Colorectal cancer (CRC) is the third most common type of cancer worldwide, with estimated 1.8 million new cases and 0.91 million deaths in 2020 [1]. Approximately 25–30% of CRC patients develop colorectal liver metastases (CRLM) during the course of the disease [2, 3]. Hepatectomy is currently the most effective and potentially curative treatment for patients with CRLM, with a 5-year survival rate of 40%–50% [4–7]. In recent years, preoperative chemotherapy has been widely used to treat CRLM, not only to downsize the lesions, ensuring their surgical resectability and the control of micrometastases [8, 9] but also to prolong progression-free survival [10]. Unfortunately, more than half of patients still develop postoperative recurrence within 2 years [11].

Many clinical factors have been used to predict the survival of patients with CRLM after hepatectomy, including the tumor size and number of lesions [8, 12–14], carcinoembryonic antigen (CEA) level [12, 15], node-positive primary disease [12, 16], clinical risk score (CRS) [17–19], and resection margin involvement [16, 20]. With the development of preoperative chemotherapy, new clinicopathological factors have emerged. Pathologic complete response (PCR), which is defined as the absence of any residual cancer cells, has been proven to confer an extremely favorable response to chemotherapy and can serve as an indicator of good prognosis [21–27]. However, its applicability to CRLM is limited because only approximately 5% of these patients achieve all-lesion PCR, with the vast majority achieving only partial-lesion PCR [23, 28]. Among the patients who achieve partial PCR, the number of resected liver metastases required to achieve PCR varies, and little research has been conducted to evaluate the relationship between the number of liver metastases exhibiting PCR and the prognosis of patients.

In this study, we applied a novel method to evaluate the pathologic response of CRLM patients. We defined the pathologic complete response ratio of liver metastases (PCRRLM) as the number of liver metastases exhibiting PCR divided by the number of total resected liver metastases. Then, we explored the value of the PCRRLM as an indicator of recurrence after preoperative chemotherapy and the need for liver resection in patients with CRLM.

In this study, we retrospectively analyzed the accumulated cases of CRLM in our hospital in an attempt to identify potential prognostic indicators for these patients. We proposed the PCRRLM, a new and potentially better measurement of pathologic response in patients with CRLM after preoperative chemotherapy and resection, and assessed its value for predicting prognosis and recurrence. Detailed analysis confirmed that patients with a PCRRLM ≥0.50 had a highly favorable prognosis and a significantly lower incidence of recurrence, suggesting that PCRRLM could be a feasible clinicopathological indicator.

Although the strength of the pathologic response after preoperative chemotherapy has generally been accepted as a prognostic factor over the years, the optimal grading system to evaluate this parameter is still a matter of debate. A conventional assessment was semi-quantitatively estimating the proportion of residual cancer cells in relation to the total tumor area, and the mean of the values was used when there was more than one metastasis [21]. Some other researches only considered patients who had achieved full PCR in all tumors [23, 28]. However, in clinical practice, most patients had multiple liver metastases, among which only a small percentage achieved full PCR in all tumors. Practically, the number of liver metastases was the most frequently included prognostic factor in all studies of CRLM [34]. Moreover, achieving PCR in all metastases proved not to be necessary, while at least one tumor showed PCR was also beneficial to prognosis [35]. The conventional method did not consider the number of metastases, and the applicability was limited due to the low odds of PCR in all tumors. In contrast, PCRRLM carefully taken into account the number and heterogeneity among multiple metastases, allowing for better characterization of pathologic response for patients with multiple CRLMs.

Another popular index for pathologic response was tumor regression grade (TRG), a 5-point scoring system firstly proposed for esophageal tumor and then introduced into the field of colorectal cancer by Dworak et al. [36] TRG is semi-quantitatively determined by the amount of residual tumor cells versus the amount of fibrosis, with a TRG of 4 indicating no evidence of any treatment effect and a TRG of 0 indicating a complete response with no viable tumor remaining. The interpretation of TRG in patients with multiple liver metastases is flawed because these patients are mostly categorized according to the worst grade (namely, the highest TRG), which may not reflect the heterogeneity observed among metastatic lesions. As has been documented in many studies[8, 12–14], the number of metastases has a significant impact on patient outcome. TRG essentially ignores the regression status of all but the worst tumor. By comparison, our methods take into account the regression status of all resected tumors from a holistic point of view. Moreover, because TRG is a semiquantitative measurement, it is inevitably influenced by the subjectivity of pathologists. The prognostic value of TRG can hardly be observed across each tier (from TRG 4 to 0); hence, a variety of complex grouping strategies have been applied by researchers, all of which have lacked rigor [37–39] (e.g., TRG 4 v 2+3 v 0+1, TRG 4+3 v 2 v 0+1, or TRG 4+3 v 2+1+0). In contrast, our methods avoid this subjectivity and grouping complexity to some extent and proved to be an excellent predictor for OS and RFS. Thus, the PCRRLM has the potential to be a better portrayal of pathologic response after chemotherapy.

One of the inherent drawbacks of applying the PCRRLM in clinical practice is the difficulty of assessing it in a noninvasive manner prior to surgery. Therefore, it would be considerably valuable to preoperatively identify patients who are likely to achieve a favorable pathologic response. In this study, we determined that age <55 years old and preoperative CEA ≤5 ng/ml are strong predictors of PCRRLM ≥0.50. In addition, a primary tumor site in the rectum, CRLM diameter of <3 cm, and TBS <5 points were all identified as predictors in the univariate analysis. The correlations among small tumor size, low CEA levels, and PCR have been well confirmed by previous studies [21, 23, 28, 40, 41]. Some researchers [23, 28] used a significantly elevated CEA level (e.g., 20 or 30 ng/ml) as the classification standard, while in our study, the critical value of a normal concentration (5 ng/ml) was applied. Our results further demonstrated the value of a normal preoperative CEA level and small CRLM diameter in predicting a favorable pathologic response, since the probability was 34.2% (25/73) when two factors were present and 6.3% (7/111) when only one was present. Another factor that has been found to be related to the PCRRLM is the primary tumor location. Several studies [42–45] have reported a relatively better response to chemotherapy and favorable survival in patients with metastases from left-side primary tumors. We did not observe a significant difference between patients with left- and right-sided primary tumors, but we found that patients with a primary tumor site in the rectum, which can be classified as left-sided colon in histogenesis [46], were more likely to have PCRRLM ≥0.50. This may be due to the genetic heterogeneity of primary tumors at different sites, as they differ in terms of embryologic development, blood supply, macroscopic pathology, and clinicopathologic parameters [47]. Moreover, we identified a nonnegligible correlation between a low TBS and better PCRRLM. TBS was first proposed by Sasaki et al. [31] by combining the size and number of CRLMs in a continuous rather than dichotomous stratification and was proven to have a higher prognostic value than the traditional CRS system. Ruzzenente et al. [48] found that a reduction in the TBS indicated a response to preoperative chemotherapy and could serve as a predictor of survival. Sasaki et al. [49] further validated that imaging-based TBS had the same strength as pathology-based TBS in predicting OS preoperatively. Considering the potential of a low TBS and high PCRRLM, our results further demonstrate the significant correlation between the TBS and pathologic response and provide evidence of a method to predict pathologic response by calculating the TBS using preoperative radiologic cross-sectional imaging, which may make it unnecessary to wait for the results from resected tumors after surgery.

The long-term prognosis was much better for patients who had a higher PCRRLM than for those who did not. In our study, the proportion of patients who achieved pathologic complete response (PCR) was 4.9% (15/305), which is similar to previous studies [21, 23, 28]. Owing to this low incidence, the application of PCR in clinical practice is limited, and few changes in patients may bias the results. Considering that PCR can improve the prognosis of patients with only one metastatic liver lesion [35], we relaxed the inclusion criteria and categorized patients into two groups: the high-PCRRLM group (≥0.50) and the low-PCRRLM group (<0.50). The new criteria categorized 14.4% (44/305) of patients into the high group and 85.6% (261/305) into the low group. After validation, a higher PCRRLM was found to serve as a good predictor of favorable survival, with a remarkable 5-year OS rate of 62.75% in the high group and 41.06% in the low group (p = 0.045). For recurrence-free survival, the 5-year RFS rate in the high-risk group was 27.18%, whereas that in the low-risk group reached only 16.82% (p = 0.016). The excellent prognosis of the high-PCRRLM group, especially in terms of recurrence-free survival, can be explained by their favorable response to preoperative chemotherapy.

The response to chemotherapy was limited among patients with a PCRRLM less than 0.50, and their prognosis was relatively poor. Hence, for these patients, the follow-up interval should be shortened, as closer clinical monitoring is needed. A low PCRRLM also suggests that the chemotherapy regimen should be modified. Some studies have reported a relatively higher tumor response after oxaliplatin-based preoperative chemotherapy than irinotecan-based preoperative chemotherapy [21, 28, 39, 50], although this correlation was not found in our research. A study conducted in a Chinese cohort [40] also reported a better pathologic response when cetuximab was combined with irinotecan-based chemotherapy. Therefore, a timely assessment of the PCRRLM after hepatectomy can provide clinicians with guidance and a reference for selecting postoperative chemotherapy, contributing to a better therapeutic effect.

This study had some limitations, which should be acknowledged. First, this was a retrospective analysis that included only a limited number of patients from a single institution. Incorporating the grouped chemotherapy and targeted therapy regimens into the analyses would make it difficult to obtain good reproducible results, as this was a small cohort study. Therefore, the findings need further validation in a larger prospective cohort of patients. Second, the mutation status of some genes, such as RAS and BRAF, have been confirmed as significant prognostic factors and widely applied in clinical practice [51, 52], but we did not include them in the analysis because these data were unavailable for the majority of patients in this study. We also did not incorporate the TRG index in this study. For a large number of liver metastases, there are many possibilities of TRG, and there is no clear standard for which TRG should be selected as the final TRG. It is inappropriate to use single-tumor TRG measurements for tumor regression, so a comparison between PCRRLM and TRG is not done. Moreover, PCRRLM was not available for patients who had undergone radiofrequency ablation. Prospective randomized trials are needed in the future to further verify the reliability and validity of the PCRRLM.

In conclusion, we proposed the PCRRLM as a novel grading system for the pathologic response of patients after the resection of CRLM treated with preoperative chemotherapy. The long-term RFS of patients who achieved a PCRRLM ≥0.50 was extremely favorable, and most of them showed a significantly lower tendency toward recurrence. The PCRRLM may provide an objective and precise evaluation of the pathologic response of patients with CRLM.