In this study, we analyzed clinical factors or treatment variables that are associated with a complete response to preoperative CRT. Our univariate and multivariate analysis results showed that a pretreatment CEA level of ≤5 ng/mL and an interval from the completion of neoadjuvant CRT to surgery of >7 weeks were significantly associated with an increased rate of pCR.

Few studies have evaluated clinical factors associated with complete response to preoperative CRT for rectal cancer [16–20]. García-Aguilar et al. [16] studied the clinical factors of 168 patients with locally advanced rectal cancer, but the authors failed to find any factors that were associated with pCR. Kalady et al. [17] identified that an interval of >8 weeks between the completion of preoperative CRT and surgical resection was the only predictor for pCR in a retrospective review of 242 patients. In the largest sample size study to date, Das et al. [18] evaluated the predictors of pCR in 562 patients and found that the circumferential tumor extent was the only factor that was significantly associated with pCR. In a group of 249 patients assessed by Park et al. [19], the pre-CRT movability, post-CRT size, post-CRT morphology, and gross change were independent predictors of pCR. More recently, Garland et al. [20] found that the tumor size and pretreatment clinical N category were independent predictors of pCR after evaluating the clinical factors of 297 patients.

The serum CEA level is widely used as a tumor marker in patients with colorectal cancer. The pretreatment CEA level is useful for assessing the prognosis, and postoperative CEA testing is used for the early detection of recurrent disease [21]. However, few studies have evaluated the value of the CEA level in predicting the response to preoperative CRT. We found that the CEA level was significantly higher in the non-pCR group than in the pCR group; 76.0% of the patients with a pCR had a normal pretreatment CEA level, versus 58.5% of the patients in the non-pCR group. In addition, a normal pretreatment CEA level was significantly associated with pCR in both univariate and multivariate analyses. Several previous studies have reported similar results [20, 22, 23]. Yoon et al. [22] found that the pretreatment CEA level was an independent predictor for pCR. Recently, Garland et al. [20] identified that the pretreatment serum CEA levels and a decrease in the pre- to post-treatment serum CEA level were significantly associated with pCR in univariate analysis. Additionally, tumor cells that have a high density of CEA may resist radiation [24]. However, the exact mechanism is unclear and remains to be elucidated.

Radiation-induced necrosis and subsequent tumor regression is a time-dependent phenomenon in which a longer interval between the completion of CRT and surgery may increase the rate of pCR [17]. In this study, we found that an interval of >7 weeks between the completion of CRT and surgical resection was significantly associated with a higher rate of pCR (31.3 vs. 14.6%, P = 0.001). Similarly, Kalady et al. [17] reported that an interval of >8 weeks was the only independent predictor for pCR (OR = 2.63, 95% CI = 1.13–6.12, P = 0.020). Wolthuis et al. [25] reported that an interval of >7 weeks was associated with increased pCR (28 vs. 16%, P = 0.030), and the 5-year cancer-specific survival rate was higher in the long-interval group than in the short-interval group (91 vs. 83%, P = 0.046). Conversely, Stein et al. [26] and Lim et al. [27] did not find that a longer interval between CRT and surgical resection was an independent predictor of pCR. To settle these disputes about the optimal time from CRT to surgery, randomized controlled trials are needed.

Increasing data indicate that the treatment response to chemoradiation associated with the oncologic outcomes [28]. To improve the response, some randomized trials have added oxaliplatin or targeted drugs into the currently widely used fluorinated, pyrimidine-based preoperative chemotherapy regimen [29–31]. However, none of these studies reported an increased pCR rate. Recently, O’Connell et al. [29] reported that adding oxaliplatin did not improve the rates of pCR, sphincter-sparing surgery, or surgical down-staging; instead, oxaliplatin added significant toxicity. We also found that the addition of oxaliplatin did not increase the pCR rate (25.0 vs. 22.8%, P = 0.729). By contrast, waiting a longer time between CRT and surgery was significantly associated with a higher rate of pCR.

There are some potential limitations of this study. Because this was a retrospective study, there may be bias. Additionally, we evaluated patients who were treated over the course of 9 years. The CRT regimen and pathologic assessment may change over time. Other studies have reported some variables that were not included in this study, such as tumor size, circumferential extent of tumor, morphology, and gross change after CRT. These factors are subjective parameters that are susceptible to inter-observer variations, and it is difficult to assess these factors in a retrospective study.

Further prospective studies are needed to identify the predictors for pCR and elucidate its potential mechanisms. In the future, it may be possible to stratify patients into different response groups before CRT is initiated, which could allow us to develop individualized therapy plans. Patients who are considered good responders to preoperative CRT may benefit little from TME, and their optimal treatment may involve nonoperative management or local excision. Patients who are stratified into the poor response group may need more aggressive therapies or alternative new therapies.

In conclusion, this large, retrospective study demonstrated that a pretreatment CEA level of ≤5 ng/mL and an interval from the completion of neoadjuvant CRT to surgery of >7 weeks were independent clinical predictors for achieving pCR. These findings may help clinicians predict the prognosis of patients and develop individualized treatment strategies.