This study was designed according to the Standards for Reporting Diagnostic Accuracy (STARD) 2015 statement and was approved by the Institutional Review Board (IRB) of the facility where the work was conducted (opinion No. GPPG-2018-0106) (Figure 1). Due to the completely retrospective design of this investigation and the absence of any intervention, there was no need for written informed consent; accordingly, a waiver was requested and obtained from the IRB. Intraoperative findings were considered the gold standard.

Figure 1.

STARD flow diagram of the study.

(0.41MB).

Inclusion criteria were as follows: all patients under 18 years old who were subjected to surgical resection of solid abdominal tumors who had undergone both sph-CT and USG for preoperative evaluation. None of the study patients had any evidence of metastatic spread. Exclusion criteria were the following: patients who underwent USG and MRI but had no CT images; patients who underwent only MRI and CT in the preoperative evaluation, without USG; patients who were not subjected to surgical resection; CT and USG performed for diagnosis, staging, or evaluation of complications; and patients with liver tumors, because the preoperative evaluation was performed by multiphase CT or MRI and their surgical team was not the same. Figure 1 illustrates the participant selection process. Between January 2009 and December 2017, 60 patients were eligible and 50 were included in the study.

USG was performed using ALOKA 4000® or PHILIPS HD11® with multi-frequency transducers (2-5 MHz convex and 5-12/5-10 MHz linear). All patients were accompanied by their parents or legal guardian, did not receive sedation or anesthesia, and the examination was performed after the ingestion of liquids.

USG was performed by an experienced pediatric radiologist and followed the institution's protocol, including tumor volume, calculated by the ellipsoid formula (height x width x thickness x 0.523); a grey-scale evaluation of each organ/structure; color Doppler flow of each vessel and circumferential involvement of the main abdominal vessels; the relationship of the tumor with adjacent organs and structures regarding the presence or absence of cleavage planes and signs of invasion; and the dynamic evaluation of the sliding sign between the tumor and adjacent structures during respiratory motion, Valsalva maneuver, gentle abdominal pressure, intestinal peristalsis, and arterial pulsatility. Invasion or encasement of adjacent structures by the tumor was defined as follows:[16] Absent, when the sliding between the tumor and adjacent structures was independent; present, when the tumor and the adjacent structures moved together (en bloc) or when there was no perceived sliding.

Sph-CT was performed using a PHILIPS Brilliance 16 or TOSHIBA Aquillion 64 according to the institution's protocol, which consisted of a contrast-enhanced single-phase image acquired only in the portal venous phase after intravenous administration of iodinated contrast under general anesthesia in small children and without anesthesia in older patients, according to the ALARA (as low as reasonably achievable) principle.18

The authors defined, as a consensus, that the statistical analysis would consider each solid organ, hollow viscus, skeletal muscle structure, and main abdominal vessel in contact or adjacent to the tumor as a study unit.

Image assessment, as well as sph-CT and USG reports of all solid abdominal pediatric tumors included in this study, were based on guidelines for the 2 main types of pediatric abdominal tumors. Neuroblastomas were evaluated according to international guidelines,6 and renal tumors were assessed according to the Umbrella protocol.5 To standardize CT and USG descriptions, the used terms followed the consensus of the neuroblastoma study group:[6] Separation meant there was a visible layer, usually consisting of fat tissue, between the tumor and the adjacent structures; contact meant there was no visible layer between the tumor and the adjacent structures, and for vascular structures, it meant that less than 50% of the circumference was in contact with the tumor; invasion of vascular structures meant that more than 50% of the circumference of the vessel was in contact with the tumor; infiltration meant that the tumor extended to other organs, vessels excluded, or that the margins were not well defined (Supplementary material 1).

Color Doppler USG was performed for all tumors that were in contact with the inferior vena cava, renal veins, superior and inferior mesenteric veins, portal vein, splenic vein, and iliac veins.5

For comparison purposes, abdominal organs and structures were divided into four groups: Group 1– solid organs (liver, spleen, pancreas, kidneys, adrenal glands, uterus, and ovaries); Group 2– major and intermediate abdominal vessels (abdominal aorta, iliac artery and vein, superior mesenteric artery and vein, inferior mesenteric artery and vein, celiac trunk, splenic artery and vein, hepatic artery, renal artery and vein, inferior vena cava, and portal vein); Group 3– gastrointestinal (GI) organs (stomach, small bowel, large bowel, and bladder); and Group 4– neuromuscular structures (diaphragm, abdominal wall, lumbar and sacral nerve roots).

All patients underwent surgical resection through an exploratory laparotomy (gold standard). The surgical team had full access to the sph-CT and USG reports. After the surgery, surgeons constructed a surgical report following a checklist based on images of all organs and structures that showed signs of invasion, circumferential involvement, or simply adherence to the tumor. The adjacent structures and organs that had been described as compromised by sph-CT or USG assessments and were found free in surgery (and vice-versa) were also included in the surgical reports. sph-CT and USG findings were correlated to the surgical reports.

Our primary endpoint was the correlation between sph-CT and USG findings and the surgical findings (gold standard). The analyzed performance metrics included sensitivity (SEN), specificity (SPE), positive predictive value (PPV), negative predictive value (NPV), and accuracy (AC). The Kappa test was used to evaluate sph-CT and USG agreement in the detection of invasion or encasement of organs and adjacent structures by the tumor.19 Statistical analysis was performed using the IBM SPSS Statistics software, version 20.0, and p < 0.05 was considered significant.

Figure 1 depicts the STARD flow diagram for this study. The demographic data and distribution of tumors according to age group are summarized in Tables 1 and 2. This study included 50 patients, of which 28 (56%) were female and 22 (44%) were male. The mean age was 25.5 months. Histologic findings indicated that 19 of the tumors were of neuroblastic origin and 20 were of renal origin; 11 patients presented other types of tumors, of which 5 were adrenal carcinomas. The mean tumor volume was 357cm³ (10-3930cm³). Complete surgical resection with negative microscopic margins (R0) was successfully performed in 44 (88%) patients. In two neuroblastoma patients, only partial tumor resection could be performed (debulking). The remaining 4 patients were considered unresectable (1-adrenal carcinoma, 1-rhabdomyosarcoma of the uterus, and 2-neuroblastic tumors).

Performance measurements are detailed in Table 3. At preoperative planning, sph-CT had an SEN of 90.3%, SPE of 86.8%, PPV of 75.3%, NPP of 95.3%, and AC of 87.9%. USG had an SEN of 86.6%, SPE of 94.6%, PPV of 86.9%, NPP of 94.4%, and AC of 92.2%. The Cohen's kappa coefficient between these two methods in the identification of the invasion/encasement of adjacent structures was 0.72 (p < 0.001). The comparison between sph-CT and USG results by the Chi-squared test with Yates correction showed statistical significance for SPE (p < 0.001), PPV (p = 0.002), and AC (p = 0.006).

The performance of sph-CT and USG in identifying the invasion/encasement of adjacent structures by the tumor was evaluated by grouping organs as described in the Methods section (Supplementary material 2). In Group 1 (solid organs), 252 units were evaluated by sph-CT and USG. At surgery, 89 units showed signs of invasion/encasement by the adjacent tumor (CT = 108; USG = 87). Sph-CT and USG presented respectively: 93.1% vs 88.7% SEN, 84.1% vs 95% SPE, 75.6% vs 90.8% PPV, 95.8% vs 93.9% NPP, and 87.3% vs 92.8% AC. The kappa correlation coefficient was 0.77 (p < 0.001). The comparison between sph-CT and USG results by the Chi-squared test with Yates correction showed statistical significance for SPE (p = 0.002) and PPV (p = 0.011).

In Group 2 (vessels), 293 units were evaluated by sph-CT and 307 by USG. At surgery, 91 units showed signs of invasion/encasement (CT = 80; USG = 84). Sph-CT and USG, respectively, demonstrated a SEN of 91.2% vs 92.3%, SPE of 91% vs 92.1%, PPV of 82.2% vs 83.2%, NPP of 95.8% vs 96.6%, and AC of 91.1% vs 92.2%. The kappa correlation coefficient was 0.79 (p < 0.001). The comparison between sph-CT and USG by the Chi-squared test with Yates correction did not show statistical significance.

In Group 3 (GI organs), 54 units (20 invasion/encasement) were identified by sph-CT and 59 (8 invasion/encasement) by USG. During surgery, 14 units were detected. Sph-CT and USG provided a SEN of 85.7% vs 63.6%, respectively, a SPE of 80% vs 97.9%, PPV of 60% vs 87.5%, NPP of 94.1% vs 92.2%, and AC of 81.5% vs 91.5%. The kappa correlation coefficient was 0.27 (p < 0.027). The comparison between sph-CT and USG by the Chi-squared test with Yates correction showed statistical significance for SPE (p = 0.016).

In Group 4 (neuromuscular structures), 135 units (42 invasion/encasement) were identified by sph-CT and 142 (29 invasion/encasement) by USG. Surgery detected 33 units with signs of invasion/encasement. Sph-CT and USG demonstrated a SEN of 81.8% vs 74.1%, respectively, a SPE of 85.2% vs 97.3%, PPV of 64.3 vs 88.5%, NPP of 93.6% vs 93.1%, and AC of 84.4% vs 92.3%. The kappa correlation coefficient was 0.57 (p < 0.001). The comparison between sph-CT and USG by the Chi-squared test with Yates correction showed statistical significance for SPE (p = 0.004).

USG allowed the diagnosis of the invasion of vascular structures (aorta, splenic artery, right renal vein, and left renal vein), which was missed by sph-CT in 4 patients (1 for each structure). Similar results were observed in 3 patients with muscular invasion (diaphragm, anterior abdominal wall, and psoas muscle). A large bowel invasion was also detected by USG and missed by the sph-CT. Therefore, in 14% (7/50) of the patients (1 patient had 2 invaded structures), the invasion of adjacent structures was diagnosed by USG but not by sph-CT.