Impact of Perioperative Chemotherapy on Survival Outcomes Among Patients with Metastatic Colorectal Cancer to the Liver

Aim: Compare overall survival (OS) between adjuvant and neoadjuvant chemotherapy and analyze the effect of chemotherapy on OS. Materials & methods: National Cancer Database was queried for patients diagnosed with metastatic colorectal adenocarcinoma with isolated liver metastases between 2004 and 2016. We evaluated the OS and chemotherapy effect using Kaplan-Meier estimates and multivariable cox regression analyses. Results: Total 6883 patients with metastatic colorectal cancer and liver metastases were included, of which 6042 patients were treated with surgery and chemotherapy and 841 patients were treated with surgery only. Patients who received neoadjuvant chemotherapy had better OS compared with patients who received adjuvant chemotherapy. Conclusion: Patients with colorectal cancer with isolated liver metastases who were treated with neoadjuvant chemotherapy had better OS compared with adjuvant chemotherapy.


Introduction
Colorectal cancer (CRC) is the third most commonly diagnosed malignancy in the United States and the world and the second leading cause of cancer death in both sexes in 2020 in the United states [1][2][3] .
Because the majority of the intestinal mesenteric drainage enters the hepatic portal venous system, hepatic metastases develop in approximately half of all colorectal cancer cases with nearly one-fth to one-quarter of newly diagnosed metastatic colorectal cancer patients present with liver metastases at the time of primary diagnosis 2,4 . Surgical resection alone is the cornerstone of treatment for patients with liver metastases from colorectal cancer in whom there is at least a one in six chance of cure after hepatectomy and apparent clinical cure is achieved in those who survive 10 years [5][6][7] . In the past 15-20 years, Profound improvements in outcomes of metastatic CRC resulted in more than a doubling of the 5-year survival 4,8,9 . Although new treatment options have doubled overall survival (OS) for advanced disease, survival is still best for those with non-metastatic disease 10 . While the role of surgical resection is established in metastatic disease to the liver, the potential bene t from peri-operative chemotherapy remains unclear 6, 11 . While some studies have shown that perioperative chemotherapy with surgery reduced the risk of progression-free survival events at 3-years in patients with resectable liver metastases 12 , others concluded that preoperative chemotherapy did not prolong survival with colorectal cancer and resectable or marginally resectable liver metastasis 13 . Also, there's a lack of data in assessing the impact of timing of perioperative chemotherapy on survival (pre-versus post-operative), as well as in comparing the effect of single-agent vs multiagent chemotherapy on survival. We studied the survival outcomes of patients with CRC with liver metastases who underwent surgical management alone vs combined surgery and chemotherapy based on a real-world cohort. In addition to that, we assessed the effect of timing of chemotherapy on survival among patients who received chemotherapy and compared the OS between patients who received single-agent and patients who received multiagent chemotherapy as the rst line of chemotherapy

Data source
The National Cancer Database (NCDB) is a nationwide database supported by the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. About 70% of all cancer cases diagnosed in the United States from more than 1,500 cancer centers are included in this database. Details of patients' demographics, malignancy staging, and histological characteristics in addition to treatment and outcome information are provided in this database 14 . The CoC's NCDB and the hospitals participating in the CoC's NCDB are the sources of the de-identi ed data used herein; they have not veri ed and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors of the current study.
After we obtained the approval on our proposed protocol and the letter of support from the cancer committee chair at Cleveland Clinic, the NCDB was queried for patients diagnosed with metastatic colorectal adenocarcinoma between 2004 and 2016. Survival follow-up for the studied cohort is available till December 2016. It's worth mentioning that the metastatic status provided in the database is related to the status at the time of the diagnosis. NCDB doesn't provide the details about the possible nonmetastatic cases that subsequently developed metastatic disease.

Patient selection
For our study, we selected the patients who were diagnosed with metastatic colorectal adenocarcinoma with liver metastases at age of 18 or older. We excluded patients with metastases to other sites other than the liver. Patients who didn't have radical surgical resection to the primary tumor and/or didn't have surgery to distant metastatic sites were also excluded in addition to patients with unknown chemotherapy status. Also, we excluded patients who lost follow up which are de ned by patients with missing details about their last contact status as alive or dead, and those with missing information about their time from diagnosis to the last contact.

Variables
Using the NCDB, we included the following patient demographic variables: Age, gender, race, median income, education level, insurance status, facility type (which includes community cancer program, comprehensive community cancer program, academic/research program, and integrated network cancer program) in addition to the geographic area which was classi ed as metropolitan, urban or rural location. Also, we collected the following disease-related variables: clinical tumor and nodal categories based on American Joint Committee on Cancer (AJCC) staging, microsatellite instability (MSI) status, Charlson-Deyo comorbidity score, and chemotherapy status.

Outcome
The primary outcome of this study was overall survival (OS) in months de ned as the time from diagnosis to the time of death for any reason.

Statistical analysis
We compared the baseline demographics and characteristics between patients who were treated with surgery and chemotherapy versus patients who were treated with surgery only using the Pearson Chisquare test.
For OS comparisons, we split the cohort into two separate groups according to the type of treatment the patients received, the rst group who was treated with surgery and chemotherapy and the second group who was treated with surgery only. Then, we compared the OS between the two groups. Moreover, we conducted additional OS comparisons (limited to patients who received perioperative chemotherapy) between patients who received adjuvant chemotherapy versus neoadjuvant chemotherapy; and between patients who received single-agent versus multiagent chemotherapy.
OS analyses were evaluated using the Kaplan-Meier survival method. A log-rank test was used to evaluate survival differences between groups. We used Cox regression analysis to conduct multivariable analyses to evaluate the factors associated with improved OS; hazard ratios (HRs) with associated 95% con dence intervals (CI) were accordingly generated, statistical signi cance was de ned as a P value less than 0.05 for all analyses. The following factors were included in the multivariable model: facility type, age, gender, race, median household income, education level, insurance status, area, Charlson score, clinical T stage, clinical N stage, MSI status, and chemotherapy use.
All the above statistical analyses were performed using the SPSS Statistics 27.0 To further reduce any potential biases related to the variation in baseline characteristics between the patient group that received chemotherapy and the one that didn't receive chemotherapy, we used nearest neighbor, 1:1, propensity score matching using RStudio Version 1.3.1093 (MA, USA).
As a sensitivity analysis, we repeated all the above analyses after excluding patients who died within 90 days of the most de nitive primary site surgery (to eliminate the effect of post-surgical complicationrelated mortality on overall survival).
To further examine the impact of MSI status on the above survival analyses, we repeated the above survival analyses among patients with MSI-H versus those with MSI-L status.

Baseline characteristics
Figure-1 provides a owchart for patient selection within the current study. We ended up with 6,883 patients with metastatic CRC and liver metastases; 6,042 (87.8%) were treated with surgery and chemotherapy and 841 (12.2%) were treated with surgery only. Compared to the surgery-only group, the perioperative group was likely to be younger (58 vs 70 years old), males (55.5% vs 47.4%), and had more MSI-L status (27.8% vs 16.6%). The baseline characteristics' comparison between the two groups is summarized in Table 1. Next, in our sensitivity analysis, we excluded all patients who died within 90 days of most de nitive primary site surgery (N = 550), which left us with 6,333 patients; 5,834 (92.1%) who were treated with surgery and chemotherapy, and 499 (7.9%) who were treated with surgery only. The baseline characteristics' comparison between the two groups is summarized in Table 2. We compared the OS between patients who were treated with surgery and chemotherapy versus patients who were treated with surgery only and found that patients who were treated with surgery and chemotherapy had better OS compared to patients who were treated with surgery only (median OS 42.9 months vs 6.4 months, P < 0.001) as illustrated in the Kaplan-Meier survival curves (Fig. 2).
Propensity score matching yielded 1,682 patients for analysis: 841 patients in the no chemotherapy group and 841 patients in the chemotherapy group (supplementary gure A shows the histogram for raw data and matched data). Comparison of baseline characteristics between the groups after propensity score matching is shown in supplementary table A.
We compared the OS between patients who received chemotherapy versus patients who didn't receive chemotherapy using Kaplan-Meier analysis on the propensity score-matched groups and found that patients who received chemotherapy had better OS than those who didn't receive chemotherapy as

Discussion
To our knowledge, this is the largest real-world study that looked at the chemotherapy effect in metastatic CRC with only liver metastases who were treated with metastasectomy and compared the OS between patients who received perioperative chemotherapy versus patients treated with surgery only in addition to comparing the OS between adjuvant chemotherapy and neoadjuvant chemotherapy and comparing the OS between single-agent and multiagent chemotherapy.
Our study showed that patients with metastatic CRC with liver metastases who were treated with chemotherapy in addition to surgery had statistically signi cant better OS compared to patients who didn't receive chemotherapy. Also, in the multivariable analysis, we found that chemotherapy was independently associated with better OS.
There is limited data about the role of perioperative chemotherapy in patients with liver metastases from CRC. Most of the studies had a small number of patients and were designed to evaluate the effect of chemotherapy on recurrence-and progression-free survival rather than overall survival. And the majority of the studies that evaluated the effect of chemotherapy on OS didn't report a signi cant difference in the OS between patients who received perioperative chemotherapy and patients who didn't.

Nordlinger et al. in a European
Organization for Research and Treatment of Cancer (EORTC) trial based on 364 patients with metastatic CRC and up to four liver metastases, who were recruited from 78 hospitals from different European countries, compared the outcomes of 182 patients who received perioperative FOLFOX-4 chemotherapy to those of 182 patients who had surgery only. Initial results showed chemotherapy was associated with better 3-year progression-free survival (PFS) compared to surgical resection alone. After a median follow-up of 8·5 years, it was found that patients who received perioperative chemotherapy had a better median OS and 5-year OS, but it wasn't statistically signi cant.
There're multiple factors in that study that could contribute to the lack of difference in the OS between the two groups. In addition to the small number of included patients, the study was designed with PFS being the primary endpoint whereas OS was a secondary endpoint. Also, the perioperative group had a higher number of non-cancer deaths compared to the surgery-only group 12,15  Among patients who received chemotherapy, we found that patients were treated with neoadjuvant chemotherapy had better OS compared to patients were treated with adjuvant chemotherapy. Also, we found patients who received multiagent chemotherapy had better OS compared to patients who received single-agent chemotherapy.
The vast majority of trials, that was done on metastatic CRC, assessed the effect of perioperative chemotherapy compared to surgical resection only, and while the comparison between adjuvant and neoadjuvant chemotherapy in non-metastatic CRC has been investigated, there are very few prospective studies done in metastatic cases 21,22 . While some studies suggest neoadjuvant chemotherapy is bene cial in unresectable disease and helps in improving the complete resection success rate 23 , other studies showed that neoadjuvant chemotherapy was associated with increased liver toxicity and suggest adjuvant chemotherapy was more bene cial in managing micro-metastases with less liver toxicity-related complications [24][25][26][27] . In the setting of not having a head-to-head comparison between adjuvant and neoadjuvant chemotherapy in metastatic disease, the NCCN guideline still recommends using perioperative chemotherapy for a total of 6 months without a preference for adjuvant or neoadjuvant chemotherapy 28 .
When it comes to the comparison between the single-agent and multiagent chemotherapy, few studies looked at that in the setting of metastatic disease. Seymour et al. evaluated 2135 patients with inoperable metastatic or locoregional CRC and compared the OS in 3 different chemotherapy strategies; the rst group received single-agent as a rst line and single-agent as a second line, the second group received single-agent as a rst line and combination chemotherapy as a second line, the third group received combination chemotherapy as a rst line. The study showed that the group treated with combination chemotherapy as a rst line had better OS compared to the group that received single-agent chemotherapy in both rst and second line. Interestingly, there was no statistically signi cant difference in the OS between the group that received single-agent chemotherapy as a rst and second line and the group that received single-agent chemotherapy as a rst line and combination chemotherapy as a second line. These results are similar to our result in one way that patients who received multiagent chemotherapy as a rst line had better OS than patients who received single-agent as a rst line.
Unfortunately, the NCDB doesn't provide details about the second line chemotherapy to be able to compare patients according to second line chemotherapy like the mentioned study 29 . The biggest limiting factor in this study is that recruited patients were selected as a poor-prognosis group who were de nitely incurable and patients with operable metastases were excluded.
Koopman et al. in phase III randomized controlled trial (CAIRO study), reported no statistically signi cant difference in the OS between patients with advanced CRC who received combination chemotherapy as a rst line and patients who received the same chemotherapy agents but were given sequentially 30 . Similar to what we mentioned before, the chemotherapy details in NCDB are limited to rst line chemotherapy.
Our study has some limiting factors that need to be discussed; rst, the type of chemotherapy that patients received was not available. This could have impacted the outcomes of some patients receiving adjuvant chemotherapy. Second, the differentiation between single-agent and multiagent chemotherapy is limited to the rst line of chemotherapy not total chemotherapy the patient may have received. Third, the distribution and completeness of surgical resection of liver metastases were not speci ed in the database. This is an important element that could have affected patients' outcomes if imbalanced between both groups. Fourth, NCDB also doesn't have details about other oncologic endpoints like disease-free survival and cancer-speci c survival (both represent important endpoints to evaluate the effect of chemotherapy). Fifth, the retrospective nature of data collection within the NCDB might have affected the veracity of the analyses as well.
It's worth mentioning as well that in our study, patients who had surgery only were older compared to patients who received perioperative chemotherapy (mean age in years 70 vs 58) which could be a signi cant factor that affects the difference in the OS between the two groups as older patients tend to have more comorbidities and as a result more likely not to receive chemotherapy. To account for this potential bias, we did a propensity score matching for age and Charlson comorbidity score as well as multivariable Cox regression analysis incorporating other relevant baseline factors.

Conclusion
In conclusion, our study shows that the addition of chemotherapy to surgery improves the OS in metastatic CRC with liver metastases treated with metastasectomy. Also, neoadjuvant and multiagent chemotherapy seem to improve the OS compared to adjuvant and single-agent chemotherapy, respectively. Further large-scale studies are needed to evaluate the effect of chemotherapy role and timing not only on recurrence and PFS but also on OS.  We compared the OS between patients who were treated with surgery and chemotherapy versus patients who were treated with surgery only and found that patients who were treated with surgery and chemotherapy had better OS compared to patients who were treated with surgery only (median OS 42.9 months vs 6.4 months, P<0.001) as illustrated in the Kaplan-Meier survival curves Figure 3 After excluding patients who died within 90 days of the most de nitive primary site surgery, we compared the OS between patients who were treated with surgery and chemotherapy versus patients who were treated with surgery only. We found that patients who were treated with surgery and chemotherapy had better OS compared to patients who were treated with surgery only (median OS was 43.6 months vs 16.6 months, P<0.001) as illustrated in the Kaplan-Meier survival curves Figure 4 After including only patients who received chemotherapy and excluding patients with the unknown number of chemotherapy agents they received, we compared the OS between patients who received single-agent chemotherapy and patients who received multiagent chemotherapy. We found patients who received multiagent chemotherapy had statistically signi cant better OS compared to patients who received single-agent chemotherapy (median OS was 43.7 months vs 36.1 months, P<0.001) as illustrated in the Kaplan-Meier survival curves After excluding patients who did not received chemotherapy and with unknown time duration between diagnosis and surgery and time duration between diagnosis and chemotherapy, we compared the OS between patients who received neoadjuvant chemotherapy versus patients who received adjuvant chemotherapy. We found patients were treated with neoadjuvant chemotherapy had statistically signi cant better OS compared to patients were treated with adjuvant chemotherapy (median OS was 48.6 months vs 38.6 months, P<0.001) as illustrated in the Kaplan-Meier survival curves

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.