Figures
Abstract
Background & Aims
Recent observational studies showed that post-operative aspirin use reduces cancer relapse and death in the earliest stages of colorectal cancer. We sought to evaluate the cost-effectiveness of aspirin as an adjuvant therapy in Stage I and II colorectal cancer patients aged 65 years and older.
Methods
Two five-state Markov models were constructed separately for Stage I and II colorectal cancer using TreeAge Pro 2014. Two hypothetical cohorts of 10,000 individuals at a starting age of 65 years and with colorectal cancer in remission were put through the models separately. Cost-effectiveness of aspirin was evaluated against no treatment (Stage I and II) and capecitabine (Stage II) over a 20-year period from the United States societal perspective. Extensive one-way sensitivity analyses and multivariable Probabilistic Sensitivity Analyses (PSA) were performed.
Results
In the base case analyses, aspirin was cheaper and more effective compared to other comparators in both stages. Sensitivity analyses showed that no treatment and capecitabine (Stage II only) can be cost-effective alternatives if the utility of taking aspirin is below 0.909, aspirin’s annual fatal adverse event probability exceeds 0.57%, aspirin’s relative risk of disease progression is 0.997 or more, or when capecitabine’s relative risk of disease progression is less than 0.228. Probabilistic Sensitivity Analyses (PSA) further showed that aspirin could be cost-effective 50% to 80% of the time when the willingness-to-pay threshold was varied from USD20,000 to USD100,000.
Citation: Soon SS, Chia W-K, Chan M-lS, Ho GF, Jian X, Deng YH, et al. (2014) Cost-Effectiveness of Aspirin Adjuvant Therapy in Early Stage Colorectal Cancer in Older Patients. PLoS ONE 9(9): e107866. https://doi.org/10.1371/journal.pone.0107866
Editor: Talitha L. Feenstra, National Institute for Public Health and the Environment, Netherlands
Received: February 1, 2014; Accepted: August 19, 2014; Published: September 24, 2014
Copyright: © 2014 Soon et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study is supported by the following program grants: (1) National University of Singapore (NUS) Undergraduate Research Opportunities Programme in Science (UROPS) (http://www.science.nus.edu.sg/undergraduate-studies/ugenh/urops-main), (2) National Medical Research Council (NMRC) (Project number: NMRC/1226/2009) (http://www.nmrc.gov.sg/content/nmrc_internet/home.html), and (3) the Silent Foundation (http://www.silent.org.sg/index.html). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Colorectal cancer (CRC) is the third most common cancer worldwide with more than 1.2 million new cases diagnosed annually [1]. More than half of the patients diagnosed with CRC die from the disease and it is the second leading cause of overall cancer deaths in the United States [2]. Over the past decade, coincident with a rapid rise in CRC incidence rates in Asia [3], there is a dramatic increase in the proportions of CRC patients diagnosed with early stage disease [4]–[6].
Adjuvant chemotherapy has been shown to reduce the risk of recurrence and improve overall survival (OS) in patients with Stage III CRC. Chemotherapy with 5-fluorouracil reduces the relative risk (RR) of cancer recurrence by approximately 30%, and absolute risk by approximately 15% [7]. However, adjuvant chemotherapy has a much more limited role in earlier stages of CRC (Stage I and II) where its benefit is modest at best, and limited to tumors with high risk features in patients under 70 years [8], [9].
Most recently, data from a series of observational studies have strongly supported a beneficial role of aspirin use after CRC diagnosis, with a halving of disease-specific mortality rates [10]. In these analyses, aspirin’s effectiveness was not restricted to Stage III tumors, but extended to Stage I and II disease. Large randomized adjuvant studies are now underway in Asia (NCT00565708) and Europe (NTR3370) to confirm the benefit of aspirin in CRC patients.
Since aspirin is cheap, easy to administer, and has a good risk-benefit profile relative to chemotherapy, we hypothesize that aspirin might represent a cost-effective strategy for the adjuvant treatment of Stage I and II CRC where the risk of cancer recurrence is low. Such patients are currently not routinely offered adjuvant chemotherapy and are followed-up with observation alone. As the number needed to treat (NNT) to prevent one CRC recurrence or death will be much larger for Stage I and II CRC than for Stage III disease, global cost-effectiveness will be an important consideration for advocating treatment in low relapse-risk cancers.
To date, although there have been several cost-effectiveness analyses of aspirin in the primary prevention of CRC [11]–[13], no studies have been undertaken to evaluate the cost-effectiveness of aspirin in the adjuvant or secondary cancer prevention setting. Given the ever escalating costs of cancer care and constraints in health resources globally, a cost-effectiveness analysis of adjuvant aspirin in the context of treatment of cancer, in particular low-risk cancer, is both timely and important. The primary objective of this study is to determine the cost-effectiveness of aspirin as adjuvant therapy for Stages I and II CRC in the United States (U.S.) population. The U.S. was chosen as the population under study due to the relative availability of data for model input. The study model focused solely on sporadic CRC as it is the most common and relevant type of CRC [14].
Methods
Model Structure
Based on literature review and clinicians’ input, two separate Markov cohort models for Stage I and Stage II CRC respectively were constructed using TreeAge Pro 2014 (TreeAge Software, Inc., Williamstown, MA). Although the health states were identical, the state-specific transition probabilities, efficacy and utility estimates differed according to cancer stage. The five health states were: ‘Remission with Intervention’, ‘Treatment of Non-fatal Adverse Event’, ‘Remission with Unplanned Discontinued Treatment’, ‘Recurrence’, and ‘Death’ (Figure 1). In Stage I, the treatment options were aspirin or no treatment; and in Stage II, aspirin, chemotherapy or no treatment. The chemotherapy regime selected was the standard protocol of capecitabine, an oral pro-drug of 5-fluorouracil. Capecitabine was used as a comparator in Stage II disease as it is an oral agent, has better side effects profile than 5-fluorouracil [15], and is commonly used in the treatment of Stage II CRC patients with high-risk tumor features.
Patients enter the model at the ‘Remission with Intervention’ state for both stages. For the ‘no treatment’ arm of both stages, fatal and non-fatal adverse events are taken to be zero. ‘Treatment of Non-fatal Adverse Event’ state is modeled as a temporary state where patients remain in that state for only one cycle.
A hypothetical cohort of 10,000 individuals at a starting age of 65 years was simulated in each model that had a cycle length of one year and ran over 20 years. The entry age of 65 years was selected since the median age at diagnosis for Stage I and II colon and rectum cancers ranged from 66 to 73 [16]. Based on the average life expectancy of 19.1 years at 65 years of age in 2010 [17], a time horizon of 20 years was chosen. All subjects entered the model at the ‘Remission with Intervention’ health state, received the intervention specified, and then progressed through the model in annual cycles. The U.S. societal perspective was adopted for the analysis and published data on cost inputs from public databases and cost studies were utilized. Health outcomes were measured in terms of incremental cost per quality-adjusted life year (QALY), and incremental cost per life year gained (LYG).
Model Assumptions
The model assumed a uniform treatment benefit effect and fatality risk across the various regions of the ascending, transverse, descending and sigmoid colon and rectum, and all treatment effects were assumed to be immediate. Disease-free survival ratios and cancer-specific survival ratios for capecitabine and aspirin respectively were used for the imputation of treatment benefit in the model [9], [18]. Patients who experienced Grade 3 or 4 adverse events (AE) from aspirin or capecitabine were assumed to discontinue their use. In addition, the model assumed that no more than one AE could occur within each cycle. The risks of treatment-related side effects were assumed to cease immediately after completion of adjuvant treatment (five years for aspirin and six months for capecitabine), and after treatment was prematurely terminated due to serious AEs. After five years, the risk of death from other causes was thought to be equal to that of the general population of the same age. Bleeding risk was estimated from cardiovascular aspirin studies and assumed to be equal in patients with resected CRC. Bleeding risk from aspirin was assumed to be uniform across the exposure period and beneficial effects of intervention were assumed to apply during the five-year aspirin regimen. Similarly, for capecitabine, the beneficial effects were assumed to apply for the first five years of the simulation.
All patients were assumed to be treatment naïve at the beginning of treatment. For simplicity, all recurrences were deemed incurable although it is recognized that 6% from Stages I, II and III could be expected to have surgically curable recurrences [19]. Hence, the transition from ‘Recurrence’ back to either of the ‘Remission’ states was not permissible in our model. Aspirin’s cardiovascular benefit and chemoprevention effects on other cancers were not included in this model.
Transition Probabilities
Transition probabilities refer to the likelihood of an event happening in a given time period and differ from rates which are instantaneous. The transition matrices (File S2) show the probability of transition from states in the rows to states in the columns.
Model Validation
To ensure face validity, the model structures and assumptions were developed in consultation with medical oncologists. The no treatment arm was then validated using data derived from the Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database (version 8.1.2) to simulate natural history of early stage CRC. The details of the validation using SEER data can be found in File S1.
Treatment Effects
To model for the beneficial effects of aspirin and capecitabine, the relative risk of disease recurrence on aspirin or chemotherapy versus no treatment was applied to the transition probabilities associated with disease progression. Treatment effect of aspirin was specific to standard oral 325 mg aspirin tablet daily[18]. Treatment effect of oral capecitabine was assumed to be equivalent to that of the intravenous administration of 5-fluorouracil using the Mayo Clinic regimen [15]. Estimates of the beneficial effects were taken from the QUASAR study as their study population, with 91% of enrolled patients having Stage II disease, was the most similar to our hypothetical cohort [9].
For aspirin, age-related fatal (hemorrhagic death) and non-fatal (major gastrointestinal bleeding and intracranial bleeding) AEs were used [20]–[22]. For capecitabine-related side-effects, both non-fatal AE (Grade 3 or 4 hand-and-foot syndrome and diarrhea) and fatal AE were included in the model [15], [23].
The QALY and LYG were summed across all model cycles. Incremental effectiveness was estimated as the difference across treatment arms in terms of QALY or LYG.
Utilities
The respective stage-specific mean utility scores for staying in remission for Stage I and II CRC were estimated from stage-specific utilities elicited from CRC patients [24]. For the recurrence state, mean utility scores for Stage IV were applied. A utility of 0.999 (i.e. disutility of 0.001) was applied to the period aspirin was taken [25]. This represents the diminution of quality of life due to inconvenience of taking a daily pill [25], [26].
Cost Inputs
As the societal perspective was adopted, direct medical costs, indirect medical costs and non-medical costs were considered (Table 1). These included costs of drugs (aspirin and capecitabine), surveillance (physician visit, blood tests, serum carcinoembryonic antigen (CEA) level test, computerized tomography and colonoscopy), medical care (cost of care for metastatic CRC), adverse events and indirect costs (patient’s time). Costs relating to surveillance used non-facility rates in the Medicare Physician Fee Schedule from the Centers for Medicare & Medicaid Services [27], while those relating to adverse events were extracted from the Healthcare Costs and Utilization Project (HCUP) based on charges billed for clinically meaningful categories developed by the Agency for Healthcare Research and Quality [28]. All costs were inflated to 2013 (November) U.S. dollars using the Medical Care Consumer Price Index [29] and were discounted at 5% in the base case, while outcomes was discounted at a lower 3% to take into account the increase in future value of health effects [30]. All costs and outcomes were taken to be incurred at the end of the year.
Indirect costs related to patient work loss can be substantial [31]. To estimate this, time estimates associated with CRC care were used [32].
Incremental cost-effectiveness output was calculated by dividing the total incremental costs by the incremental effectiveness and reported as cost per QALY or cost per LYG. The adopted societal willingness-to-pay threshold was USD100,000 [33].
Sensitivity Analyses
Cognizant of the many underlying assumptions and the limited randomized evidence base available, one-way sensitivity analyses, extensive one-way sensitivity analyses, and multivariable probabilistic sensitivity analyses (PSA) were performed to evaluate the impact of model assumptions on the study findings.
One-way sensitivity analyses, represented in the form of tornado diagrams, were conducted for the variables shown in Table 2. The incremental cost-effectiveness ratio (ICER) is contingent on the accuracy of estimates of these variables. The 95% confidence intervals from primary sources were used whenever such data were available; where absent, a ±20% range was applied with the exception of transition probabilities that were between 0 and 0.01. For these transition probabilities, the lower and upper bound limits of 0 and 10 times base case value (or 0.01, whichever greater) were applied respectively. For variables with values that varied during the simulated cycles, the ±20% range was calculated using the largest base case value. In addition, costs more than USD100,000 were varied widely from 50% to 200% of base case values to reflect the impact of outliers. Variables excluded from sensitivity analyses were starting age, background mortality, surveillance costs, and transition probability of non-fatal AE and fatal AE for the no treatment arm.
Variables with high levels of uncertainty, identified as those with spread exceeding 50,000 from the initial one-way sensitivity analyses, were first subjected to extensive one-way sensitivity analyses to elucidate the robustness of the base case results before being included in the multivariable PSA. The ranges of values tested in the extensive one-way sensitivity analyses are 0 to 1 for all variables with this lower and upper bound limit, up to 10% for transition probabilities relating to fatal AE rates, up to 30% for transition probabilities relating to non-fatal AE, up to USD100,000 for drug cost, and up to USD600,000 for cost of metastatic CRC care. Multivariable PSA was conducted with a Monte Carlo simulation of 10,000 iterations using the appropriate distribution for the corresponding type of parameter (Table 3) [34]. Due to the limitations of the evidence available for the construction of the models in this study, a pragmatic approach to fitting distributions to parameters based on available information has to be taken [34]. Cost-effectiveness (CE) acceptability curves were then plotted with the percentage of cost-effective iterations against willingness-to-pay thresholds ranging from USD0 to USD100,000.
Results
Base Case Analysis
The results of the base case cost-effectiveness analysis are shown in Table 4. In both Stage I and II CRC, the base case analyses provide preliminary results to suggest that aspirin is a cost-effective option as compared to the other options.
The no treatment strategy remained dominated (i.e. more expensive and less effective than aspirin) in both stages. Similarly, capecitabine was also dominated. In general, although the differences in QALY and LYG of the dominated strategies were only 0.15 to 0.28 less than that of aspirin, the cost differences were substantial with a range of USD9,864 to USD61,277. The additional application of a utility of 0.999 during the period aspirin was taken did not appear to have an impact on the results.
Sensitivity Analyses
One-way Sensitivity Analyses.
Based on the one-way sensitivity analysis (Figure 2), the sensitive variables for Stage I CRC were: (i) utility of taking aspirin, (ii) transition probability of fatal AE when on aspirin, (iii) cost of metastatic CRC in recurrence state, (iv) utility score of staying in remission without intervention, (v) relative risk of disease progression when on aspirin, (vi) utility score of staying in remission with intervention, (vii) transition probability of non-fatal AE when on aspirin.
AE: Adverse event; CRC: Colorectal cancer.
For Stage II CRC, the sensitive variables were: (i) utility of taking aspirin, (ii) utility score of staying in remission with intervention, (iii) transition probability of fatal AE when on aspirin, (iv) transition probability of fatal AE when on capecitabine, (v) relative risk of disease progression when on capecitabine, (vi) relative risk of disease progression when on aspirin, (vii) cost of 8 cycles of capecitabine, (viii) transition probability of non-fatal AE when on aspirin, and (ix) cost of metastatic CRC in recurrence state.
Extensive One-way Sensitivity Analysis.
Overall, the extensive one-way sensitivity analysis (File S3) showed that the results of the base case were not affected to a large extent over wide ranges of the variables identified in the initial one-way sensitivity analyses. In Stage I CRC, aspirin was found to be cheaper and more effective (i.e. dominant) than the no treatment strategy when one of the following conditions occurred: (i) utility of taking aspirin was 0.948 or more, (ii) annual probability of fatal aspirin-related AE did not exceed 0.03%, (iii) cost of care for metastatic CRC was more than USD7,200, (iv) utility score of staying in ‘Remission without intervention’ was 0 to 1, (v) relative risk of disease progression when on aspirin was 0 to 0.96, (vi) utility score of staying in ‘Remission with intervention’ was 0 to 1, (vii) annual probability of non-fatal AE when on aspirin was 16.8% or less. Aspirin was dominated (i.e. more expensive and less effective) by the no treatment strategy when the relative risk of disease progression when on aspirin was 0.997 or more. The no treatment strategy could be a cost-effective alternative when the utility of taking aspirin was 0.909 or less, or when the annual probability of fatal aspirin AE was 0.57% or more.
For Stage II CRC, aspirin was the dominant option when: (i) utility of taking aspirin was 0.959 or more, (ii) utility score of staying in ‘Remission with intervention’ was 0.311 or more, (iii) annual probability of fatal aspirin AE was 0.31% or less, (iv) annual probability of fatal capecitabine AE was 10% or less, (v) relative risk of disease progression when on capecitabine was 0.507 or more, (vi) relative risk of disease progression when on aspirin was 0.959 or less, (vii) cost of 8 cycles of capecitabine was USD0 to USD100,000, (viii) annual probability of non-fatal aspirin AE was 12% or less, or (ix) cost of care of metastatic CRC in recurrence state was USD9,000 or more.
The no treatment strategy was a cost-effective alternative in Stage II when utility of taking aspirin was less than 0.931, annual probability of fatal aspirin AE was 0.49% or more, or relative risk of disease progression when on aspirin was 0.976 or more. Capecitabine could be a cost-effective option when the relative risk of disease progression when on capecitabine was less than 0.228.
Multivariable Probabilistic Sensitivity Analyses.
Using the assigned distributions of the variables identified to have high levels of uncertainty for each stage (Table 3), the CE acceptability curves generated using multivariable PSA are shown in Figure 3.
In Stage I CRC, aspirin was consistently cost-effective about 70% to 80% of the time as compared to the no treatment strategy when willingness-to-pay was varied from USD20,000 to USD100,000. At USD0, the no treatment strategy could be cost-effective 80% of the time. However, it decreased steeply to about 30% when the willingness-to-pay was USD20,000. For Stage II CRC, when willingness-to-pay was likewise varied, aspirin was cost-effective at least 50% of the time when the threshold laid between USD20,000 to USD100,000. Similarly, the no treatment strategy quickly became cost-effective at about 45% of the time at a threshold of USD20,000 despite being cost-effective for more than 80% of the time when the willingness-to-pay was USD0. Capecitabine could be cost-effective at about 2% of the time throughout the range of threshold tested.
Discussion
Emerging evidence highlights some benefits of aspirin in several solid tumor cancers. In this first study of the hypothetical cost-effectiveness of aspirin in the adjuvant treatment of cancer, we found aspirin to be more cost-effective as compared to a no treatment strategy in Stage I and II CRC. Aspirin was also more cost-effective compared to capecitabine. Our PSA further showed aspirin to be cost-effective approximately 50% to 80% of the time in both stages when the willingness-to-pay threshold was between USD20,000 to USD100,000.
In our model, both no treatment and capecitabine were dominated by aspirin over wide ranges. However, no treatment or capecitabine (Stage II only) can be cost-effective alternatives in both stages if the utility of taking aspirin is below 0.909, aspirin’s annual fatal AE probability exceeds 0.57%, aspirin’s relative risk of disease progression is 0.997 or more, or when capecitabine’s relative risk of disease progression is less than 0.228.
Unlike capecitabine which has a well-defined regimen for use in Stage II CRC, there is a dearth of literature especially on the optimal dose and duration of aspirin therapy. In this study, we modeled aspirin to be a five-year therapy, covering the critical period where recurrence is most likely. Both aspirin and capecitabine were also assumed to exert their effects for the first five years of the simulation. Despite applying a utility of 0.999 for the period during which aspirin was taken in order to take into account the inconvenience of pill-taking, it did not bring about an appreciable difference. This is possibly due to the small margin of disutility assumed.
More recently, the cost of treatment of metastatic CRC has increased rapidly [35] with the incorporation of new biological treatments such as bevacizumab, cetuximab, panitumunab, aflibercept and regorafenib [36]–[38]. Thus an annual medical treatment cost of up to USD600,000 for metastatic CRC is no longer an obscure possibility. Additional analyses showed that aspirin remained cost-effective in both CRC stages even in an extreme scenario where the annual cost of care for recurrent metastatic CRC was USD6million.
We recognized that this study is not without limitations, mainly due to the uniform assumptions required. First, model inputs were estimated from a variety of sources. For example, indirect cost in the form of patient time was factored into cost inputs using certain wage and time estimates. These estimates did not include those incurred for AE and may not be generalizable to all CRC patients. Nevertheless, our findings in the sensitivity analyses remained similar over a wide range of estimates. Second, a number of the studies we drew data from, although consistent in their findings, were observational in nature [31], [32], [35], [39]. As such, limitations associated with observation studies (e.g. bias) would apply. Third, our model may be criticized for being overly simplified. However, given the paucity of data, a simpler model is probably more suited to the intended purpose of this study. In addition, our model did not permit individuals in the ‘Recurrence’ state to return to ‘Remission’ state even though this is clinically plausible [19]. However, as this happens only in a small number of CRC patients with liver metastases who could return to remission after surgical resection, it was not efficient to increase the complexity of the model to account for this low event probability.
In an attempt to give a conservative estimate of the cost-effectiveness of aspirin in CRC patients, potential cardio-protective and primary cancer prevention benefits of aspirin were not included. Although there is a recent study with a preliminary link of the use of aspirin to age-related macular degeneration [40], given the relatively rudimentary evidence and the small increase in risk, this effect was also not modeled. More recently, observational data has suggested that the tumor PIK3CA mutation or high tumor COX2 expression may serve as useful biomarkers for aspirin benefit [41]–[43]. Gene-expression analyses, although useful in prognosticating cancer relapses, have not yet been shown to predict adjuvant chemotherapy benefit. For these reasons, we have chosen to restrict our analysis to unselected CRC populations.
The National Cancer Institute has labeled aspirin’s activity in reducing CRC incidence and mortality as one of the most provocative questions in cancer [44], underscoring the importance and broad relevance of this treatment approach. Whereas primary cancer prevention with aspirin requires the treatment of large numbers of health individuals over prolonged periods of time, with toxicity and benefits finely balanced; aspirin’s ascendant role in the secondary prevention of resected cancers remains extremely attractive. Thus aspirin, if proven effective in prospective randomized trials is likely to play a unique role in the adjuvant treatment of Stage I and II cancers where large numbers of patients will need to be treated in order to prevent one cancer death. Our findings have two important implications. First, aspirin’s high cost-effectiveness in extremely low risk cancers alters the therapeutic paradigm of extremely low risks cancers and offers potential for adjuvant cancer treatment in a group of patients (i.e. Stage I CRC) that would currently undergo only observation. It supports a model of drug development away from traditional cytotoxics, where the risk of over-treatment is highest, towards repurposed old drugs such as aspirin. Second, the high cost-effectiveness of adjuvant aspirin underscores its broad social relevance to low income countries operating under constrained healthcare budgets. Lastly, the findings that aspirin is cost-effective even up to an extremely low therapeutic benefit ratio (i.e. a 1% relative risk reduction), draws attention to the difficulty in producing the requisite clinical evidence that is necessary to change clinical practice. A trial adequately powered for a hazard ratio of 0.99 in low risk cancer populations would require more than 300,000 subjects and would be impossibly expensive under existing development paradigms. Nonetheless, the potential benefits of aspirin as an adjuvant agent and its high cost-effectiveness justifies robust public support for research into its expanded use in the secondary prevention of cancer.
Supporting Information
File S2.
Transition Matrices of Stage I and II CRC.
https://doi.org/10.1371/journal.pone.0107866.s002
(DOCX)
File S3.
Model Input and Output of Extensive One-way Sensitivity Analyses.
https://doi.org/10.1371/journal.pone.0107866.s003
(DOCX)
Author Contributions
Conceived and designed the experiments: SSS WKC MSC HLW. Performed the experiments: SSS MSC. Analyzed the data: SSS WKC MSC HLW. Contributed reagents/materials/analysis tools: HLW. Wrote the paper: SSS WKC MSC HLW. Provided critical input on analysis: SSS WKC MSC GFH XJ YHD CST AS ES SM RA HCT HLW.
References
- 1.
Ferlay J, Shin HR, Bray F, Forman D, Mathers C, et al.. (2010) Cancer Incidence and Mortality Worldwide. Database: GLOBOCAN 2008 v2.0. Available: http://globocan.iarc.fr/. Accessed 2013 May 6.
- 2.
American Cancer Society (17 January 2013) Colorectal Cancer - What are the key statistics about colorectal cancer. Available: http://www.cancer.org/cancer/colonandrectumcancer/detailedguide/colorectal-cancer-key-statistics. Accessed 2013 May 6.
- 3. Sung JJY, Lau JYW, Goh KL, Leung WK (2005) Increasing incidence of colorectal cancer in Asia: implications for screening. The Lancet Oncology 6: 871–876.
- 4. Ponz de Leon M, Benatti P, Di Gregorio C, Fante R, Rossi G, et al. (2000) Staging and survival of colorectal cancer: are we making progress? The 14-year experience of a Specialized cancer Registry. Digestive and Liver Disease 32: 312–317.
- 5. Di Gregorio C, Benatti P, Losi L, Roncucci L, Rossi G, et al. (2005) Incidence and survival of patients with Dukes' A (stages T1 and T2) colorectal carcinoma: a 15-year population-based study. Int J Colorectal Dis 20: 147–154.
- 6. Ito Y, Nakayama T, Miyashiro I, Sugimoto T, Ioka A, et al. (2012) Trends in 'cure' fraction from colorectal cancer by age and tumour stage between 1975 and 2000, using population-based data, Osaka, Japan. Jpn J Clin Oncol 42: 974–983.
- 7. Moertel CG, Fleming TR, Macdonald JS, Haller DG, Laurie JA, et al. (1995) Fluorouracil plus levamisole as effective adjuvant therapy after resection of stage III colon carcinoma: a final report. Ann Intern Med 122: 321–326.
- 8. Figueredo A, Charette ML, Maroun J, Brouwers MC, Zuraw L (2004) Adjuvant therapy for stage II colon cancer: a systematic review from the Cancer Care Ontario Program in evidence-based care's gastrointestinal cancer disease site group. Journal of Clinical Oncology 22: 3395–3407.
- 9. Quasar Collaborative Group (2007) Adjuvant chemotherapy versus observation in patients with colorectal cancer: a randomised study. The Lancet 370: 2020–2029.
- 10. Chia WK, Ali R, Toh HC (2012) Aspirin as adjuvant therapy for colorectal cancer—reinterpreting paradigms. Nat Rev Clin Oncol 9: 561–570.
- 11. Ladabaum U, Chopra CL, Huang G, Scheiman JM, Chernew ME, et al. (2001) Aspirin as an Adjunct to Screening for Prevention of Sporadic Colorectal Cancer, a Cost-Effectiveness Analysis. Ann Intern Med 135: 769–781.
- 12. Hur C, Simon LS, Gazelle GS (2004) The cost-effectiveness of aspirin versus cyclooxygenase-2-selective inhibitors for colorectal carcinoma chemoprevention in healthy individuals. Cancer 101: 189–197.
- 13. Suleiman S, Rex DK, Sonnenberg A (2001) Chemoprevention of Colorectal Cancer by Aspirin: A Cost-effectiveness Analysis. Gastroenterology 122: 78–84.
- 14. Weitz J, Koch M, Debus J, Höhler T, Galle PR, et al. (2005) Colorectal cancer. The Lancet 365: 153–165.
- 15. Twelves C, Wong A, Nowacki MP, Abt M, Burris H 3rd, et al. (2005) Capecitabine as Adjuvant Treatment for Stage III Colon Cancer. New England Journal of Medicine 352: 2696–2704.
- 16.
Surveillance Epidemiology and End Results Program (2013) Incidence. Database: SEER*Stat SEER 18 Regs Research Data + Hurricane Katrina Impacted Louisiana Cases, Nov 2012 Sub (1973–2010 varying) - Linked To County Attributes - Total U.S., 1969–2011 Counties. Available: http://www.seer.cancer.gov.Accessed 2013 November 30.
- 17.
National Center for Health Statistics (2013) Health, United States, 2012, with Special Feature on Emergency Care. National Centre for Health Statistics. Available: www.cdc.gov/nchs/data/hus/hus12.pdf. Accessed 2013 June 18.
- 18. Chan AT, Ogino S, Fuchs CS (2009) Aspirin use and survival after diagnosis of colorectal cancer. JAMA 302: 649–658.
- 19.
Mant D, Perera R, Gray A, Rose P, Fuller A, et al.. (2013) Effect of 3–5 years of scheduled CEA and CT follow-up to detect recurrence of colorectal cancer: FACS randomized controlled trial. J Clin Oncol 31: abstr 3500.
- 20. Hernandez-Diaz S, Garcia Rodriguez LA (2006) Cardioprotective aspirin users and their excess risk of upper gastrointestinal complications. BMC Med 4: 22.
- 21. United States Preventive Services Task Force (2009) Aspirin for the Prevention of Cardiovascular Disease: U.S. Preventive Services Task Force Recommendation Statement. Annals of Internal Medicine 150: 396–404.
- 22.
Zhao Y, Encinosa W (2008) Hospitalizations for Gastrointestinal Bleeding in 1998 and 2006, HCUP Statistical Brief #65. Agency for Healthcare Research and Quality. Available: http://www.hcup-us.ahrq.gov/reports/statbriefs/sb65.pdf. Accessed 2013 June 18.
- 23. Scheithauer W (2003) Oral capecitabine as an alternative to i.v. 5-fluorouracil-based adjuvant therapy for colon cancer: safety results of a randomized, phase III trial. Annals of Oncology 14: 1735–1743.
- 24. Ramsey SD, Andersen MR, Etzioni R, Moinpour C, Peacock S, et al. (2000) Quality of Life in Survivors of Colorectal Carcinoma. Cancer 88: 1294–1303.
- 25. Naglie IG, Detsky AS (1992) Treatment of Chronic Nonvalvular Atrial Fibrillation in the Elderly: A Decision Analysis. Medical Decision Making 12: 239–249.
- 26. Fontana M, Asaria P, Moraldo M, Finegold J, Hassanally K, et al. (2014) Patient-Accessible Tool for Shared Decision Making in Cardiovascular Primary Prevention: Balancing Longevity Benefits Against Medication Disutility. Circulation 129: 2539–2546.
- 27.
Centers for Medicare and Medicaid Services (2013) Overview of Physician Fee Schedule Search. Database: Physician Fee Schedule. Available: http://www.cms.gov/apps/physician-fee-schedule/. Accessed 2013 May 16.
- 28.
Agency for Healthcare Research and Quality (2014) HCUPnet - a brief description of HCUPnet definitions. Agency for Healthcare Research and Quality. Available: http://hcupnet.ahrq.gov/HCUPnet.jsp?Id=F1C69BCCCE6C025F&Form=Help&JS=Y&Action=%3E%3ENext%3E%3E&HCUPnet%20definitions.x=1. Accessed 2014 April 30.
- 29.
United States Department of Labor (2013) U.S. Bureau of Labor Statistics - Consumer Price Index. Available: http://www.bls.gov/cpi/tables.htm. Accessed 2014 January 1.
- 30. Gravelle H, Smith D (2001) Discounting for health effects in cost-benefit and cost-effectiveness analysis. Health Econ 10: 587–599.
- 31. Chu E, Shi N, Wei W, Bendell JC, Cartwright T (2009) Costs associated with capecitabine or 5-fluorouracil monotherapy after surgical resection in patients with colorectal cancer. Oncology 77: 244–253.
- 32. Yabroff KR, Warren JL, Knopf K, Davis WW, Brown ML (2005) Estimating Patient Time Costs Associated with Colorectal Cancer Care. Medical Care 43: 640–648.
- 33. Hunt TL LB, Page MJ, Pokrzywinski R (2009) Willingness to Pay for Cancer Prevention. Pharmacoeconomics 27: 299–312.
- 34.
Briggs A, Claxton K, Sculpher M (2006) Decision Modelling for Health Economic Evaluation. Great Britain: Oxford University Press. 237 p.
- 35. Song X, Zhao Z, Barber B, Gregory C, Cao Z, et al. (2011) Cost of illness in patients with metastatic colorectal cancer. J Med Econ 14: 1–9.
- 36. Grothey A, Barone C, Adenis A, Tabernero J, Yoshino T, et al. (2013) Regorafenib monotherapy for previously treated metastatic colorectal cancer (CORRECT): an international, multicentre, randomised, placebo-controlled, phase 3 trial. Lancet 381: 303–312.
- 37. Tol J, Punt CJA (2010) Monoclonal antibodies in the treatment of metastatic colorectal cancer: A review. Clinical Therapeutics 32: 437–453.
- 38. Douillard J-Y, Jassem J, Rivera F, Kocákova I, Ruff P, et al. (2010) Randomized, phase III trial of panitumumab with infusional fluorouracil, leucovorin, and oxaliplatin (FOLFOX4) versus FOLFOX4 alone as first-line treatment in patients with previously untreated metastatic colorectal cancer: the PRIME study. Journal of Clinical Oncology 28: 4697–4705.
- 39. Manfredi S, Bouvier AM, Lepage C, Hatem C, Dancourt V, et al. (2006) Incidence and patterns of recurrence after resection for cure of colonic cancer in a well defined population. Br J Surg 93: 1115–1122.
- 40. Liew G, Mitchell P, Wong TY, Rochtchina E, Wang JJ (2013) The association of aspirin use with age-related macular degeneration. JAMA Internal Medicine 173: 258.
- 41. Kelley RK, Venook AP (2011) Prognostic and Predictive Markers in Stage II Colon Cancer: Is There a Role for Gene Expression Profiling? Clinical Colorectal Cancer 10: 73–80.
- 42. Chan AT, Ogino S, Fuchs CS (2007) Aspirin and the Risk of Colorectal Cancer in Relation to the Expression of COX-2. The New England Journal of Medicine 356: 2131–2142.
- 43. Liao X, Shima K, Sun R, Nosho K, Meyerhardt JA, et al. (2012) Aspirin use, tumor PIK3CA mutation, and colorectal-cancer survival. The New England Journal of Medicine 367: 1596–1606.
- 44.
National Cancer Institute (2013) Provocative Questions - Identifying Perplexing Problems to Drive Progress Against Cancer. National Cancer Institute. Available: http://provocativequestions.nci.nih.gov/. Accessed 2013 August 31.
- 45.
Centers for Medicare and Medicaid Services (2013) Fee Schedule. Database: Fee Schedule. Available: http://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/ClinicalLabFeeSched/clinlab.html. Accessed 2013 May 16.
- 46. Pence BC, Belasco EJ, Lyford CP (2013) Combination aspirin and/or calcium chemoprevention with colonoscopy in colorectal cancer prevention: cost-effectiveness analyses. Cancer Epidemiol Biomarkers Prev 22: 399–405.
- 47.
United States Department of Labor (2013) Real earnings in November 2013. Bureau of Labor Statistics. Available: http://www.bls.gov/schedule/archives/realer_nr.htm. Accessed 2014 January 1.
- 48. Gorelick PB, Weisman SM (2005) Risk of hemorrhagic stroke with aspirin use: an update. Stroke 36: 1801–1807.
- 49. Rosamond WD, Folsom AR, Chambless LE, Wang CH, McGovern PG, et al. (1999) Stroke Incidence and Survival Among Middle-Aged Adults: 9-Year Follow-Up of the Atherosclerosis Risk in Communities (ARIC) Cohort. Stroke 30: 736–743.
- 50.
Genentech USA Inc (2011) Highlights of Prescribing Information - XELODA (capecitabine). Genentech USA, Inc. Available: www.gene.com/download/pdf/xeloda_prescribing.pdf. Accessed 2013 May 14.
- 51. Benns M, Carr B, Kallan MJ, Sims CA (2013) Benchmarking the incidence of organ failure after injury at trauma centers and nontrauma centers in the United States. J Trauma Acute Care Surg 75: 426–431.
- 52.
Agency for Healthcare Research and Quality (2014) Statistics for U.S. community hospital stays, principal diagnosis based on CCS (Clinical Classifications Software) 2011. Agency for Healthcare Research and Quality. Available: http://hcupnet.ahrq.gov/.Accessed 2014 April 30.
- 53.
USA Social Security Administration (2012) Retirement Planner: Benefits By Year Of Birth. USA Social Security Administration. Available: http://www.socialsecurity.gov/retire2/agereduction.htm. Accessed 2013 May 15.
- 54.
Surveillance Epidemiology and End Results Program (2013) Incidence. Database: SEER*Stat SEER 9 Regs Research Data, Nov 2012 Sub (1973–2010) <Katrina/Rita Population Adjustment> - Linked To County Attributes - Total U.S., 1969–2011 Counties. Available: http://www.seer.cancer.gov. Accessed 2013 November 30.
- 55. McQuaid KR, Laine L (2006) Systematic review and meta-analysis of adverse events of low-dose aspirin and clopidogrel in randomized controlled trials. Am J Med 119: 624–638.
- 56.
Levit K, Ryan K, Elixhauser A, Stranges E, Kassed C, et al.. (2007) Healthcare Cost and Utilization Project (HCUP) Facts and Figures: Statistics on Hospital-based Care in the United States in 2005. Agency for Healthcare Research and Quality. Available: http://www.hcup-us.ahrq.gov/reports/factsandfigures/figures/2005/2005_2_4.jsp. Accessed 2013 May 16.