Influence of concurrent capecitabine based chemoradiotherapy with bevacizumab on the survival rate, late toxicity and health-related quality of life in locally advanced rectal cancer: a prospective phase II CRAB trial

In brief, patients with histologically proven stage II/III adenocarcinoma of the rectum within 15 cm of the anal verge and without contraindications for ChT or targeted agents were included. Local extend of the disease was determined by magnetic resonance imaging.

Radiotherapy (RT) was delivered using three-dimensional conformal computed tomography (CT)-based treatment planning. Four-field box technique with all fields treated daily was used. Patients received 45 Gy to the pelvis plus 5.4 Gy as a boost to the primary tumor in 1.8 Gy over 5.5 weeks. Capecitabine 825 mg/m² twice daily was administered concomitantly continuously throughout of RT without interruptions on weekends. Patients received bevacizumab intravenously at a dose 5 mg/kg 14 days prior and on days 1, 15 and 29 during chemoradiotherapy (CRT). Resection (low anterior resection or abdominoperineal amputation with total mesorectal excision) was performed 6–8 weeks after the completion of CRT. Patients with histopathological R0 resection received 6 cycles of adjuvant chemotherapy with capecitabine, while in those with R1 additional 2 cycles were given. Pathologic response after CRT with bevacizumab was determined according to Dworak tumor regression grade (TRG) system. The pathological complete response (pCR) was defined as TRG 4, meaning no tumor cells in surgical specimen.

Results for primary endpoint of this prospective phase II study, the pCR rate, and on some secondary endpoints (pathological response rate, rate of sphincter-sparing surgical procedure, radical resection rate, acute and perioperative toxicity of combined modality treatment) have been reported previously.15 Here we report results on other secondary endpoints of this updated analysis: local control (LC), disease free (DFS) recurrence-free (RFS) and overall survival (OS), late toxicity and health related QOL. Survival rates were calculated using Kaplan-Meier technique. All time intervals were calculating from the date of inclusion. The end day for LC was the date of last follow-up or recurrence in the pelvis; for OS the date of last follow up or death from any cause; for DFS the date of relapse, secondary cancer, death for any cause or the last follow-up.

Separate analysis was performed for health related QLQ applying the questionnaires launched by the EORTC. For this study the core questionnaire EORTC QLQ-C3016 adjoined with the colo-rectal module EORTC QLQ-CR3812 was delivered to the patients twice: before starting the treatment and one year after the treatment was finished. The EORTC QLQ-C30 is a questionnaire assessing individual HRQL during the previous week. The EORTC QLQ-C30 has 30 items and is divided into five function scales (physical, role, cognitive, emotional and social functions); three symptom scales (fatigue, pain, nausea or vomiting) and one global health-status/quality of life dimension. The six single items address specific symptoms: dyspnoea, appetite loss, insomnia, constipation, diarrhoea and a question addressing the financial impact of the disease. The EORTC QLQ-CR38 has 38 questions and is divided into four functional and seven scales of symptoms/problems. The answers recorded by the questionnaires were transformed into dimensions ranged 0–100 according to the EORTC scoring instructions.17 For functional scales and single items higher scores represent a higher level of functioning, but for symptom scales and single items, a higher score represents a higher level of symptoms. To examine the statistically significant changes in QLQ scores over time the Wilcoxon signed-rank test was applied.

All statistical analysis was performed using the SPSS statistical software package, version 24 (SPSS Inc, Chicago, IL, USA). Values of p < 0.05 were considered statistically significant.

The 5-year OS was 72.2% (95% CI 58.2–84). Recurrences were observed in 14 patients (23.3%) and in one secondary cancer occurred. The latest distant recurrence was observed 54 months after the operation. The 5-years recurrence-free survival and DFS were 75.6% (95% CI 64.6–81.3) and 70% (95% CI 58.5–81.5), respectively. Local recurrence as any component of first failure occurred in 4 patients (6.7%), with isolated local recurrence in 1 (1.7%). The 5-year LC rate was 92.4% (95% CI 85.4– 99.4). Figure 1 illustrates OS, DFS and recurrence-free survival and LC of patients treated in CRAB trial.

Figure 1

Based on the Cox proportional hazards regression model, there were no significant association between OS or DFS and gender, age, performance status, cT, cN, TRG and adjuvant ChT. Patients with pathological positive nodes or pathological T3–4 tumors had significantly worse survival than patients with pathological negative nodes or T0–2 tumor (Table 1 and Figure 2).

Figure 2

Nine patients (14.8%) developed grade 3 late complications of combined modality treatment (Common Terminology Criteria for Adverse Events version 4.0) for which hospitalization and/or urgent intervention was needed. We recorded the first event 12 months and the last 87 months after the operation with the median time to occurrence of 48 months. All these complications, time to occurrence after surgery of primary cancer and procedures needed are presented in Table 2. Among patients without mentioned complications, we observed permanent defecation complications in 11: constipation in 4, defecation urgency in 3 and fecal incontinence in 4 colostomy free patients. Two patients with stoma experienced permanent urinary complications: 1 incompleted bladder emptying and 1 urinary incontinence.

The results of the health related QOL analysis is summarized in Table 3. One year after treatment there were no significant changes in global quality of life, however the physical and social functioning significantly decreased. On the other hand, there was a significant decrease in three symptoms: pain, insomnia and diarrhea.

Based on QLQ-CR38 scores there were no significant change on three functional scales (i.e. future perspectives, sexual functioning and enjoyment), but the body image scores significantly increase. One year after treatment has been completed our patients reported significantly less problems with weight loss, but there was a significant increase in ChT side effects (dry mouth, thin or lifeless hair and different taste) and sexual dysfunction in men.

Compared to our study, Willett et al. reported better 5-year LC rate (92,4% vs. 100%) and OS (72.2% vs. 100%).10 However, no difference in 5-year DFS (Willett et al. vs. CRAB, 75% vs. 75.6%) was observed. Pathologic complete response rate and gender ratio between studies were comparable. Some differences were seen regarding pathological nodal stage and age of patients. Our results revealed that pathological positive nodes and pathological T3/4 tumors were significantly associated with worse survival (Table 1 and Figure 2). Among patients with pathological positive nodes and pathological T3/4 tumors, the 5-year OS and DFS were 37.5% and 60.9%, 41.7% and 61.8%, respectively. These poor results for high-risk subgroup suggest that more aggressive approach is needed for such a patient. Proportion of patient included in CRAB study with pathological negative nodes was lower than in American study (19.7% vs. 28%).10 This fact could affected poorer survival results in CRAB study. Other studies have also shown that pathologic nodal status may represent a superior predictor of better survival for patients with LARC.21, 22

Furthermore, patient characteristics (sex and age) predict survival. In the largest analysis examining the impact of demographic characteristics on the survival of patients with rectal cancer, older age and male gender are associated with worse cancer-specific and OS.23 One of the possible causes for our worse OS could be in the older population that was included in the CRAB study compared to American study (median age with range; 60 years [31–80] vs. 51 years [35–72]).10

A comparison between neoadjuvant single-agent capecitabine CRT and the current study revealed promising oncologic outcome with adding bevacizumab in the standard neoadjuvant treatment of LARC in Slovenia (5-year OS, DFS and LC; 61.4, 52.4% and 87.4 vs. 72.2, 75.6 and 92.4%).24 Thus, this difference is not statistically significant and we cannot state with certainty that the improvement depends only on adding bevacizumab to standard neoadjuvant capecitabine-based CRT due to differences in the prescribed radiation dose. In CRAB study we used additionally concomitant radiation boost for which it is known that might have a positive effect on OS.25

Bevacizumab may cause severe late side effects in metastatic setting but late side effects are relatively rare.26 Most often described bevacizumab related late adverse events are spontaneous intestinal perforation and delayed anastomotic leak.27, 28 The addition of bevacizumab to standard CRT could be one of the possible causes for a higher proportion of fistulas in our study.

Only few studies analyzed the incidence of the late anastomotic leakage after low anterior resection (LAR). Delayed anastomotic leakages that develop after 30 days after surgery are not uncommon. The incidence is between 0.3% and 9.8%.29, 30, 31, 32, 33, 34, 35 It is not clear whether early and late anastomotic leakage after LAR are different entities because some patients with late anastomotic leakage may show uneventful postoperative clinical recovery.29 Late leakages more frequently involve the fistula type (22–100%) than early postoperative leakage.29, 30, 31, 32, 34, 35 Pelvic abscess and anastomotic-vaginal fistula were the most common causes of delayed complications.35

In our study only one female patient developed anastomotic vaginal fistula 12 months after rectal surgery. In the two studies with higher number of female patients included, the overall rate of anastomotic-vaginal fistula after LAR was higher, 3% (11/371) and 5.1% (20/390).36, 37 Anastomotic-vaginal fistulas in both studies were diagnosed mostly late, on median postoperative day 83 (15–766) in the first study and 25 (5–172) in the second. Despite that only 3% of the female patients in the first study were treated with preoperative ChT and no one with RT, pre-op ChT was one of the independent risk factor for anastomotic-vaginal fistula formation.36 In the second study, risk factors for anastomotic-vaginal fistula were preoperative radiotherapy, anastomosis < 5 cm above the anal verge and cancer stage IV.37

The incidence of the delayed anastomotic leakage in present study was high (13.1%), but it is important that only three fistulas were diagnosed before 5 years of follow up. Comparing the incidence of the delayed anastomotic leakage in our study with other studies is controversial. Follow up in our study was the longest than it was in other published studies. The longest interval between the surgery and fistula formation we found in the literature was 5.7 years and in our study 7.25 years. RT and ChT were identified as an independent risk factor for the development of the late anastomotic complications.32, 34, 35 In all published studies patients with very mixed type of neo-adjuvant and adjuvant treatment or without treatment were included. On the other side, in our study all patients were treated with neo-adjuvant CRT. According to previous facts, it is important that we pay attention to the appearance of early signs of the fistula during follow up in the patient previous treated with bevacizumab even several years after the completion of treatment.

The global health status (mean 68), which refers to the general assessment of the health and quality of life of the last seven days, was similar over time, and also comparable to the Slovenian general (mean 71.1) and CRC patients (mean 68.3).38, 39

The values of social and physical functioning indicated by patients were significantly influenced by the time between diagnosis and 1 year after treatment indicating a high incidence of problems in this area. On the other hand, 1 year after treatment patients reported less pain, insomnia and diarrhea. These results are consistent with QLQ-C30 scores of Slovenian CRC patients after surgical treatment.39 However, compared above mentioned items with the general Slovenian population, patients 1 year after treatment report more diarrhea, lower physical and social functioning, but less pain and insomnia. The population-based QOL reference values should be taken into account in the interpretation of disease progress and treatment effects. For Slovenian general population it was shown that gender, age and self-rated social class are important confounders in the QOL scores descriptions.38

Our long-term trends in longitudinal QOL are in agreement with study published from Couwenberg et al. including rectal cancer patients treated with CRT and surgery.20 They demonstrated that treatment of rectal cancer has larger impact on QOL decline within 3–6 months after the start of treatment, but still gradually improves within 1 year after treatment. Moreover, within two years all scores normalize towards pretreatment levels, although compared to general population lower functioning, more insomnia and fatigue persist for more than 2 years from diagnosis.

Based on QLQ-CR38 significantly more ChT side effects and sexual dysfunction of men were observed after 1 year of treatment. However, body image has improved and patients reported less weight loss. Male sexual dysfunction is common and remains high after multimodality treatment for rectal cancer, more precisely, surgical nerve damage remain the main cause.40 Patients treated with well-defined and standardized technique of total mesorectal excision together with minimally invasive techniques experienced less sexual dysfunction compared to conventional surgery.41 In addition, laparoscopic resection contributes more to the maintenance of the nerves.40

The main limitations of this study include the single center design of the study with small number of patients, which limits statistical power. To assess

the QOL we have chosen a QLQ-CR38 and not revised successor QLQ-CR29, because the latter was published later in 2007. Consequently, our results are less comparable to others that have been used QLQ-CR29.