was collected and managed using REDCap (Research Electronic Data Capture), hosted by the Region of Southern Denmark 20, 21 .
Statistics were performed using Stata/BE 17 (StataCorp, Texas US) with continuous data reported by mean and standard deviation if normally distributed, and compared using Student's t-test. Discrete or non-normal data were presented by median and range. Categorical data was represented as prevalence, and differences tested using chi-square or Fisher's exact test. Statistical significance level was defined as 0.05.

Results
Of 802 referrals (178 from the diagnostic centre and 624 from GP or specialist practice), 248 matched the inclusion criteria (123 men and 125 women) ( Figure 1). Table 1 presents baseline demographics and characteristics.
Patients presented to the referring physician with 0-8 symptoms with a median of 2 ( Table 1). The most frequent symptom was weight loss (56% of the patients), and 78/248 (32%) had only one symptom. Some symptoms were more frequently associated with cancer: changed bowel habits (8/16), blood in the urine (1/2), or pain (23/80) ( Figure 3). When no symptoms were reported, the reasons for referral was usually abnormal results of blood samples or suspicious imaging findings.
After correlation of scans with final diagnoses, the diagnostic yield of initial CT and [ 18 F]FDG-PET/CT was calculated as presented in Table 2. Cancer prevalence was 22% (41/190) in the CT group and 19% (11/58) in the [ 18 F]FDG-PET/CT group, respectively.
Seven patients with findings suspicious of cancer on the initial CT scan had no biopsies. When CT scans were reanalysed, four of the patients had obvious cancer based on imaging alone and died during follow-up; these could arguably be classified as true positives.
Two non-solid haematological cancers (leukaemia) were not detected by [ 18 F]FDG-PET/CT ( Figure 2). These should be diagnosed by blood samples and not imaging. [ 18 F]FDG-PET/CT is suitable for the detection of solid tumours 22 and hematologic cancers could be categorised as true negatives with respect to solid tumours.
The diagnostic yield after applying these modifications can be found in Table 2.
After the initial scan, patients often underwent additional examinations; patients without cancer went through 0-5 subsequent examinations (median of 1). Of those, 62 underwent one examination and 44 underwent more. CT scans and endoscopies were the most common supplementary examinations (Table 3); 87 underwent endoscopies, in 40 cases with biopsy performed.

Discussion Summary
Twenty-one percent of included patients were diagnosed with cancer, most frequently in the digestive organs, respiratory system, and lymphoid or hematopoietic system. The most common initial symptoms were weight loss, pain, or fatigue. CT and [ 18 F]FDG-PET/CT had comparably high sensitivity and NPV, whereas CT had superior specificity and PPV.

Strengths and limitations
This is a retrospective study with the inherent limitations by this design, e.g. important data may unavailable and data is prone to bias and confounders that are difficult to control. To minimize the risk of selection bias, we included all referrals to CT received as part of a NSSC-CPP at our institution in 2020 regardless if they originated from GP, specialist clinic, or diagnostic centre. We also used biopsy as the reference standard and retrieved cancer diagnoses directly from the national pathology database to minimize the risk of misclassification and recall bias. This ensured that we did not miss a diagnosis even if a patient moved to a different region of Denmark.
On the negative site, the number of patients and scans in our dataset is relatively limited leading to some wide confidence intervals of all diagnostic properties (sensitivity, specificity, PPV, NPV) especially in the case of [ 18 F]FDG-PET/CT. Thus, a misclassification would have a considerable impact on diagnostic properties. Also, our study was observational with no interventions, meaning that the patients in our study were preselected for CT and [ 18 F]FDG-PET/CT depending on the clinical setting and the referring physicians discretion. Therefore, the two groups were not directly comparable, although the cancer prevalence was similar in both groups.
Some patients were undoubtedly diagnosed with non-malignant diseases that where relevant as differential diagnoses in the context of NSSC, but due to technical issues after a regional switch to a new electronic patient record, we did not have full access to historic electronic patient charts. Therefore, we could not investigate this further or verify any post-scan clinical procedures or examinations except endoscopies.
We registered initial symptoms based on referral text and there could be reporting bias if referring doctors disregarded some symptoms or findings in the referrals.

Comparison with existing literature
Our study found 21% had a biopsy proven malignancy. Another 1.6 % (4 / 248) had imaging findings in keeping with malignancy not confirmed by biopsy. Arguably, our prevalence is 23 % and within the range in the literature. Møller et al. found a prevalence of 20 % in a cohort from GP 23 . Prevalence of cancer with NSSC in Denmark, Sweden, UK, The Netherlands, Australia and Spain are found to be 9-35% 13, 14, 23-30 . The COVID-19 pandemic stressed healthcare systems in 2020 and a general decrease in cancer incidence was observed 31,32 . The prevalence of cancer in our study is similar to studies before COVID-19 14,23,26,33 and any influence of the COVID-19 pandemic could not be detected in our result.
In adherence to Danish GDPR legislation, we had to keep our overall groupings of malignant findings relatively broad; the most common cancer sites were the digestive organs (27%), respiratory system (15%), and lymphoid or haematopoietic malignancies (14%). These results were in keeping with the literature, for instance Chapman et al. found the three most common cancers to be GI-cancers (upper and lower) (35%), lung (22%), and haematological (13%) 24 . Several other national and international studies found comparable results albeit with variations in numbers 25,26,29,34 .
In agreement with other studies 13,15,26,34 , more than half of patients (56% overall, 12% of those with a cancer diagnosis) were referred with weight loss. For instance, in a large British study, Chapman et al. found weight loss in 66%, in 20% it was monosymptomatic 24 . Unintended weight loss is associated with cancer, but not often explored in a standardized manner 35,36 . Interestingly, it was recently discussed at our institution whether monosymptomatic weight loss is enough to qualify a patient for the NSSC-CPP. Given the frequency of this symptom among NSSC patients, weight loss may warrant further studies to investigate if it could predict cancer in itself. The second and third most common symptoms/findings in our study was pain (32% overall, 9% of cancer patients) and fatigue (28% overall, 5% of cancer patients). Chapman et al. found similar results, i.e. pain and fatigue was the third and fifth most common (32% and 19%, respectively) 24 .
Of note, 6.4% of the patients in our cohort presented with changed bowel habits, which would actually qualify them for the national colorectal CPP, and the reason why they entered the NSSC-CPP is unknown, but probably just signifies the complexity of this population.
Our cohort consisted of patients with NSSC from both GPs and the diagnostic center of our institution; NSSC are common and can be a challenge to GPs 34,36,37 . Organisation of CPP vary among institutions both nationally and internationally 29,34,38,39 and there is an ongoing effort to gather information to optimize the efforts 37 .
Imaging is routinely used in the diagnosis, staging and follow up of cancer, and although several studies address the diagnostic yield of advanced imaging in patients with NSSC, timing, first-line choice, and cost-effective use of imaging in patients with NSSC remains controversial 13,23,29,33,37,40 . For instance, a current protocol is testing rapid CT in this context 41 , and in Denmark there are disagreements over conventional CT or [ 18 F]FDG-PET/CT as first-line modality 15,16 .
Initial scans detected 85% (44/52) of cancers in our series. Møller et al. investigated the diagnostic properties of contrast enhanced CT in NSSC-CPP referred from GP. Cancer prevalence was 20%, and 92% had CT results classified as possible or probable cancer; a positive CT raised the probability of a cancer diagnosis to 62%, whereas a negative one decreased the probability to 2% 23 . Similar results were reported by Ormstrup et al. 33 . Most malignant diagnoses in our study were established during the initial work-up process, but in two cases, the diagnoses were not established until six months and 1 year, respectively, after referral. One patient presented with fatigue, weight loss, and gastrointestinal symptoms and was diagnosed with metastatic breast cancer 150 days later. The other presented with fatigue and dizziness and findings in keeping with arteritis on [ 18 F]FDG-PET/CT. This patient was diagnosed with chronic lymphocytic leukaemia 300 days later. It is unclear if their symptoms were related to their final malignant diagnosis.
Suspicion of prostate cancer was raised twice on [ 18 F]FDG-PET/CT but not reported as suspicious on CT (Figure 2). Hypertrophy of the prostate was reported in both patients with prostate cancer, but CT is not considered diagnostic in the routine workup for prostate cancer 43 .
Some patients were initially CT scanned before entering the NSSC-CPP and were subsequently referred for [ 18 F]FDG-PET/CT. Generally, patients without cancer went through multiple examinations like endoscopy illustrating the diagnostic challenges in this patient group. More than half of the non-cancer patients underwent at least one additional examination after the initial scan before the workup process was concluded. Other studies showed a similar trend 16,44 . [ 18 F]FDG-PET/CT may identify incidental gastrointestinal findings potentially representing malignancy 45 , thus, patients without cancer with initial [ 18 F]FDG-PET/CT underwent relatively more endoscopies compared to those with initial CT.

Conclusions
Our findings in NSSC-CPP patients were in accordance with the literature. We found a cancer prevalence of 21%, most frequently in the digestive organs. The most frequent symptom was weight loss as reported by more than half of the patients and even as the only symptom, it is a potential marker for cancer. CT and [ 18 F]FDG-PET/CT were sensitive with high NPV, whereas PPV was superior in CT. Patients without a cancer diagnosis underwent subsequent examinations following initial imaging with CT or [ 18 F]FDG-PET/CT.

Implications for clinical practice
Our data underlines the heterogeneous presentation of NSSC-patients with a multitude of potential symptoms and findings. Our data also supports the current imaging strategy in NSSC-CPP. CT and [ 18 F]FDG-PET/CT both have a place, but based on our data, it was not possible to establish why patients were referred for CT or [ 18 F]FDG-PET/CT. Future, prospective studies are needed to better stratify patients according to presentation to further optimise the NSSC-CPP workup strategy.

Ethical approval
Relevant permissions according to Danish legislation were obtained from the regional council (22/24999) and the hospital (22/21408) before the study began. In retrospective studies, informed consent is not required under Danish law.

Funding
Nothing to report.