Study cohort

Data on low dose aspirin use (80–100 mg), derived from the PHARMO Database Network (PHARMO, Netherlands), were linked to the Eindhoven Cancer Registry (ECR). The validity of the linkage of these cohorts was described previously [18]. The ECR serves about 1.5 million inhabitants in the southern region of the Netherlands and is part of the nationwide Comprehensive Cancer Organisation (IKNL). The PHARMO Database Network is a population-based network and combines data from different healthcare settings in the Netherlands. The Outpatient Pharmacy Database was used for this study, which comprises drug dispensing records from all community pharmacies. The records in this database contain information on the type of product, date prescribed, dose and regimen, quantity, and route of administration. Drugs are coded using the Anatomical Therapeutic Chemical classification [19]. The Comprehensive Cancer Organisation is obliged to work according to the law on data protection; informed consent of the patients for this specific study was not applicable.

As previously published, aspirin initiated or continued after diagnosis was associated with improved survival for patients with colon cancer, but not for patients with rectal cancer, in our cohort [20]. Therefore, only patients with colon cancer were included in this study.

The vital status of patients (alive/dead) was established from medical records or through linkage of cancer registry data with the municipal population registries. As information on hospital dispensing was not available, follow-up started 14 days after diagnosis of colorectal cancer (T0), and was continued until last contact date (January 2012), date of loss to follow-up, or date of death—whichever occurred first.

Patients who only used aspirin before diagnosis were also excluded (n = 40, see Fig 1). Non-users were classified as those who never had a dispensing for aspirin or had a dispensing for less than 14 days after diagnosis of colon cancer. Users were defined as those who had been given a dispensing of aspirin for 14 days or more after a colon cancer diagnosis. The median duration of one dispensing was 30 days and the mean dispensing number was 12 (range 1–220).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart of patients selected for analysis.

https://doi.org/10.1371/journal.pone.0170775.g001

BRAF and KRAS tumor mutation analyses

The ECR-PHARMO cohort, as previously published by Bastiaannet et al, contained 3,586 patients [20]. Of this cohort, Formalin-Fixed Paraffin-Embedded (FFPE) tumor tissues were retrieved of 1,026 colon cancer patients who underwent a surgical resection between 2002 and 2008 [21]. Twenty-seven patients with more than one colon tumor at the time of diagnosis were excluded from this cohort (Fig 1). Additionally, 63 patients were excluded because they used aspirin before diagnosis or with a follow up less than 14 days. Of these patients, 599 patients were randomly selected with a ratio 1:2 for aspirin user: non-user, as was previously described [22].

No significant demographic differences were calculated between the total cohort (n = 999) and the randomly selected patients (n = 599) [22].

Of the included patients (n = 599), tumor areas on hematoxylin and eosin (H&E) stained tumor sections were marked by an experienced pathologist/researcher. Guided by the H&E-stained slides, 1–2 punches with a diameter of 2.0 mm diameter and variable length were taken from the tumor focus, followed by DNA extraction as described by de Jong et al [21]. For determination of KRAS and BRAF mutations status, hydrolysis probes assays were performed for the major known mutations (hotspots) in codon 600 for BRAF, c.1799T>A; p.V600E and codon 12 and 13 for KRAS; c.34G>A; p.G12S, c.34G>C; p.G12R, c.34G>T; p.G12C, c.35G>A; p.G12D, c.35G>C; p.G12A, c.35G>T; p.G12V, c.38G>A; p.G13D and c.37G>T; p.G13C, as previously described [21]. Hydrolysis probe assays were analyzed using qPCR analysis software (CFX manager version 3/0, Bio-Rad). Mutation detection was performed by two independent observers (M.R. and R.E.). All primers used for the assays were previously described [23].

Statistics

Statistical analyses were performed using the statistical packages SPSS (version 20.0 for Windows, IBM SPSS statistics) and Stata (version 12 for windows, StataCorp LP). Statistical tests were two-sided and considered significant at a p-value below 0.05.

A parametric survival model with an exponential (Poisson) distribution was used, with the use of aspirin as time varying covariate. This method prevents the introduction of time-related biases [24]. Non-users were defined from T0 until date of death or end of follow-up. Patients were considered aspirin users from the moment of first prescription, mimicking an intention-to-treat analysis. In order to investigate differential associations of aspirin use with overall survival by tumor molecular subtype, stratified analyses were performed for BRAF wild-type / BRAF mutation and KRAS wild-type / KRAS mutation, followed by an interaction analysis. The interaction analysis was performed by including a cross product of BRAF mutation status in the survival analysis and the use of aspirin and significance was assessed with the Wald test.

Adjustments for potential confounders were made for sex, age (groups), stage (pathological stage and clinical stage if pathological stage was unknown), adjuvant chemotherapy (yes/no), co-morbidity (yes/no) and tumor grade.

Survival curves were calculated according to the Simon-Makuch method, an alternative for Kaplan Meier, with the ability to include time-varying covariates [25].

A subgroup analysis was performed by excluding patients with stage IV disease.

Overview of findings

Increasing attention is paid to a personalized treatment approach, by stratifying patients into subgroups based on biomarkers. This study investigated whether the survival benefit observed in patients with colon cancer using aspirin could be associated with BRAF or KRAS mutational status. This study found that BRAF mutation status and KRAS mutation status were not distinctive in the association between low-dose aspirin use and a survival benefit in patients with colon cancer. In the multivariate analysis, patients with wild-type BRAF tumors, aspirin use after diagnosis was associated with a significantly better outcome. However, the crude hazard ratios in both groups (BRAF wild-type and mutation) are equal and the P-value for interaction was non-significant. Because no statistical interaction was observed, the distinctiveness of BRAF mutational status on the association between aspirin use and survival in the multivariate analysis could very well be based on chance. Therefore, it could not be concluded from this study that patients with BRAF mutated tumors should be withheld from using aspirin. The subgroup analysis in patients with stage I-III colon cancer showed a reduced effect size. However, due to limitations in power, no firm conclusions can be drawn from the results of this subgroup analysis.

Comparison with other studies

Nishihara et al [26] previously studied the effect of BRAF mutational status on colorectal cancer incidence and survival in patients using aspirin. BRAF mutational status showed to be of influence on the incidence of colorectal cancer. BRAF wild-type was associated with a lower risk of colorectal cancer, multivariable hazard ratio; 0.73 (95% CI 0.64–0.83) whereas BRAF mutated tumors did not show a reduced risk of colorectal cancer (HR 1.03, 95% CI 0.76–1.38). A survival analysis in this study was performed as an exploratory analysis, and in both subgroups (BRAF mutation and wild-type tumors) no association between the use of aspirin and improved survival was found, in line with our study.

Strengths and limitations

Our study has several strengths. To the best of our knowledge, no studies have assessed the association between KRAS, aspirin and survival in patients with (colon) cancer. Information regarding aspirin use and dose was derived from prescriptions rather than patient recall, resulting in a precise definition of regular aspirin use. By using a time-varying covariate for the use of aspirin, the risk of non-differential misclassification is reduced [24]. Lastly, a robust and reliable method was used to determine BRAF and KRAS mutational status, resulting in a 98% successful determination of mutational status and therefore a relatively large cohort.

However, several limitations must be acknowledged. First, over-the-counter aspirin use and adherence was unknown and could be a potential source of bias. Nevertheless, it has been shown that pharmacy data can give valid estimates, despite over-the-counter availability of aspirin [28]. It seems unlikely that a large fraction of patients bought aspirin over-the-counter: low-dose aspirin is only indicated for secondary cardiovascular prevention in the Netherlands and therefore this should always be made available through a doctor’s prescription. The main reason for over-the-counter purchase of aspirin is its use as analgesic, however low-dose aspirin does not suffice as analgesic. Moreover, the possible benefits of aspirin as treatment for cancer were not widely known during the analysis period.

Second, this is a retrospective study in which patients were not randomized. Even after adjustment for potential confounders, residual confounding may still be present. Confounding by indication could, in general, have resulted in overestimation of the results. For cancer patients to be prescribed aspirin, patients should have a cancer prognosis which outweighs the risk of cardiovascular disease. Patients to whom this does not apply should, in theory, not be prescribed aspirin. These patients are then assigned into the non-user group which could have resulted in an overestimation of the association between aspirin use and survival. The variation in length of use of aspirin and the moment patients start using aspirin makes it difficult to conclude any causality from this study, only associations were observed. Therefore, the current ongoing randomised controlled trials are highly warranted. However, this is a limitation of all retrospective studies.

Third, no information regarding disease-specific survival was available in this study. However, a large meta-analysis of individual patient data found that the benefit of patients using aspirin as secondary prevention for cardiovascular disease is only 0.91 (95% CI 0.82–1.00) [29]. This can therefore not fully explain the observed overall survival benefit for the aspirin users in the current study.

To the best of our knowledge, this study is one of the largest cohorts analyzing the association between the use of aspirin, overall survival in colon cancer patients and mutational status of BRAF and KRAS, however numbers were too small for any additional subgroup analyses.

Clinical implications

Precision medicine has gained more attention over the last couple of years and multiple publications were dedicated to the discovery and development of clinical prognostic and predictive biomarkers [11]. Nevertheless, conflicting results have been observed for every previous appointed biomarker regarding the association between aspirin use and survival. Proposed biomarkers associated with aspirin use and survival are COX-2, HLA class I, PIK3CA mutation status and several specific genetic profiles [22, 30–32]. Mutations in BRAF and KRAS, acting in the RAS-RAF-MAPK kinase cascade and mutated PIK3CA, acting in PI3K-PTEN-AKT signaling pathway, are known for their contribution to the development of CRC and are associated with cancer prognosis [11, 33]. The strong survival benefit in patients with a PIK3CA mutation can only partly explain the effect of aspirin found in the general cancer population. The magnitude of the clinical benefit as found in CRC cohorts, cannot be explained by patients with a PIK3CA mutation solely, because of the low mutation frequency (15%). Therefore, additional biological processes must be responsible for the effect of aspirin on survival.

In this study we were focusing on the RAS-RAF-MAPK cascade, known for the crosstalk with the PIK3CA pathway, in relation to aspirin use [17]. No differentiating effect of aspirin use in BRAF or KRAS mutated tumors could be detected. It could be (cautiously) concluded that biomarkers from the RAS-RAF-MAPK cascade and an activated PI3K-PTEN-AKT signaling pathway may not be able to fully unravel the complexity and versatility of the aspirin effect on cancer. Therefore, the evidence points more towards a generalized, systemic effect [5, 34].

One suggested hypothesis is the role of aspirin as thrombocyte aggregation inhibitor [34]. By inhibiting the aggregation of thrombocytes, which naturally shape around circulating tumor cells, the immune system is able to detect and subsequently clear tumor cells from the circulation. Another hypothesis could be found in the anti-inflammatory effects of aspirin [35]. In the past years, several publications focused on the identification of subtypes of colorectal cancer, highlighting the heterogeneity of the disease and aiming to improve optimal allocation of treatment modalities [11, 36, 37]. Linnekamp et al advocate that the development of new agents should take place in a disease sub-type-specific fashion and in that manner generate more effective therapies [36]. These subtypes could also be the key to personalized treatment with aspirin. With this information and growing consensus on these subtypes, new research could focus on the effect of aspirin in (inflammatory) specific subtypes.