Second-Opinion Interpretations of Gynecologic Oncologic MRI Examinations by Sub-Specialized Radiologists Influence Patient Care.

The need to simultaneously improve the quality of health care (patient outcomes) and reduce health care costs has been widely recognized nationally and internationally [ 1 – 4 ]. In the United States, a value-based health care delivery system – where payments are based on outcomes rather than the volume of services rendered – is increasingly being pursued as a solution, thanks in part to the passage of the Affordable Care Act [ 5 ; 6 ]. An important feature of a value-based system is the organization of health care delivery around integrated practice units, in which physicians with relevant expertise work jointly to provide coordinated, evidence-based care for patients with a single medical condition. Today, many oncology centres, including our tertiary care cancer centre, rely on multidisciplinary disease management teams (DMT) who care for patients with specific types of cancer. Because precise diagnosis and assessment of disease extent are considered essential for determining an appropriate treatment plan, each DMT, in addition to specialists in surgery, medical oncology, radiation oncology, and pathology, includes radiologists with relevant, focused expertise. Before DMT meetings at our institution, imaging studies obtained elsewhere are often re-interpreted by such sub-specialized radiologists and an official second-opinion report is issued. Rates of disagreement between initial and second-opinion imaging interpretations have been examined in a number of studies, in fields such as oncologic imaging, abdominal imaging, neuroradiology, paediatric imaging, and emergency radiology [ 7 – 18 ]. Some of these studies emphasized the frequency of minor and major discrepancies rather than their impact on patient care as determined by referring clinicians. The aim of this study was to determine the added value of second-opinion review of gynaecologic oncologic (GynOnc) MRI examinations by GynOnc radiologists, as assessed by treating physicians. We decided to focus on GynOnc MRI because of the limited prior literature in this area and the fact that GYN MRI scans are commonly interpreted by general body radiologists, who may have limited subspecialty training in gynaecologic MRI [ 19 ; 20 ].

The median patient age was 51 years (range: 18–90 years). Of 469 women included in the study, 130 women had a known diagnosis of cancer; 123 of these 130 women had gynaecologic malignancies and 7 had either non-gynaecologic cancers or cancers of unknown primary ( Table 1 ). The GynOnc MRI examinations were obtained for a variety of clinical indications, summarized in Table 2 and Figure 2 . GynOnc MRI studies were submitted from 178 different institutions. Four hundred and twenty-three (90.2%) of the 469 MRI studies were first interpreted by radiologists at private community hospitals or outpatient radiology facilities, and 46 (9.8%) were initially read by radiologists at academic tertiary care centres; 438 (93%) included contrast-enhanced sequences. The median time interval between initial and second-opinion interpretations was 19 days (range: 1–181 days). Second-opinion review found 82 (17.5%) of 469 MRI studies to be of limited imaging quality ( Figure 3 ); repeat imaging was recommended in 25 (30.5%) of 82 MRI examinations, and 7 MRIs were immediately repeated to compensate for the technical limitation of the initial study. The reasons for this limited imaging quality assessment are summarized in Table 3 . The preliminary review by two radiologists found that for 288/469 MRI studies (61.4%; 95%CI:56.8–65.8), there were no disagreements between the initial and second-opinion interpretations, while for 181/469 (38.6%; 95%CI:34.2–43.2), there was some disagreement (either clinically unimportant or important). The discrepancies concerned the number and locations of lesions detected (for example, presence of pelvic lymphadenopathy or peritoneal carcinomatosis) in 37/469 (7.9%; 95%CI:5.7–10.8), interpretation of findings (for example, determination of the origin of an adnexal mass) in 124/469 (26.4%; 95%CI:22.6–30.7) or both in 20/469 (4.3%; 95%CI:2.7–6.6) patients ( Figure 3 ). For 124 (68.5%; 95%CI:61.1–75.1) of the 181 cases with discrepant MRI reports, the discrepancies were deemed clinically important by at least one of the gynaecologic oncologic surgeons. In other words, discrepancies between initial and second-opinion interpretations would theoretically have affected patient care in up to 124/469 (26.4%; 95%CI: 22.6–30.7) patients. Second-opinion review removed the suspicion of gynaecologic malignancy raised by the initial report for 46/124 (37%; 95%CI:28.7–46.3) patients, suggested a new diagnosis of cancer omitted from the initial report for 12/124 (10%; 95%CI:5.3–16.6) patients, upstaged or downstaged disease for 18/124 (14%; 95%CI:9.1–22.2) patients, and provided a more specific diagnosis (for example, epithelial ovarian neoplasm versus ovarian mass) in 48/124 (39%; 95%CI:30.2–47.9) patients. Comparison to histopathology or follow-up imaging showed that the second-opinion interpretation was correct for 103 (83%; 95%CI:75.0–89.0) of the 124 MRI studies for which at least one surgeon identified clinically important discrepancies between initial and second-opinion interpretations. Neither initial nor second-opinion report was correct in 12/124 cases (9.7%; 95% CI: 5.3–16.6). Initial reports were correct in only 9/124 (7.3%; 95% CI: 3.6–13.7) of cases with clinically important disagreements. Detailed summaries of the results for each surgeon are provided in Tables 4 and 5 . Briefly, second-opinion review of GynOnc MRI would have theoretically affected patient care in 94/469 (20%; 95% CI:16.5–24.0) patients for surgeon 1 and 101/469 (21.5%; 95%CI:17.9–25.6) patients for surgeon 2. The treatment approach would have been altered in 71/469 (15.1%; 95%CI:12.1–18.8) and 60/469 (12.8; 95%CI:10.0–16.2) patients for surgeons 1 and 2, respectively. According to surgeons 1 and 2, these treatment changes would have prevented unnecessary surgery in 35 (7.5%) and 31(6.6%) patients, respectively, and would have altered surgical procedure type/extent in 19 (4.1%) and12 (2.5%) patients, respectively. Additionally, discrepancies between initial and second-opinion interpretations would have led to changes in counselling for 92/469 (19.6%; 95%CI:16.2–23.6) patients for surgeon 1 and 101/469 (21.5%; 95%CI:18.0–25.6) patients for surgeon 2.

Gynaecologic MR imaging plays an important role in the care of patients with gynaecologic tumours, often contributing to initial diagnosis, determination of disease extent, treatment selection and treatment follow-up [ 23 ]. For example, in patients with endometrial cancer, MR imaging is a valuable tool for pretreatment risk stratification, aiding the identification of patients who stand to benefit from pelvic and para-aortic lymph node dissection [ 22 ; 24 ; 25 ]. In women with cervical cancer, MR imaging improves the accuracy of the FIGO clinical stage determination, leading to more precise treatment selection and planning [ 26 – 28 ]. MR imaging is crucial for confirming eligibility for fertility-sparing surgical or medical procedures in patients with early-stage endometrial or cervical cancer who desire fertility preservation [ 29 – 32 ], and in women with adnexal masses of an indeterminate nature on ultrasound, MR imaging is the diagnostic problem-solving modality of choice [ 33 ; 34 ]. In our study of 469 MRI examinations, initial reports and second-opinion interpretations by GynOnc MRI subspecialists disagreed in more than a third (181/469) of cases. Some of these discrepancies were not clinically relevant and would not have affected patient care. However, according to the opinions of experienced gynaecologic oncologic surgeons, second-opinion review would have changed some aspect of clinical management in at least one fifth (20–21.5%) of patients, led to a change in the treatment approach for 12.8–15.1% of patients, prevented unnecessary surgery in 6.6–7.5% of patients, improved the surgical approach in 2.5–4.1% of patients, and replaced a non-operative management strategy with a more appropriate, surgical intervention for 1.9–2.3% of patients. The second-opinion report proved correct in the majority (103/124, 83%) of cases for which at least one surgeon found clinically important differences between initial and second-opinion MRI reports. We hypothesize that the focused, subspecialty expertise of the radiologists who performed the interpretations largely accounts for considerable differences in interpretations. This proficiency has likely developed as a result of repeated exposure to a large volume of cases with similar clinical findings and questions, a concept that is supported by prior literature, including GYN imaging. As a part of UK nationwide audit, Duncan et al evaluated the performance of various centres across the UK for the assessment of endometrial cancer stage using MRI. That study found that the centres with higher case load were significantly more accurate at the evaluation of the depth of myometrial invasion using histopathology as a gold standard [ 35 ]. To our knowledge, no prior studies have explored the added value of second-opinion radiology subspecialty review of GynOnc MRI examinations. Two previous reports examined the impact of a gynaecological oncology tumour board on the management of women with known or suspected gynaecologic malignancies referred to a tertiary care centre [ 19 ; 20 ]. Cohen et al reported on 509 patients discussed at the meetings of a gynaecological oncology tumour board and found that major discrepancies were seen for 30 of the patients (5.9%) [ 19 ]. The study did not specify the total number of patients who had outside imaging or pathology examinations reviewed or what types of examinations were reviewed. Greer et al reported on 215 patients whose cases were presented for radiological review (mostly of CT and PET scans) at a weekly multidisciplinary gynaecologic oncology tumour board; they found that the secondary radiological review led to a new diagnosis of gynaecologic malignancy or upstaging of known cancer in 19 (10%) of patients [ 20 ]. Our findings assessing the value of second-opinion image interpretations are in agreement with those of several studies involving other radiology sub-specialties. Gollub et al. reported on 143 body CT scans reinterpreted at a tertiary care cancer centre and found major disagreements between initial and second-opinion interpretations of 24/143 (17%) scans; the disagreements led to changes in management in 5/143 (3.5%) patients [ 10 ]. Eakins et al. analyzed 733 paediatric cases for which diagnostic imaging examinations were re-reviewed at a paediatric hospital; they found major discrepancies between initial and second-opinion reports for 168 (21.7%) examinations [ 8 ]. The second-opinion interpretations were more accurate than the initial reports in 83 (90%) of 92 discrepant cases with proven diagnoses. In a review of 4,534 neuroradiology cases, Zan et al. found clinically significant discrepancies between initial and second-opinion reports for 347 (7.7%) imaging studies [ 13 ]. The second-opinion review was correct in 163 (84%) of 194 discrepant cases with pathologically proven final diagnoses. When comparing initial and second-opinion reports for 396 patients who were referred to surgical oncologists at four academic centres, Dudley et al. found that reports disagreed in 162 (41%) of all cases, and the second-opinion interpretations were correct in 153 (94%) discrepant cases [ 18 ]. Brook et al reported on 383 consecutive patients presented at a Radiology Conference of the Division of Oncology and found these second-opinion consultations lead to major changes in management of 37% cancer patients and provided additional important information in up to 50% of patients [ 14 ]. Our study had several limitations. The study design is retrospective potentially leading to a number of selection biases; for example, only GYN MRI examinations submitted for second-opinion review were included; it is possible that the rate of clinically relevant discrepancies would have been different if all outside GYN MRI studies were submitted for review. Second, the distinction between clinically important and unimportant discrepancies is subjective. We tried to limit this concern by recruiting two senior gynaecological oncologic surgeons to evaluate all de-identified MRI reports with any discrepancies between initial readings and second-opinion interpretations. In conclusion, our results indicate that second-opinion review of GynOnc MRI by sub-specialized radiologists can impact patient care, allowing for more informed medical decision-making. In conjunction with previous studies in other imaging disciplines, our findings support the notion that subspecialty training and focused expertise influence patient management in a setting of multidisciplinary, disease-specific team-based care. Furthermore, second-opinion consultations should be viewed as a valuable and reimbursable clinical service.

Our institutional review board approved and issued a waiver of informed consent for this retrospective study, which was compliant with the Health Insurance Portability and Accountability Act. A retrospective search of our institutional clinical database was performed to identify all patients fulfilling the following inclusion criteria:(i) consecutive submitted GynOnc MRI performed and initially interpreted at an outside institution between January 1 st 2008 and August 1 st 2013; (ii) second-opinion GynOnc MRI interpretation documented in an official report issued by one of four sub-specialized gynaecologic radiologists at our institution; and (iii) histopathology or ≥ 6-month follow-up imaging after the MRI. This search yielded 525 outside GynOnc MRI studies with both initial reports and second-opinion interpretations. To ensure that second-opinion review did not benefit from extra clinical information, we excluded 56 MRI studies obtained in patients who had biopsy and/or surgery (45 patients) or additional diagnostic imaging test(s) (11 patients) performed between the initial report and second-opinion interpretation. The final sample consisted of 469 consecutive submitted GynOnc MRI studies obtained outside our institution, each of which had been reinterpreted by one of four gynaecologic oncologic radiologists with at least 5 years of sub-specialty imaging experience gained by actively participating in the gynaecologic oncologic tumour board. Two radiologists (with 5 and 7 years of experience in diagnostic imaging), who did not participate in the second-opinion interpretations, examined the initial and second-opinion interpretations of each MRI and, in consensus, divided all the MRI studies into two groups: i) NO disagreement and ii) ANY disagreement between initial and second-opinion interpretations. Two board-certified gynaecologic oncologic surgeons, each with 20 years of experience, independently reviewed initial and second-opinion reports for all cases in which the two reports contained any disagreements. The surgeons were provided with relevant clinical history but were blinded to the origin of each MRI report (i.e., each surgeon was presented with de-identified written summary of key imaging findings for each MRI report as illustrated in Figure 1 . The two radiologists described above created these summaries. The surgeons were blinded to the origin of each MRI report throughout the study). For every pair of MRI reports, each surgeon recorded whether differences between the reports were clinically important, i.e. would theoretically have led to a change in patient management ( Figure 1 ). A change in patient management was defined as an alteration in any of the following: i) treatment approach; ii) patient counselling; iii) patient referral (tertiary care centre versus community-based practice). In instances where the difference between the initial and second-opinion reports would have led to a modification in treatment approach, each surgeon was asked to record their preferred treatment strategy and any additional required tests, basing their judgment on each MRI report and the relevant clinical history provided. When clinically relevant discrepancies were identified between initial and second-opinion interpretations, the precision of the second-opinion interpretation was evaluated by using histopathology (in 103/124 cases [83.1%]) or by using minimum 6-month imaging follow-up (in 21/124 cases [16.9%]) as the reference standard. The number of MRI examinations that were considered limited due to their image quality by the sub-specialized radiologists was noted. The reasons for such assessment were obtained from the second-opinion report and classified as follows: 1) artefacts (for example, motion-induced image blurring), 2) absence of key pulse sequence(s) (for instance, the lack of oblique axial T2-weighted images for the assessment of parametria in studies acquired for the initial staging of cervical cancer or the absence of multiphase contrast-enhanced sequences in the studies obtained for the initial staging of the endometrial cancer), 3) suboptimal imaging technique (for example, low magnet strength, large field-of-view). Confidence intervals (CI) were estimated using the Wilson score interval with continuity correction [ 21 ]. The programming language R (version 3.1.1; R Foundation for Statistical Computing, Vienna, Austria) was used to perform all statistical computations.