Added value of quantitative, multiparametric 18F-FDG PET/MRI in the locoregional staging of rectal cancer

With over 1.9 million new cases and 935,000 deaths in 2020, colorectal cancer is the third most common cancer and second leading cause of cancer-related death worldwide, with rectal cancer (RC) accounting for about one-third of patients and associated deaths [1]. Overall declines in colorectal cancer incidence mask the increasing rates of early-onset colorectal cancer among patients under 50 years of age, largely driven by RC [2].

Accurate primary locoregional staging in RC is pivotal for optimal multidisciplinary disease management, as well as prognostication. Initial disease stage is the most important predictive factor for overall survival (OS) since 5-year survival rate decreases from 91% in stage I to 72% in locally advanced rectal cancers (LARC; T ≥ 3, N +), and drastically to 15% in stage IV RC [3, 4]. In the last few decades, advances in surgical techniques, including standardized total mesorectal excision (TME), chemotherapy, radiotherapy (RT) and chemoradiotherapy (CRT), as well as a better understanding of surgical oncology, have led to a significant reduction in locoregional recurrence rates and longer OS [5]. In LARC, the standard of care consists of neoadjuvant CRT followed by surgery, which has improved control of local disease with local recurrence rates of 5–7% [6, 7]. However, recurrence of distant disease remains high. Recently, the results of the RAPIDO and PRODIGE-23 phase III clinical trials using total neoadjuvant chemo(radio-)therapy in LARC have demonstrated promising results, improving disease free survival and overall survival [8, 9]. Of note, the majority of oncologic therapy in those regimens (i.e., CRT) is administered before any surgical histology becomes available. Thus, the treatment decision heavily relies on accurate staging information and not post-resection histopathology. In addition, precise staging has an economic impact since treatment costs increase exponentially from early to late stages [10, 11].

According to the European Society of Gastrointestinal and Abdominal Radiology, as well as the National Comprehensive Cancer Network, pelvic MRI remains the gold standard for locoregional staging of RC [12, 13]. MRI delivers anatomical tumor information with depiction of T-staging, involvement of the mesorectal fascia (MRF), extramural vascular invasion (EMVI), and locoregional lymph node metastases [14]. However, despite being superior to other imaging modalities, the ability of MRI to predict the exact tumor stage remains suboptimal, primarily due to poor detection of metastatic lymph nodes and differentiation between T2 and T3 tumors [15].

Positron emission tomography/magnetic resonance imaging (PET/MRI) enables the simultaneous acquisition of metabolic information from PET along with a superb MRI anatomic layout and functional data from dynamic contrast-enhanced MRI and DWI, reflecting the pathophysiologic characteristics of the tumor [16]. Furthermore, PET/MRI has shown its ability to predict even more complex tumor features such as tumor phenotype and circulating miRNA [17, 18]. Where available, it has emerged as a “one-stop shop” examination, not only for colorectal cancer, but also for various other oncologic diseases [19‐23]. However, sensitivity of small lung nodule detection remains lower than in standard of care imaging [24]. Particularly for local staging of RC, PET/MRI, due to the higher soft-tissue signal and contrast-to-noise ratios of MRI, has been shown to outperform both positron emission tomography/computed tomography (PET/CT) and MRI alone [3, 25, 26]. Nevertheless, the exact impact on the accuracy of locoregional staging remains unclear, since, in most studies, many patients with RC have received neoadjuvant CRT, thus altering the T, N, and MRF status in the resected specimen compared to treatment-naïve tissue.

Therefore, the aim of this study was to determine whether locoregional staging and tumor assessment can be improved by multiparametric PET/MRI compared to MRI alone, in patients with early and intermediate-stage RC, who underwent resection without neoadjuvant CRT. Moreover, the prognostic value of mpPET/MRI parameters on long-term survival was evaluated.

Patients with histopathologically confirmed, untreated adenocarcinoma of the rectum who underwent routine clinical 18F-fluorodeoxyglucose (FDG) PET/MRI for staging of primary RC between July 2015 and December 2020 were potentially eligible for this study. This study was approved by the local institutional review board of the Medical University of Vienna (IRB-No. 1403/2015) and performed in accordance with the Declaration of Helsinki. Written, informed consent was waived for this retrospective analysis. Low- to intermediate-risk patients, according to current European Society of Medical Oncology guidelines [27] (T < 3 for low tumors, up to T3ab for middle or high tumors, N0, but N1 allowed if upper-third tumors, MRF not involved, no EMVI), who had been treated with surgical resection without neoadjuvant chemotherapy or chemoradiotherapy, were selected via a hospital information system database search. Treatment decisions were based on routine multidisciplinary tumor boards after reviewing all available imaging material at the time of diagnosis. Exclusion criteria were (i) age < 18 years, (ii) pretreatment with chemotherapy or chemoradiotherapy, (iii) surgical resection not performed, (iv) poor examination quality, and (v) blood glucose level > 150 mg/dl. From an initial cohort of 119 patients who had undergone PET/MRI for primary RC, 57 patients were excluded due to neoadjuvant treatment, six patients did not undergo surgery, five studies had poor image quality or incomplete studies, and, in five patients, histopathological data was incomplete. The final study group consisted of 46 patients who fulfilled all the criteria.

The goal of this retrospective study was to analyze the value of multiparametric PET/MRI in the staging of primary RC without preoperative treatment to gain histologically correlated metabolic and functional information that could potentially improve locoregional staging, and thus, influence treatment stratification. The results of our study demonstrate an association between quantitative PET parameters and unfavorable tumor features, leading to an improved identification of locally advanced tumors by the combination of PET and MRI, possibly influencing treatment stratification and providing predictive information.

The treatment of non-metastatic rectal cancer is heavily driven by the local tumor stage and adverse locoregional tumor features, which tend to increase the likelihood of local and distant recurrence. However, major limitations of the current imaging gold standard, MRI, include poor performance in differentiation between T2 and T3ab stage, mainly due to the desmoplastic reaction of the adjacent mesorectal fat, with reported sensitivities and specificities of 43–72% and 76–94%, respectively [25, 28], which concurs with our results. Although very similar 5-year survival rates in pT3a and pT2 have been reported [29], understaging may lead to omission of RT or CRT and an increase of local relapse, while overstaging may lead to unnecessary treatment and functional consequences. Another major limitation of MRI is its performance in the detection of positive lymph nodes with reported sensitivities of 56–85% and specificities of 54–94%, respectively [30, 31]. Although the combination with morphologic criteria has improved the diagnostic performance [31], it is still subpar given how important treatment decisions, such as total neoadjuvant therapy, rely on its accuracy.

Functional MRI imaging techniques, namely DWI, may contribute valuable, additional information in rectal cancer imaging, particularly regarding therapy response assessment. While DWI is recommended by radiological guidelines [13] due to its proven benefit for tumor detection, as well as evaluation of treatment response, it appears to play only a minor role in the primary staging of rectal cancer. However, quantification of diffusion properties and expressing these properties as an apparent diffusion coefficient could potentially be used as imaging biomarkers of tumor aggressiveness. Although not statistically significant, lower ADC values were associated with poorly differentiated tumors in our study. A recent study shows a similar trend toward low ADC values in high-grade tumors [32].

In previous studies, mainly relying on PET/CT, the PET component often failed to provide additional locoregional staging information. In our quantitative analysis of the PET data, MTV was independently associated with higher T-stage, as well as with LARC, similar to what is known from other tumor entities, such as cervical or esophageal cancer [33, 34]. Compared to MRI staging, MTV yielded an excellent sensitivity of 91.7% and specificity of 68.2% for LARC, significantly superior to anatomical MRI assessment. The MTV value might be used as an additional tumor feature to support the differentiation between borderline T2 and early T3 tumors, as demonstrated in our reported performance Figs. 3, 4, and 5. However, an independent confirmatory study in a different, prospective, patient cohort would be required before any clinical recommendation can be made. MTV also demonstrated an association with overall survival, indicating potential prognostic value, as has been proposed in previous studies [35]. This might contribute to the diagnostic value of volumetric PET parameters in the future. SUV_max, SUV_peak, and SUV_mean did not show any independent association with tumor stage, which reflects current data from PET/CT studies [36]. Furthermore, the early assessment of FDG uptake (30 vs. 60 min) did not show any benefit in gaining information on staging or prognosis compared to the standard acquisition.

The combined assessment of metabolic and functional imaging using PET/MRI is gaining attention in the evaluation of RC as a robust hybrid imaging technology because of its ability to combine high soft-tissue anatomic depictions with PET data in a single study. Queiroz et al. [37] evaluated PET and MRI parameters for the prediction of synchronous metastases in RC patients. In their study, the PET volumetric features TLG (352.9 vs 242.7, P = 0.046) and MTV (36.1 vs 26.2 mL, P = 0.03) significantly differed in patients with and without distant metastases and, thus, demonstrated higher values in patients with advanced tumors, while SUV_max and SUV_mean did not facilitate differentiation, strongly corresponding to our findings.

Other recent studies have mainly focused on comparison of PET/MRI to standard-of-care imaging and PET/CT, as well as the consequent impact on patient management. Only a few studies have investigated the diagnostic value of locoregional PET/MRI in the staging of untreated and treated patient cohorts. One study by Catalano et al. [25] found potential superior diagnostic accuracy of PET/MRI in the primary staging of untreated LARC, with PET/MRI outperforming MRI for evaluation of tumor length, external sphincter infiltration, and lymph node involvement, with a sensitivity of 92% and specificity of 86% vs. 88% and 43% for MRI, respectively. A composite reference standard of clinical, imaging follow-up, and histopathology was used as the gold standard for staging. Another study investigated the benefit of PET/MRI in the staging and restaging of RC, but did not demonstrate a significant improvement in PET/MRI versus MRI diagnostic accuracy for locoregional T- and N-staging compared to MRI [38]. However, in that study, a mix of treated and untreated patients rendered the reference standard of histopathology difficult to assess. In most diagnostic studies that include advanced RC, the reference standard is quite heterogenous, since it mostly consists of a combination of endoscopic biopsies, clinical and imaging follow-up, or surgical specimens after neoadjuvant CRT.

To the best of our knowledge, there has been no study investigating the diagnostic value of PET/MRI in the primary staging of untreated RC in a homogeneous patient cohort using the surgical specimen as a reference standard in each patient. In the authors’ opinion, this is the most reliable way to properly assess the true locoregional tumor stage and compare it to imaging performance.