PET/MRI versus PET/CT in oncology: a prospective single-center study of 330 examinations focusing on implications for patient management and cost considerations

PET/MRI (positron emission tomography/magnetic resonance imaging) is a relatively novel hybrid imaging technique that has recently been introduced into routine clinical practice. Today, the vast majority of PET/MRI systems are installed in tertiary care centers, where this imaging technique is partly or mainly used for research [1]; worldwide, the number of PET/MRI systems is gradually increasing [1]. Compared with PET/CT (computed tomography)—the standard hybrid imaging technique—PET/MRI offers reduced radiation exposure and higher morphological soft-tissue contrast.

For oncologic imaging, several comparative studies between PET/MRI and PET/CT have been performed within the last couple of years, the majority in smaller-sized patient populations. Depending on the type of cancer investigated, the clinical setting, and the choice of PET radiotracer, these studies have either reported that PET/CT and PET/MRI perform equally well [2‐13], or that PET/MRI has minor to moderate advantages [14‐21]. It is questionable whether the latter results justify the use of PET/MRI in a routine setting, because the costs for purchase, installation, and maintenance of a PET/MRI system exceed those of PET/CT, and because the scan duration is typically longer with PET/MRI, which leads to a lower patient throughput [1]. In addition, clinical implications—such as changes in treatment strategy—of using PET/MRI instead of PET/CT have been documented in only a few studies [7, 17‐19, 22].

The aim of this study was therefore to (1) prospectively investigate the clinical impact of PET/MRI, compared with PET/CT, in a mixed population of cancer patients, and to (2) perform an economic evaluation of PET/MRI through comparison with PET/CT, using clinically oriented cost-effectiveness analyses.

The results of our prospective study suggest that PET/MRI provides additional clinical value in terms of changes to more appropriate management in 8% of cancer patients who undergo PET/CT in routine clinical practice. This percentage is lower than that in the largest study thus far: with regard to clinical impact, Catalano et al. reported a superiority of PET/MRI over PET/CT for 16% of cancer patients in a retrospective analysis [17]. This discrepancy may be explained by the differences in relative frequencies of cancer types between our study and theirs (e.g., NSCLC, 17% vs. 7%; melanoma, 10% vs. 3%; breast cancer, 2% vs. 26% of the entire cohorts, respectively). Unlike Catalano et al., we also used PET radiotracers other than [¹⁸F]FDG, namely, [⁶⁸Ga]Ga-DOTANOC for well-differentiated NETs. Notably, patients with well-differentiated NETs, for which [⁶⁸Ga]Ga-DOTA-peptide-PET/CT is the technique of choice [24, 25], and which accounted for 13% of our population, were not included by Catalano et al. [17]. While our results clearly do not justify a general replacement of PET/CT with PET/MRI, they do suggest that certain subgroups—such as advanced-stage NSCLC and melanoma (see Table 3)—could benefit from undergoing PET/MRI instead of PET/CT.

In our study, the overall superiority of PET/MRI over PET/CT in terms of diagnostic accuracy (+ 13%) was mainly due to the superior performance of PET/MRI for the detection of brain and liver metastases (see Table 3, Figs. 1 and 2), which has been documented in previous smaller-sized studies [2, 8, 10, 11]. Similarly, the superiority of PET/CT for the detection of lung lesions—another previously reported finding [2, 26‐28]—was also confirmed in our study. While the additional brain and liver metastases detected exclusively by the MRI component of PET/MRI had implications for management in 19/21 patients (see Table 3), the additional lung lesions detected exclusively by the CT component of PET/CT, but not by PET/MRI, did not have implications for management in any patient.

Importantly, in NSCLC, it has been shown that treatment of early brain metastases, while still asymptomatic, is associated with better control of neurologic manifestations and longer survival [29], and hence, the American College of Chest Physicians recommends cranial MRI (preferred over cranial CT) for routine imaging of clinical stage III–IV NSCLC patients [30], a strategy comparable with that proposed by the European Society of Medical Oncology [31, 32]. Similar recommendations on the use of cranial MRI to detect brain metastases exist for melanoma in stages III–IV, such as the German S3 Guideline (https://​www.​leitlinienprogra​mmonkologie.​de/​fileadmin/​user_​upload/​Downloads/​Leitlinien/​Melanom/​Melanom_​Version_​3/​LL_​Melanom_​Langversion_​3.​1.​pdf). A closer look at the patients in whom PET/MRI led to a management change in our study reveals that, in 9/10 NSCLC patients (all stage III or IV), this change was due to detection of brain metastases on the MRI component. Contrary, in melanoma (five patients, all stage III or IV), reasons for management changes in our study were more balanced: brain metastases in 3/5 patients and liver metastases in 2/5 patients. In pancreatic adenocarcinoma and NETs (3 patients each), and colorectal cancer (one patient), management changes were exclusively due to liver metastases on the MRI component of PET/MRI. Notably, in the single patient with cervical cancer (referred as stage II) in whom PET/MRI lead to a management change, the higher soft-tissue contrast provided by morphological MRI enabled correct locoregional staging, ruling out bladder infiltration (a criterion for stage IV disease). Although more data are needed to confirm these findings, they nevertheless suggest that MRI protocols in the setting of PET/MRI may need to focus on different anatomic sites in different types of cancer, e.g., the brains in NSCLC and melanoma, the liver in melanoma, colorectal and pancreatic cancer, and NETs.

In the above scenarios, the use of PET/MRI instead of PET/CT obviates the need to perform additional, single-region MRI. Such considerations must be considered when looking at our per-examination costs, which show that PET/MRI is almost 50% more expensive than PET/CT. A clinically indicated addition of MRI to PET/CT may, depending on the body region and MRI protocol, result in a similar, or even higher overall cost than for PET/MRI, with the possible disadvantage of a prolonged time interval until treatment initiation. Furthermore, in our study, PET/MRI prevented inappropriate surgery in several cases (e.g., patient nos. 27, 106, 115, and 127 in Table 3), the cost for which exceeds the cost difference between PET/MRI and PET/CT. In one melanoma patient, treatment failure with ipilimumab was detected only by PET/MRI (patient no. 212 in Table 3 and Fig. 2), which was then discontinued and replaced by pembrolizumab—here, the cost for an inappropriate, additional cycle of ipilimumab (10,000–15,000 EUR, depending on body weight) would have exceeded the cost difference between PET/CT and PET/MRI by far.

Our study is limited by its monocentric design, which also affected our sample size. However, with 263 patients (330 examinations), we prospectively evaluated about twice as many patients as the largest study on this topic so far (Catalano et al., with 134 retrospectively included patients and 134 examinations) [17]. We enrolled only patients scheduled to undergo PET/CT for routine purposes, and thus, our sample reflects standard clinical care, in terms of cancer types and radiotracers, without any relevant selection bias. However, this strategy prevented us from including patients with untreated prostate cancer, who, in our institution, undergo [⁶⁸Ga]Ga-PSMA-11-PET/MRI rather than PET/CT, and for whom participation in our study would have meant a purely study-related radiation exposure. Consequently, we cannot exclude the possibility that inclusion of prostate cancer patients would have influenced our study results—in view of the current literature, probably in favor of PET/MRI [21, 33, 34]. Furthermore, similar to the majority of studies on this topic [2, 8, 11, 17], we used a composite reference standard that relied chiefly on follow-up imaging, and, in a smaller number of patients, on biopsies, for verification of additional lesions detected exclusively by either PET/MRI or PET/CT, because it would have been unethical to perform strictly study-related invasive procedures. Our study design prevented us from performing an economic evaluation of changes in patient management due to the use of PET/MRI, as this would require (1) randomization of patients to either PET/MRI or PET/CT, (2) homogeneous cohorts in terms of cancer type(s) and predefined treatment trajectories, and (3) long-term follow-up including assessment of both clinical outcome and quality of life. Cost estimates—and in particular, absolute numbers—reflect the situation at our tertiary care center, and to a certain extent, trends within the country where our institution is located, but may not necessarily be applicable to other countries. Finally, our cost-effectiveness analyses focused on a direct comparison of the two hybrid imaging techniques, PET/MRI and PET/CT, but not on combinations with single-modality techniques, such as PET/CT combined with cranial MRI or PET/MRI combined with chest CT.

In conclusion, the results of our prospective study in a mixed oncologic patient population suggest that the choice of PET/MRI over PET/CT has implications for management in a non-negligible fraction of patients who routinely undergo PET/CT. In particular, patients with NSCLC and melanoma may benefit from PET/MRI, which detects brain and liver metastases that go undetected on PET/CT. Since the cost per PET/MRI examination is almost 50% higher than that of PET/CT, a histology-based triage of patients to either PET/MRI or PET/CT could be meaningful.