Diagnostic accuracy of non-contrast abdominal CT scans performed as follow-up for patients with an established cancer diagnosis: a retrospective study

Abstract

Objectives: To determine the accuracy and non-detection rate of cancer related findings (CRFs) on follow-up non-contrast-enhanced CT (NECT) versus contrast-enhanced CT (CECT) images of the abdomen in patients with a known cancer diagnosis.

Methods: A retrospective review of 352 consecutive CTs of the abdomen performed with and without IV contrast between March 2010 and October 2014 for follow-up of cancer was included. Two radiologists independently assessed the NECT portions of the studies. The reader was provided the primary cancer diagnosis and access to the most recent prior NECT study. The accuracy and non-detection rates were determined by comparing our results to the archived reports as a gold standard.

Results: A total of 383 CRFs were found in the archived reports of the 352 abdominal CTs. The average non-detection rate for the NECTs compared to the CECTs was 3.0% (11.5/383) with an accuracy of 97.0% (371.5/383) in identifying CRFs. The most common findings missed were vascular thrombosis with a non-detection rate of 100%. The accuracy for non-vascular CRFs was 99.1%.

Conclusion: Follow-up NECT abdomen studies are highly accurate in the detection of CRFs in patients with an established cancer diagnosis, except in cases where vascular involvement is suspected.

Introduction

Increased physician reliance on computed tomography (CT) for confirmation or exclusion of many clinical diagnoses has resulted in a continued surge in CT utilization. Some estimate three million CT examinations were performed yearly in 1980 compared to 62 million currently [1].

This highlights continued concerns surrounding estimates that one-third of examinations may not be medically necessary and that 1.5–2.0% of all cancers in the United States may be attributable to the radiation from CT studies [1,2].

In light of these considerations, scrutiny of CT usage and protocol appropriateness are necessary to provide evidence-based support for the safe and cost effective use of CT.

Contrast-enhanced CT (CECT) has been widely used in the diagnosis, staging and follow-up of malignancy. While its efficacy has been established [3–6], adequate scrutiny should be given to the usefulness and safety of each intervention. Such analysis requires assessment of costs and benefits, looking at both financial cost and associated patient risk. Practitioners must use diagnostic ionizing radiation responsibly by using appropriate study selection, dose reduction techniques, patient education and provider consultation. Moving in this direction, there has been increased emphasis upon, and increased public awareness of radiation exposure and its stochastic risks [1,7–9].

A similar analysis of usefulness and cost must also be applied to the variable or elective components of a procedure. This includes questioning the use of iodinated contrast as a mainstay component of many CT imaging protocols. While current iodinated contrast options have a relatively reduced risk profile compared to their predecessors [10–12], the incidence of adverse clinical outcomes still approaches 3% [13]. Such risks include the controversial (e.g., contrast-induced nephropathy), and nonspecific (nausea, vomiting, heat) [10,13–21].

Iodinated contrast has an important role in many imaging applications. It results in superior sensitivity and specificity to non-contrast studies in many clinical scenarios partly due to increased contrast resolution. However, we suspect there are clinical scenarios in which it may not result in a significant enough benefit to justify its use.

The utility of iodinated contrast has been assessed within the context of many disease processes, patient presentations and clinical settings, including both the acute [22–27], and non-acute/oncologic indications [3–6,28].

To our knowledge, there are no published studies comparing the usefulness of non-contrast-enhanced CT (NECT) of the abdomen to combined CECT plus NECT performed for the follow-up of patients with an established cancer diagnosis. We hypothesize that NECT alone will demonstrate equivalent efficacy in lesion identification to CECT plus NECT in follow-up oncologic imaging, with expected limitations as detailed within this article. Given the high volume of examinations within this patient population, showing such equivalency, could result in significant benefits to both the health care system and patients. Replacing CECTs with NECTs would result in significant financial saving and reduced patient complications even given the low risk profile of intravenous iodinated contrast.

Material and methods

Study cohort

Our study was approved by our Institutional Review Board Committee. Informed consent was waived because it was retrospective. Between March 2010 and October 2014, 352 consecutive abdominal CT scans performed at our institution meeting our inclusion criteria were included. We included all CTs of the abdomen and pelvis on patients over 18 years old and performed for the follow-up of previously diagnosed cancer at our institution during this time period. All CTs of the abdomen and pelvis during this time were performed with and without intravenous contrast and had field of view spanning from the lung bases through the pelvis. Patients were between 21 and 94 years of age with a mean age of 61.5 years. The population was 53.5% male (n = 152) and 46.5% female (n = 132). Ages of the male patients ranged between 21 and 94 years of age with a mean age of 62 years. Ages of the female patients ranged between 27 and 86 years of age with a mean age of 61 years. The study was conducted at an academic medical center in the United States, with most patients coming from the Department of Oncology.

CT scan protocol

CT scans were obtained using a 16-slice multidetector CT scanner (Toshiba Aquilion) and 64-slice multidetector CT scanner (Toshiba Aquilion). Imaging parameters for the 16-slice multidetector CT scanner were as follows for abdomen/pelvis studies: 120 kV and 350 mAs, with 0.5 s rotation time and helical pitch 23. Imaging parameters for the 64-slice multidetector CT scanner were as follows for abdomen/pelvis studies: 120 kV and 500 mAs (maximum), with 0.5 s rotation time and helical pitch 95. Contrast-enhanced studies were obtained with intravenous administration of 100 ml Omnipaque 350 mgI/ml (GE Healthcare; Princeton, NJ) by dual power Medrad Stellant injector at a rate of 3.0 ml/s with the use of 22-gauge antecubital vein plastic angiocatheter. Oral contrast was utilized for some studies at the discretion of the ordering physician and included 225 ml of Redicat 2 (E-V-EM Canada Inc; Lake Success, NY) administered 1 h and then again half an hour following the non-contrast portion of the exam.

Data analysis

Independent interpretation of all CT scan images including scout images was performed by two general radiologists (14 years and 1 year of experience). The only history provided was the primary cancer diagnosis and only the most recent prior NECT was available as a comparison. Each independent read included special attention to each abdominal and pelvic organ system within the field of view and was compared to the archived report as a gold standard. When only one radiologist identified an important finding, the discrepancy was resolved by consensus. Consensus results were also compared to the CRFs mentioned in the archived reports as the gold standard and the error rates were calculated. Inter-reader reliability statistics were determined on an excel spreadsheet. CRFs were classified according to the type of finding and the location is shown in Table 2. Lymph nodes were considered significant if they were >1 cm smallest dimension. Findings found only on the CECT due to differences in anatomic coverage between the CECT and NECT scans occurring at the superior or inferior margins of the study were excluded from our statistics.

Results

Our study included 352 abdomen and pelvic CT scans performed on 284 patients. CT scans were performed between 3 and 24 months following the initial diagnosis of cancer. Colorectal, renal, lymphoma and pancreatic cancer were the most common cancers. Nine patients had two prior cancers. Table 1 lists the number of patients who were being followed for each type of cancer. Oral contrast was used in 204 exams (58%) at the discretion of the ordering physicians.

Table 1. Distribution of cancer according to the organ involved.

Colon/rectal cancer	35 (12.3%)
Kidney cancer	26 (9.2%)
Lymphomas	24 (8.5%)
Pancreatic cancer	24 (8.5%)
Breast cancer	17 (6%)
Prostate cancer	16 (5.6%)
Lung cancer	14 (4.9%)
Ovarian cancer	13 (4.6%)
Melanoma	12 (4.2%)
Uterine/endometrial and cervical cancers	12 (4.2%)
Urinary bladder cancer	11 (3.9%)
Soft tissue and retroperitoneal sarcomas	10 (3.5%)
Liver cancer	9 (3.2%)
Testicular cancer	8 (2.8%)
Esophageal cancer	6 (2.1%)
Intestinal/gastric cancers	5 (1.8%)
Unknown primary cancer	5 (1.8%)
Adrenal cancer	4 (1.4%)
Leukemia	4 (1.4%)
Gallbladder cancer	4 (1.4%)
Pseudomyxoma	3 (1%)
Laryngeal cancer	2 (<1%)
Thyroid cancer	2 (<1%)
Brain cancer	2 (<1%)
Nasal cancer	1 (<1%)
Squamous cell carcinoma of skin (SCC)	1 (<1%)
Vulvar cancer	1 (<1%)
Ganglioneuroma	1 (<1%)
Urethral cancer	1 (<1%)
Bile duct cancer	1 (<1%)
Femur cancer	1 (<1%)
Cervical and lymphoma cancers	2 (<1%)
Cervical and colon cancers	1 (<1%)
Lymphoma and prostate cancers	1 (<1%)
Lymphoma and kidney cancers	1 (<1%)
Melanoma and kidney cancers	1 (<1%)
Breast and pancreatic cancers	1 (<1%)
Breast and ovarian cancers	1 (<1%)
Urinary bladder and sarcoma cancers	1 (<1%)
Total	284 patients (100%)

Table 2. Number of findings according to category, non-detection rate and inter-rater reliability.

Category	NECT & CECT	NECT onlyReader # 1	NECT onlyReader # 2	NECT onlyaverage	Non-detection %	Match
Mass or neoplasm (68)
Renal mass	15	15	15	15	0%	1
Pancreatic mass	13	13	13	13	0%	1
Sigmoid-rectal mass	10	10	9	9.5	5%	1
Adrenal mass	6	6	6	6	0%	1
Adnexal mass	5	5	5	5	0%	1
Liver & biliary ducts masses	4	4	3	3.5	12.50%	0
Gastric mass	3	3	3	3	0%	1
Colonic mass	2	2	2	2	0%	1
Esophageal mass	2	2	2	2	0%	1
Presacral mass	2	2	2	2	0%	1
Urinary bladder mass	1	1	0	0.5	50%	0
Lung mass	1	1	1	1	0%	1
Retroperitoneal mass	1	1	1	1	0%	1
Cervical mass	1	1	1	1	0%	1
Gluteal mass	1	1	0	0.5	50%	0
Upper extremity mass	1	1	1	1	0%	1
Lymphadenopathy (98)
Retroperitoneal LN	30	30	29	29.5	1.70%	0
Mesenteric LN	16	16	16	16	0%	1
Para-aortic LN	11	11	11	11	0%	1
Mediastinal LN	9	9	9	9	0%	1
Porta-hepatis LN	8	8	7	7.5	6.20%	0
Inguinal LN	8	8	8	8	0%	1
Internal Iliac LN	7	7	6	6.5	7.10%	0
External Iliac LN	6	6	6	6	0%	1
Axillary LN	3	3	3	3	0%	1
Metastatic disease (149)
Lung and pleural metastasis	55	55	55	55	0%	1
Liver metastasis	53	53	53	53	0%	1
Skeletal metastasis	24	24	24	24	0%	1
Adrenal metastasis	16	16	16	16	0%	1
Subcutaneous metastasis	1	1	1	1	0%	1
Organ wall thickening (17)
Ureteral wall thickening	8	8	8	8	0%	1
Bladder wall thickening	3	3	3	3	0%	1
Gastric wall thickening	2	2	2	2	0%	1
Distal esophageal wall thickening	1	1	1	1	0%	1
Sigmoid wall thickening	1	1	1	1	0%	1
Colon wall thickening	1	1	1	1	0%	1
Duodenal wall thickening	1	1	1	1	0%	1
Venous thrombus (8)
Portal vein thrombus	2	0	0	0	100%	1
Splenic vein thrombus	2	0	0	0	100%	1
IVC thrombus	1	0	0	0	100%	1
Sup mesenteric vein thrombus	1	0	0	0	100%	1
Right hepatic vein thrombus	1	0	0	0	100%	1
Right atrium thrombus	1	0	0	0	100%	1
Effusion (39)
Pleural effusion	14	14	14	14	0%	1
Pericardial effusion	11	11	11	11	0%	1
Suspicious malignant ascites	14	14	14	14	0%	1
Miscellaneous (4)
Pancreatic ducts irregularities	1	1	1	1	0%	1
Mesenteric deposits	1	1	1	1	0%	1
Retroperitoneal deposits	1	1	1	1	0%	1
Omental Infiltration	1	1	1	1	0%	1
Total	383	375	368	371.5	Average 3.0%	368/375 IRR = 98.1%

CRFs were classified according to their type and then by location. The number of CRFs found in each subcategory is listed in Table 2. CRFs were found in 182 of the 352 exams (48%) with none being found in the remainder. The total number of cancer-related findings (CRFs) in the archived radiological reports was 383. Metastatic disease accounted for 149 findings, lymphadenopathy 98, primary neoplasms 68, pleural/pericardial effusions 25, organ wall thickening 17, venous thrombus 8 and 4 miscellaneous findings. The archived radiological reports based on NECT and CECT scans contained 383 CRFs whereas our 2 radiologists found up to 375 CRFs using the NECTs alone. The first radiologist found 375 CRFs while the second found 368 CRFs. The combined non-detection rate was 2.1% (8/383) and the diagnostic accuracy of the NECT was 97.9% (375/383). The average non-detection rate between the two radiologists was 3.0% (11.5/383) and the diagnostic accuracy of the NECT was 97.0% (371.5/383). The non-detection rate for venous thrombus was 100% (8/8 cases). The non-detection rates for all other categories excluding venous thrombosis was 0.9%. Table 2 summarizes the non-detection rates for all findings. There was excellent reliability between the two readers. The inter-reader reliability (IRR) was calculated at 98.1% (368/375) and is listed in Table 2.

Discussion

Our study showed that CRFs are common in the follow-up CT scans of previously diagnosed cancer patients. We reviewed 352 abdominal CT scan images performed 3–24 months following the initial diagnosis of cancer in 284 patients and found a 52% prevalence of CRFs. Metastatic disease was the most common finding found in 149 cases. Lymphadenopathy was the second most common finding with 98 cases.

Our study also showed that follow-up NECT has a low non-detection rate for CRFs of the abdomen as compared to combined with and without contrast CT. The average non-detection rate for all CRFs was 3.0%. Excluding venous thrombosis the non-detection rate was only 0.9%.

The high degree of accuracy for detecting CRFs with NECT for follow-up of patients with known diagnosis of cancer has many implications depending on the clinical scenario. If the clinical question is the presence or absence of CRFs and/or new lesions following treatment or during surveillance then clinicians should feel confident in the ability of a NECT if there is low concern or suspicion of vascular involvement. However, if the clinical question is related to degree of enhancement or exact size of a CRF then an enhanced study should be obtained but only if these parameters would affect the further management of the patient. It should be stressed that the use of contrast should only be employed if the additional information obtained would alter the future treatment or management of the patient. As mentioned previously one such scenario would be tumors with a propensity for and a staging system highly dependent on the presence or absence of venous invasion such as a renal cell carcinoma among others. Allowing for circumstances in the imaging algorithm for NECT for follow-up of patients with known cancer reduces healthcare costs (contrast as well as renal function labs), reduces patient discomfort (IV access) and decreases the real albeit low risk of IV contrast. It also gives the clinician some confidence in the utility of ordering a NECT in patients with a contraindication to IV contrast as it can still provide relevant information regarding CRFs. We believe further evaluation of the utility of NECT versus CECT in specific clinical situations and specific malignancies needs further investigation to whether or not the choice affects patient management and even outcomes.

Additionally, if widely adopted the non-detection rate for NECT would be expected to be lower as reader familiarity and experience in image interpretation of these studies increases. The 98% inter-reader reliability was excellent and could in part be explained by the difference in experience between the two readers.

Our study had several limitations. One was that examinations from outside institutions were not obtained and therefore our study may not have taken into account non-detected CRFs found outside our institution. Another is that the clinical significance of these findings and any changes on their detection or non-detection on treatment in unknown. More specifically our study identified the presence of the abnormality but did not characterize it for changes in size or appearance which may or may not have implications on treatment. It should be noted however that some cases showed the presence of new lesions and this alone may be sufficient to guide management, rather than an interval lesion size comparison. Another limitation is that we did not isolate the number of cases or findings attributable to the use of oral contrast, however it would not be more than 3% (the average non-detection rate) and thus oral contrast likely did not contribute much as only cases related to venous IV contrast were missed. Lastly, though our study has a relatively large number of cases, the numbers are small with respect to individual findings/organs involved and types of cancer. So while our study does demonstrate the utility of NECT alone for CRFs in the abdomen and pelvis in general, its ability to influence clinical decision-making in specific circumstances may be limited except in the presence of a strong contraindication to contrast. Additional studies will likely be required more narrowly focused on individual organs/findings or types of cancer.

In conclusion, our study demonstrates that NECT alone performed for follow-up of known malignancy has the potential to detect most CRFs in the abdomen and pelvis as compared to a combined NECT and CECT with an average accuracy of 97%. The only finding that was commonly missed was venous thrombosis. Excluding venous thrombosis the accuracy of NECT was 99.1% for CRFs. Given the expense and real albeit low risks of IV contrast, the utility of NECT alone should be an area of further investigation. Additional study with respect to individual findings/organs/cancers will be required to influence clinical practice except in the setting of strong contraindication to iodinated contrast.

Disclosure statement

No potential conflict of interest was reported by the authors.