Morphological response and tumor shrinkage as predictive factors in metastatic colorectal cancer treated with first-line capecitabine, oxaliplatin, and bevacizumab

Morphologic response (MR) is a novel chemotherapeutic efficacy predictor of solid tumors, especially those treated with anti-vascular endothelial growth factor antibodies. Nevertheless, the importance of systemic chemotherapy MR for colorectal liver metastases (CLM) remains unclear. We aimed to evaluate the usefulness of MR as a factor associated with the therapeutic effects of chemotherapy plus bevacizumab for initially unresectable CLM cases. We retrospectively evaluated the associations between MR and/or Response Evaluation Criteria in Solid Tumors (RECIST), progression-free survival (PFS), and overall survival (OS) in patients who received first-line capecitabine, oxaliplatin, and bevacizumab treatment for initially unresectable CLM using multivariate analysis. Patients who showed a complete or partial response based on the RECIST, or an optimal response based on MR, were defined as “responders.” Ninety-two patients were examined, including 31 (33%) patients who responded optimally. PFS and OS estimates were comparable in MR responders and non-responders (13.6 vs. 11.6 months, p = 0.47; 26.6 vs. 24.6 months, p = 0.21, respectively). RECIST responders showed better PFS and OS than non-responders (14.8 vs. 8.6 months, p < 0.01; 30.7 vs. 17.8 months, p < 0.01, respectively). The median PFS and OS estimates of MR and RECIST responders were better than those of single responders or non-responders (p < 0.01). Histological type and RECIST response were independently associated with PFS and OS. MR predicts neither PFS nor OS; nevertheless, it may be useful when combined with the RECIST. The Ethics Committee of The Cancer Institute Hospital of JFCR approved this study in 2017 (No. 2017-GA-1123): retrospectively registered.


Introduction
Colorectal cancer is the third most common cancer and the fourth main cause of cancer-related deaths worldwide [1]. The median overall survival (OS) of patients with metastatic colorectal cancer (mCRC), which accounts for approximately 20% of all colorectal cancer cases at initial diagnosis, is approximately 30 months. This estimate is increasing with the use of targeted therapies, including epidermal growth factor receptor (EGFR) and antivascular endothelial growth factor (VEGF) antibodies [2][3][4]. Fluorouracil-based chemotherapy is the first-line therapy for mCRC and includes three standard cytotoxic regimens: FOLFOX (fluorouracil/leucovorin plus oxaliplatin), FOLFIRI (fluorouracil/leucovorin plus irinotecan), and CAPOX (capecitabine plus oxaliplatin). Bevacizumab (BEV) is a monoclonal antibody for VEGF and a molecularly targeted agent used in combination with cytotoxic regimens to treat mCRC. CAPOX plus BEV combination therapy is a standard first-line therapy for mCRC, with a reported median progression-free survival (PFS) and OS of 9.3-12.4 and 21.4-26.7 months, respectively [5][6][7].
The liver is the most common metastatic site in mCRC, and radical resection of liver metastases may improve prognosis [8]. The Response Evaluation Criteria in Solid Tumors (RECIST), which are based on the tumor size, are considered important in such cases; resectability improves when metastases are shrunk by neoadjuvant chemotherapy [9]. Chun et al. reported that metastases decrease in size and undergo unique morphological changes that can be detected by computed tomography (CT) scans, referred to as "morphologic response" (MR) [10]. MR correlates with residual tumor cell counts at liver metastatic sites and may be a more sensitive predictor of therapeutic effect than the RECIST in patients treated with neoadjuvant chemotherapy before liver metastasis resection [10,11]. The frequency of optimal MR (OMR) is high in cases receiving combination therapy with BEV [11,12].
Although some reports have demonstrated the importance of MR in patients who undergo hepatectomy for colorectal liver metastasis (CLM), the utility of MR in patients receiving systemic chemotherapy for mCRC remains unclear. Therefore, we investigated the clinical significance of MR in CAPOX plus BEV combination therapy as first-line treatment for patients with initially unresectable CLM.

Patients
This retrospective study included patients with synchronous initially unresectable CLM who received first-line CAPOX plus BEV combination therapy at our hospital between November 2009 and August 2016 to eliminate patient bias due to differences in treatment regimens. The exclusion criteria were as follows: (1) BEV dose intensity of < 80% during the first four courses; (2) liver metastases of < 15 mm and undetectable by CT scanning; (3) previous treatment for liver metastasis; and (4) appendix, anal canal, or mucinous cancer diagnosis. This study was reviewed and approved by the Institutional Review Board of our hospital (No. 2017-GA-1123) with a waiver of informed consent from patients. The protocol was described on the hospital website, and participants had the opportunity to opt out of the study. All methods adhered to the principles of the 1964 Declaration of Helsinki and its later amendments.

Study endpoints
To assess the clinical utility of MR, we examined the incidence and clinicopathological characteristics of MR and the associations between MR and PFS or OS estimates. We collected data on variables, such as age, sex, Eastern Cooperative Oncology Group Performance Status, primary tumor location, metastatic site, liver metastasis count, primary tumor resection status, histological type, RECIST response, KRAS status, carcinoembryonic antigen (CEA) levels, carbohydrate antigen 19-9 (CA19-9) levels, and alkaline phosphatase (ALP) levels. For comparison with MR, we also analyzed the associations between the RECIST and PFS or OS estimates as well as MR.

Treatment characteristics
CAPOX plus BEV consisted of a 2-h intravenous infusion of oxaliplatin at a dose of 130 mg/m 2 on day 1, plus oral capecitabine at a dose of 1000 mg/m 2 twice daily for 2 weeks in a 3-week cycle. The first capecitabine dose was given in the evening of day 1, and the last dose was given in the morning of day 15. BEV was administered at a dose of 7.5 mg/kg by intravenous infusion for 90 min on day 1 of the first cycle (60 min in subsequent cycles). Indication criteria for CLM resection are based on the Japanese Society for Cancer of the Colon and Rectum guidelines 2019 for the treatment of colorectal cancer [13]. Finally, we decided to develop a treatment plan after a discussion among the members of a multidisciplinary board.

Imaging analysis
Contrast-enhanced CT scanning was performed before starting chemotherapy and every 2-3 months thereafter. Multislice CT scans were obtained using a triphasic liver protocol or a single-phase technique. The morphologic criteria were as follows: Group 1 metastasis was characterized by homogeneous low attenuation with thin, sharply defined tumor-liver interface; Group 3 metastasis was characterized by heterogeneous attenuation and thick, poorly defined tumor-liver interface; Group 2 metastasis was characterized by morphology findings distinct from those observed in Groups 1 and 3. OMR was defined as metastatic change from Groups 3 or 2 to Group 1 classification; incomplete response (IR) was defined as change from Group 3 to Group 2; no-response (NR) was defined as no change or change to a higher group [10,11] (Fig. 1). MR was determined based on morphologic criteria assessed by two certified radiologists blinded to other data; any disagreements were resolved by consensus. According to previous reports, MRs were assigned based on the response observed in the majority of tumors in patients with multiple liver metastases [10]. Furthermore, in our study, if there was an equal number of MR assessments for each liver metastatic lesion, we adopted the best MR assessment for that patient. Response to chemotherapy was also determined based on the RECIST [14]. Patients with OMR defined by the MR criteria, or patients with complete or partial response defined by the RECIST, were defined as "responders"; all other patients were defined as "non-responders."

Statistical analysis
PFS was calculated from the initiation of chemotherapy to the date of confirmed tumor progression or death. The cut-off for patients who underwent hepatectomy was the resection date. OS was calculated from the start of chemotherapy to the date of death or last follow-up examination. Survival curves were drawn using the Kaplan-Meier method, and the log-rank test was used for comparisons. Continuous variables were compared with the Mann-Whitney U test, and between-group comparisons were performed with the χ 2 test and Fisher's exact test. Univariate and multivariate analyses of PFS and OS were performed using the Cox hazard ratio (HR); the multivariate analysis included variables with p-values of < 0.05 in univariate analyses. All analyses were performed using JMP 10 (SAS Institute, Cary, NC, USA), and statistical significance was defined as p-values of < 0.05.

Patient characteristics
Of 316 patients with mCRC, 159 had liver metastases. Of these, 92 patients met the study eligibility criteria. Metastases were present only in the liver in 27% of the patients, and the median number of liver metastases was seven (1-40). Besides the liver, the most common sites of distant metastasis were the lungs (42%), peritoneum (31%), extra-regional lymph nodes (18%), and other sites (9%). The histological type was poorly differentiated adenocarcinoma in 18% of the patients. The primary tumor was resected, and hepatectomy was performed in 62% and 24% of the patients, respectively. KRAS status was "mutant" in 43% of the patients. The median levels of CEA, CA19-9, and ALP were 53 (range: 1-10,944) ng/mL, 138 (2-50,000) U/mL, and 317 (95-91,614) IU/L, respectively. No clinical factor was associated with OMR (Table 1).

PFS and OS estimates
The median PFS was 9.4 (2.0-52.2) months, and the median OS was 24.7 (4.1-100.5) months during the median Fig. 1 Morphologic criteria assessed by contrast-enhanced computed tomography in the late phase. Group 1 metastasis was characterized by homogeneous low attenuation with a thin, sharply defined tumorliver interface. Group 2 metastasis was characterized by morphology that did not meet either Group 1 or Group 3 criteria. Group 3 metastasis was characterized by heterogeneous attenuation and a thick, poorly defined tumor-liver interface. a Optimal response: morphologic change from Group 2 to Group 1 categorization. b Incomplete response: morphologic change from Group 3 to Group 2 categorization follow-up period of 23.7 months. PFS and OS estimates were comparable in MR responders and non-responders (PFS: 13.6 vs. 11.6 months, p = 0.47, OS: 26.6 vs. 24.6 months, p = 0.21, Fig. 2a, b). Meanwhile, RECIST responders had better PFS and OS than non-responders (14.8 vs. 8.6 months, p < 0.01; 30.7 vs. 17.8 months, p < 0.01, respectively; Fig. 2c, d). Regarding the impact of hepatectomy on prognosis, patients who could undergo hepatectomy due to the reduction in liver metastases by systemic chemotherapy had a better OS than other patients (OS: 35.4 vs. 21.6 months, p < 0.01).
As MR alone was not a reliable prognostic predictor in this study, we further investigated the relationship between the combination of the RECIST and MR, and prognosis. Patients who were responders by both sets of criteria (MR and RECIST) were defined as "super-responders." As a  Fig. 3a). The corresponding OS estimates were 45.4, 27.0, 17.8, and 16.5 months, respectively (p < 0.01, Fig. 3b).

Univariate and multivariate analysis of prognostic factors for PFS and OS
In univariate Cox proportional hazard analysis, liver metastasis count, histologic type, RECIST response, and the CA19-9 levels were associated with PFS (Table 2), while primary tumor location, liver metastasis count, histologic type, RECIST response, and ALP levels were associated with OS (Table 3). In multivariate analysis, histologic type and RECIST response were associated with PFS (histologic type: HR = 2.54, p < 0.01; RECIST response: HR = 2.77, p < 0.01, Table 2). Meanwhile, histologic type and RECIST response were associated with OS (histologic type: HR = 2.71, p < 0.01; RECIST response: HR = 2.65, p < 0.01, Table 3). In contrast, MR was not associated with either outcome.

Discussion
In this study, we evaluated the usefulness of MR as a factor associated with the therapeutic effects of chemotherapy plus bevacizumab for initially unresectable CLM cases. Interestingly, we found that MR predicts neither PFS nor OS. Standard chemotherapy for mCRC includes FOLFOX, FOLFIRI, and CAPOX regimens combined with molecularly targeted agents, such as anti-VEGF and anti-EGFR antibodies. On treatment with anti-EGFR antibody cetuximab, early tumor shrinkage (ETS) evaluated by the RECIST (20% reduction in tumor size at approximately 6-8 weeks after chemotherapy initiation) helps predict treatment response [15][16][17]. In addition, the depth of response, defined as the percentage of maximum reduction from the initial tumor size, has been associated with prognosis in mCRC [18,19]. Thus, the size-based evaluation of ETS and depth of response by CT scanning is used in clinical practice to assess  the treatment response of solid tumors to chemotherapy and anti-EGFR antibody cetuximab. While size-based evaluation remains common, morphological change-based assessments attract attention in the context of CLMs treated with anti-VEGF antibodies. Chun et al. first reported that MR in CLM after neoadjuvant chemotherapy was correlated with the number of residual tumor cells in liver metastatic foci, estimated at 20%, 50%, and 75% in the OMR, IR, and NR groups, respectively, and the OMR group achieved favorable OS [10]. Frentzas et al. reported that CLM featured by vessel co-option instead of angiogenesis, was associated with a poor response to the BEV combination chemotherapy and poor morphologic change in patients undergoing hepatectomy [20]. Thus, tumor angiogenesis, the therapeutic effect of anti-VGEF antibodies, and MR are likely closely related. Moreover, Nishioka et al. reported that MR was better than size-based response criteria at predicting the presence of microscopic cancer spread beyond a 1-mm width from the CLM [21]. Shindo et al. found that MR, but not the RECIST, was an independent predictor of OS in patients undergoing hepatectomy for CLM after chemotherapy [11].
Thus, the clinical importance of MR for CLM after hepatectomy has been demonstrated, but only a few reports have focused on MR in patients who had received systemic chemotherapy for mCRC. In our study, MR was not a predictor of either OS or PFS in patients with mCRC, including extrahepatic metastasis, although patients having MR tended to have a better prognosis than those having non-MR. Similarly, Hosokawa et al. reported that MR was not associated with either PFS or OS, while both ETS and the depth of response were significantly associated with prolonged PFS and OS in patients receiving first-line chemotherapy containing BEV for liver-limited mCRC [22]. In contrast, Yoshita et al. reported that MR was a better predictor of PFS than the RECIST in patients with CLM treated with fluorouracilbased chemotherapy with or without BEV as the first-line chemotherapy [23].
When discussing the effectiveness of MR, the differences in cohort background between studies must be considered as this may have a significant impact on prognosis. Liver metastasis is the most common site of distant spread in mCRC, and hepatectomy may contribute to prolonged survival in patients with mCRC [9]. In our study, hepatectomy was performed in 23.9% of the 92 patients, and in 31.8% of the 31 patients with OMR. Regarding initially unresectable CLM, conversion to resectable liver metastases, not by morphologic change but by tumor shrinkage due to systemic chemotherapy, might have had a strong impact on prognosis. Moreover, the inclusion criteria for our study were not limited to liver metastases. In our study, 72.8% of the 92 patients had extrahepatic metastasis, including to the lung, peritoneum, paraaortic lymph node, and others. Although CLM is suitable for assessing MR in terms of frequency and size, it remains unclear whether a similar effect of MR occurs in other metastatic lesions. Although MR was not a reliable predictor in our study, the combination of the RECIST and MR was found to be a better predictor of both PFS and OS than the RECIST alone. To our knowledge, this is the first report to evaluate the relevance of MR and the RECIST for predicting chemotherapeutic effects and prognosis in patients with mCRC treated with CAPOX plus BEV. We hypothesized that patients with both types of response, presented with reduced tumor size (quantity) and reduced residual tumor cell density (quality), accounted for improved outcomes. Although MR can be clinically useful when combined with the RECIST, evaluating it using contrast-enhanced CT images has limitations, such as the lack of objectivity or estimated precision. In our study, the concordance rate between the MR assessment of the two blind radiologists was 85%, which is comparable to that of a previous report [21]. These results suggest that it is difficult to correctly evaluate MR in some cases. As an alternative, a method of predicting the pathological viability of CLM after chemotherapy using fluorine-18-fluorodexyglucose positron emission tomography has been proposed [23]. Furthermore, in patients who used anti-VGEF antibodies for mCRC, MR may be objectively evaluated by examining CT attenuation values of CLM presented by Hounsfield Unit. Therefore, our future task is to find a new method to objectively evaluate MR, including PET-CT and CT values, and to investigate their relationship and the prognosis of patients with mCRC.
Our study had some limitations. First, this was a retrospective study with a small sample size. Second, MR evaluations tend to be subjective, and the associated findings may change depending on the observing radiology specialist. Third, our study included patients who underwent surgery for liver metastasis after chemotherapy; subsequent treatment of such patients may have a major effect on prognosis.

Conclusions
MR does not predict chemotherapeutic efficacy in patients treated with first-line CAPOX plus BEV combination therapy for initially unresectable CLM. Nevertheless, a combination of MR and the RECIST may be a better predictor of OS than the RECIST alone.