Bevacizumab for metachronous metastatic colorectal cancer: a reflection of community based practice

Data from the population-based Netherlands Cancer Registry (NCR), more specifically from the Eindhoven area, were used. The Eindhoven Cancer Registry (ECR) collects data of all patients with newly diagnosed cancer in a large part of the Southern Netherlands. The ECR covers an area of approximately 2.4 million inhabitants, six pathology departments, ten hospitals and two radiotherapy institutions. Patient and tumor characteristics are collected from medical records by specially trained registry staff after notification by pathologists and medical registration offices, resulting in high quality of the data. The completeness of cancer registration is estimated to exceed 95 %. In the ECR, primary tumors are classified according to the TNM classification of Malignant Tumors by the international Union Against Cancer (UICC), 7^th edition [7]. Additional data were retrospectively collected on metachronous metastases for patients diagnosed between 2003 and 2008 with stage I-III CRC. Hospitals were asked to participate in the study by giving permission to use their data from the ECR and by giving permission for the retrospective registration of additional data. All hospitals voluntarily participated.

Metachronous metastases were defined as distant metastases of primary CRC in other organs, diagnosed at least 3 months after CRC diagnosis. However, the majority of metachronous metastases diagnoses (94 %) occurred at least 6 months after CRC diagnosis. Patterns of metastatic disease were determined based on the site of metastasis according to the International Classification of Diseases for Oncology (ICD-O), which could involve multiple localizations. Median time from primary diagnosis to data collection was 5.3 years (range 1.5–8.8 years). All consecutive patients with metachronous metastases from primary resected CRC were selected (n = 1010). Patients diagnosed with metachronous metastases before 2005 (n = 100) were excluded as bevacizumab is registered and recommended as a therapeutic option in addition to first-line chemotherapy in the Netherlands since 2005 [8]. Subsequently, patients undergoing surgery for metastases were excluded (n = 232), resulting in a study population treated with palliative intent (n = 678) of whom 361 received palliative chemotherapy (with or without palliative procedures; bypass, anastomosis, stoma). These latter patients were categorized into two treatment groups according to the prescription of bevacizumab in addition to palliative chemotherapy (Fig. 1). In the current study, we focused on the first-line palliative treatment as this is the indication for which bevacizumab is registered in the Netherlands.

Variation in the prescription of bevacizumab between hospitals of diagnosis in the ECR-region was assessed using a χ² test. Also, differences in patient and tumor characteristics and the prescription of bevacizumab between chemotherapeutic regimens were tested by means of a χ² test. To discriminate independent predictors of treatment with bevacizumab, a multivariable logistic regression model was used. Adjustments were made for relevant patient and tumor characteristics: gender, age, comorbidity at time of CRC diagnosis, primary tumor localization, adjuvant chemotherapy, time to metastases, period of metastases diagnosis, number of metastases and the prescribed first-line chemotherapeutic regimen. In order to limit potential endogeneity bias due to the population-based nature of the data, a propensity score matched sample was created. Propensity scores were determined with a logistic regression model in which bevacizumab was the variable of interest and the independent variables were factors potentially associated with the use of bevacizumab (similar to variables taken into account in the multivariable logistic regression analysis). Patients were then matched within tight bounds of the propensity scores (probability could vary by no more than 1 %). Overall survival time was defined as the time from diagnosis of the first metachronous metastatic site to death or lost to follow-up. Patients still alive at the end of follow-up (January 1^st, 2014) and those who emigrated were censored. Crude survival estimates according to the prescription of bevacizumab were calculated with the Kaplan-Meier method and presented up to 48 months in both the total study population and the propensity score matched sample. Median survival (MS) was presented in months and corresponding 95 % confidence intervals (CIs). A log-rank test was carried out to evaluate significant differences between survival curves. Multivariable Cox regression analyses were performed in both the total study population and propensity score matched sample to evaluate the independent effect of additional bevacizumab on the risk of death. Adjustments were made for the clinically relevant variables age, comorbidity, localization of primary tumor, adjuvant chemotherapy, time to metastases, period of metastases diagnosis, number of metastases, prescribed first-line chemotherapeutic regimen and the total number of systemic lines for the treatment of metastases. All analyses were performed with SAS/STAT® statistical software (SAS system 9.3; SAS institute, Cary, NC).

In the Netherlands, the NCR and Dutch hospitals have a formal agreement that all cancer patients are informed about registration in the Cancer Registry and the possibility to decline registration. According to the Dutch law, all cancer patients are included in the NCR unless the patient has objected to be registered. Therefore, consent of the patient for this specific study was not applicable.

The NCR retrospectively collects data from medical records and is obligated to work according to laws in which the privacy of patients and doctors is fixed in regulations; the law about protection of privacy and the law “Geneeskundige BehandelOvereenkomst”. An independent Committee of Privacy reassures that the NCR works compliant to these regulations. In the Netherlands, retrospective studies with data collected from medical charts do not fall under the scope of the Medical Research Involving Human Subjects Act (‘Wet Medisch-wetenschappelijk Onderzoek”) as patient integrity is not violated in these studies. Therefore, this study was exempted from further medical ethics review.

Of the 361 patients treated with first-line systemic therapy, 219 patients received combination chemotherapy (CAPOX/FOLFOX in 96 %) and 142 patients received single-agent chemotherapy (capecitabine 74 %, irinotecan 20 %). Patient and tumor characteristics of patients treated with these chemotherapeutic regimens are shown in Table 1. Patients receiving combination chemotherapy were younger, had less comorbidities, were more often diagnosed with rectal tumors and less often received prior adjuvant chemotherapeutic treatment than patients receiving single-agent chemotherapy. Moreover, patients treated with combination chemotherapy more frequently received bevacizumab (n = 153, 70 %) than patients treated with single-agent chemotherapy (n = 32, 23 %, p < 0.0001).

In multivariable regression analysis including adjustment for the type of prescribed first-line chemotherapeutic regimen, several factors were shown to influence the probability to receive additional first-line bevacizumab (Table 2). It was confirmed that patients treated with combination chemotherapy were more likely to receive bevacizumab than patients treated with single-agent chemotherapy (OR 9.666, 95 % CI 5.43–17.05). Moreover, the odds for treatment with bevacizumab was higher for patients diagnosed with metastases in a recent time period than patients diagnosed with metastases shortly after the introduction of bevacizumab in Dutch guidelines (2005–2006). The probability to receive bevacizumab was lower for patients with ≥2 comorbidities than patients without comorbidity (OR 0.4, 95 % CI 0.21–0.81) No association was observed between age and the use of bevacizumab. However, elderly patients (≥75 years) were less likely to receive combination chemotherapy (OR 0.2, 95 % CI 0.11–0.30).

As shown in Fig. 3, the addition of bevacizumab to first-line palliative chemotherapy was associated with an improved median overall survival, from 14 months (95 % CI 11–16) to 22 months (95 % CI 19–24) (log rank p < 0.0001). In the propensity score matched sample, including 60 patients (with bevacizumab n = 30, without bevacizumab n = 30), comparable results were found with a median overall survival of 13 months (95%CI 7.62–18.92) versus 25 months (95%CI 7.62–18.92) (log rank p < 0.05). In multivariable analysis, the addition of bevacizumab to palliative chemotherapy resulted in a reduced hazard ratio on death, in both the total study population (HR 0.6, 95 % CI 0.45–0.73) and propensity score matched sample (HR 0.3; 95 % CI 0.14-.079, Table 3). After stratification for the prescribed first-line chemotherapeutic regimen, the beneficial effect of the addition of bevacizumab was observed in the subset of patients receiving combination chemotherapy (HR 0.6; 95 % CI 0.40–0.81), but not in patients treated with single-agent chemotherapy (HR 0.9, 95 % CI 0.60–1.54).