Early response evaluation during preoperative chemotherapy for colorectal liver metastases: Combined size and morphology‐based criteria predict pathological response and survival after resection

Abstract Background Short treatment‐duration with early restaging is crucial to avoid liver injury after preoperative chemotherapy (preopCTX) for colorectal liver metastases (CRLM). Response evaluation according to response evaluation criteria in solid tumors (RECIST) criteria implies several limitations. Early tumor shrinkage (ETS; ≥20% size reduction <6‐12 weeks) or morphological criteria (MC) may better predict oncological outcome. Methods In patients undergoing resection after preopCTX between 2003–2017 pathological and radiological response was reassessed according to Blazer classification, ETS, MC, and RECIST within 90 days and correlated with survival. Results Seventy‐two patients were included, with a median of two (1‐10) liver lesions, 53% bilobar involvement, and 7% extrahepatic disease. PreopCTX was applied for 3 months in median (1‐6). During restaging after a median of 62 days, presence of ETS was associated with improved median overall survival (OS; 57.1 vs 33.7 months; P = .010) and disease‐free survival (16 vs 7.2 months; P = .025). MC significantly correlated with major pathological response (P = .021). When combining ETS with optimal MC, presence of one or both factors was associated with pathological response (61.5% and 92.3%; P = .044) and OS in log‐rank (P = .011), and multivariable analysis (hazard ratio [HR] 0.41; 95% confidence interval [CI], 0.19‐0.90 and HR 0.32; 95%CI, 0.11‐0.97). Conclusion Response‐grading by combined ETS/MC criteria less than 90 days after preopCTX initiation predicts pathological response and postoperative survival in CRLM.


| INTRODUCTION
The liver represents the most frequent site of metastases in colorectal cancer (CRC), affecting about 15% to 20% of newly diagnosed CRC cases simultaneously, 1,2 and 20% to 50% of patients in the further course of the disease. 3 Surgery and ablation remain the major options when aiming for curation. [4][5][6] Due to utilization of progressive oncosurgical techniques resection rates of metastatic CRC (mCRC) within multimodal concepts currently approach 50% in specialized centers. 1,7 Hereby, preoperative combination chemotherapy (preopCTX) plays a key role in downsizing advanced colorectal liver metastases (CRLM) to achieve resectability while preserving sufficient future liver volume. [8][9][10][11][12] Evidence moreover suggests a benefit of preopCTX in a subgroup of patients with primary resectable disease but unfavorable tumor characteristics. 10,[13][14][15] Prolonged preopCTX is associated with considerable toxicity and increases postoperative morbidity and mortality through chemotherapy-associated liver injury. 16,17 Hence, restricting preoperative treatment to 3 months or less is recommended by guidelines 10 to avoid complications, therefore early radiological response assessment is crucial.
Traditionally, response is classified at the time point of maximum tumor reduction according to the response evaluation criteria in solid tumors (RECIST). 18 Despite advantages in terms of objective standardized assessment, this classification implies several limitations. 16,19 Besides the prerequisite of inevitable radiological expertize and time-consuming application, the RECIST stable disease (SD) subgroup is criticized for including both cases with minor tumor shrinkage (0%-29% size reduction) and minor tumor progression (1%-19 size increase). Also, patients with complete response (CR) or progressive disease (PD) are rarely represented in typical surgical CRLM cohorts, limiting reasonable risk stratification in clinical practice. To facilitate early restaging, a number of recent trials investigating first-line CTX in mCRC have evaluated early tumor shrinkage (ETS) as a simple marker for treatment guidance after first restaging. While no consensus definition exists so far, ETS is commonly assessed within 6 to 12 weeks after CTX initiation and defined as a minimum of ≥20% to 30% size reduction in diameters of target lesions. [19][20][21][22][23][24][25] ETS has been associated with improved progression-free survival (PFS) and overall survival (OS) in these studies. Since resectability rates in first-line CTX trials usually range below 15%, and ETS has not yet been validated in a purely surgical cohort, applicability of these results in CRLM patients receiving CTX in a potentially preoperative setting is indeterminate.
Furthermore, a previous single center study showed, that response assessment according to morphological criteria (MC) with evaluation of changes in radiological CRLM appearance (tumor density, tumor-liver border) better predicts histological viability and prognosis after liver resection than size-based criteria like RECIST. 26,27 However, response was evaluated after a median of six (up to 24) cycles of CTX, thus transferability to a modern setting of short preoperative treatment with earliest possible restaging remains unclear.
The present study aims to investigate the prognostic value of early response assessment by ETS and MC within 90 days after preopCTX initiation in CRLM patients undergoing liver resection.

| Patient cohort and data
Patients undergoing curative-intent liver resection for newly diagnosed CRLM at Medical University of Innsbruck between 2003 and 2017 were reviewed from our database ( Figure 1) 28,29 OS was defined as time from metastasectomy to death or to last follow-up with known alive status, derived from patient records and cross-checked with national survival data. 30 Disease-free survival (DFS) was defined as time to occurrence of any relapse or to last follow-up date of known tumor-free status.   Figure   2A). Accordingly, metastases were allocated to MC group 3 (heterogeneous with ill-defined border) or MC group 1 (homogeneous with sharp border) or MC group 2 (mixed appearance).

| Radiological assessment
During restaging, the same lesions were again assessed according to these criteria and response was rated as optimal MC response when previous MC group 3 or 2 lesions had changed to group 1, suboptimal MC response when MC group 3 lesions had changed to group 2 or no MC response when the restaging group was identical or even increased ( Figure 2B).
Computed tomography were performed with intravenous contrast media evaluated in arterial phase with slice thickness ≤5 mm. A maximum two target liver lesions and one additional lesion per organ in case of extrahepatic disease was assessed. In cases with mixed response, the lesion with worst response was recorded for overall grading. For diverging gradings recorded by the two radiologists, consensus was achieved through individual case discussion. RECIST groups included CR, partial response (PR), SD, or PD. ETS was divided into presence or absence of ETS. Combination of ETS and optimal MC was applied as exemplarily shown in Figure 2C.

| Pathological response
All surgical specimens were re-evaluated by an expert oncopathologist (S.S.) for extent of response according to the classification described by Blazer. 31 Briefly, this assesses the percentage of viable tumor cells and grades into three groups: CR (no viable tumor cells), major response (1%-49% viability) and minor response (≥50% viability).

| Statistical analysis
Nominal variables were reported as frequencies/percentages, continuous variables as medians with standard deviation or range.
Differences between groups were analyzed by χ 2 or Fisher's exact

test. Correlation between nominal variables was assessed with
Cramer's V, whereby a level of more than 0.250 was considered indicating a strong relationship. Kaplan-Meier survival curve estimations were plotted with the log-rank test for subgroup comparison.
Cox-proportional hazards regression analysis (enter method) was performed to assess associations of clinical factors with OS. All F I G U R E 2 Grading according to the morphological criteria (MC) classification A, Liver target lesions are grouped at baseline and during restaging according to the MC. B, Response is evaluated according to the change between groups. C, Combination of optimal MC (lesions with homogeneous attenuation and sharp tumor-liver interface) and early tumor shrinkage (ETS; ≥20% size reduction of target lesions). Left: during restaging, neither optimal MC response nor ETS is present. Middle: optimal MC but no sufficient size reduction (<20%). Right: both optimal MC and size reduction ≥20% are present [Color figure can be viewed at wileyonlinelibrary.com] calculations were performed with SPSS Version 21 (IBM Inc, Armonk, NY), a two-tailed P value of <.05 was considered significant.

| Radiological and pathological response
The median time from CTX start to first restaging was 62 days (SD 16.5). Results of radiological assessment at first restaging according to the different classifications are reported in Table 2. ETS defined by ≥20% tumor shrinkage was present in 70.8%, suboptimal or optimal MC response in 19.4% and 26.4%, respectively. Evaluation according to the RECIST criteria resulted in only few patients classified as PD (n = 2; 2.8%) or CR (n = 3; 4.2%) compared to 37.5% and 55.6% SD or PR cases, respectively. Neither the type of CTX nor additional biologicals were associated with presence of ETS or optimal MC response MC (all P > .1).
Pathological response was determinable in 68 patients (5.6% missing) with grading according to Blazer as follows: minor response n = 24 (35.3%), major response n = 39 (57.4%), CR n = 5 (6.9%).  To evaluate the value of combined size and morphology-based response assessment, we further consolidated both the ETS and grouped-MC criteria. As depicted in Figure 3C, this resulted in three subgroups of patients with reasonable case numbers and significantly different outcome. The median OS was 60.9 months

| Correlation of radiological and pathological response
Analyzing the association between radiological and pathological response ( Figure 5) Cramer's V = 0.172). 0.41 and 0.28; P = .016 and P = .012, respectively). In contrast, nodal positivity and RAS mutation of the primary tumor (both HR 2.24; P = .035) were linked to worse OS, as was the need for major liver resection (HR 1.99; P = .034). All of these factors except nodal positivity remained statistically significant in the multivariable regression model. histopathologists blinded to oncological outcomes. We first determined, that early imaging assessment within 90 days according to ETS provides significant risk stratification regarding OS and DFS after liver surgery. ETS has previously already been established as a valuable factor to assess early response and predict PFS and OS in oncological first-line CTX trials. 19 However, in these studies, only a limited number of patients finally underwent liver resection and no specific surgical cohort sub-analysis has been published.

| Multivariable analysis of factors associated with OS
Moreover, our results support findings of others, that morphology-based criteria such as the MC are more precisely correlating with pathological response than size-based criteria such as RECIST or ETS. 26 The limitations of this study primarily include its retrospective nature and potential inclusion bias. A prospective trial including all patients undergoing chemotherapy for advanced but potentially ever resectable CRLM would provide a more complete picture in this challenging cohort. Currently ongoing large-scale projects on complex CRLM surgery such as the EORTC-ESSO 1409/CLIMB registry 32 will hopefully also provide insights in these specific issues. Although first-line chemotherapy trials for mCRC have already established ETS as a prognostic marker, the low resection rates in these studies do not reflect the typical surgical cohort of specialized hepatobiliary surgeons. We therefore aimed to address the following clinical scenario: In patients with potentially resectable, borderlineresectable or primary unresectable CRLM deemed to undergo preopCTX, which classification should be used by hepatobiliary surgeons to assess radiological response early within 2 to 3 months of initiation of chemotherapy to predict pathological response and postoperative prognosis? In this regard, our study represents the first of its kind, with statistically sufficient sample size and a methodology that attempts to limit the restrictions of its' retrospective design.
However, applicability of the combined ETS/MC classification in other centers or geographical regions (eg Asia) or in patients undergoing modern triple CTX (eg FOLFOXIRI 8,33 ; only 7% in our cohort) justifies further investigation. Therefore, prospective, international validation, ideally with centralized radiological and pathological assessment would be desirable in the future.

| CONCLUSIONS
This study presents a novel combination of established imaging criteria to assess early response during short-term preopCTX for advanced CRLM. The combined ETS/MC classification correlates with pathological response and OS, facilitates practical clinical applicability and fosters risk stratification to guide treatment by liver surgeons.