In his 2015 State of the Union Address, President Obama launched the era of precision medicine, which proposes medical decisions and practices tailored to the individual patient. While the concept of precision medicine is not new, what is new is the broader scope in which precision medicine will eventually apply. Broader application of precision medicine will soon be possible because of new and powerful tools of molecular biology, genomics and bioinformatics.
As appealing as it seems, precision medicine will complicate the delivery of health care. Providers will need to consider much more information than they currently do in order to personalize prevention, diagnosis, treatment and prognosis. Simple is the example of how a single genetic mutation (in the KRAS gene) determines whether anti-EGF therapy is useful in the treatment of advanced stage colorectal cancer (CRC). Broader use of precision medicine will likely require consideration of combinations of several genetic, proteomic, phenotypic and other features when attempting to optimize whether and how an individual patient is screened for a particular disease, or when deciding on a diagnostic test or drug for a suspected or established diagnosis, with the goal of maximizing clinical benefit, minimizing risk of adverse effects or both. Although the near-term focus of precision medicine is on tailoring cancer treatment based largely on genetic and molecular profiling, can concepts of it apply now to screening for CRC?
Precision medicine is currently present to a modest extent in CRC prevention. The American College of Physicians recommends that “clinicians perform individualized assessment of risk for colorectal cancer in all adults,”1 but this process is limited to consideration of inflammatory bowel disease, genetic polyposis and non-polyposis syndromes, and “family history.” Genetic biomarkers such as mutant APC, the most common genetic cause of FAP, or mutant MLH1 or MSH2, the most common genetic causes for Lynch syndrome, have a profound effect on CRC risk and on screening and treatment decisions, but affect no more than 3 percent of the population. Among the remainder, only family history is used to determine when to begin and how to screen for CRC, despite its only fair discrimination as a risk factor.
The remaining 85 to 90 percent of the population is indiscriminately and unhelpfully labeled as “average risk.” Several phenotypic (age, sex and obesity) and behavioral (smoking, diet and physical activity) factors are well-established risk factors for CRC, but these are not used for making decisions about screening despite their well-known effects. For example, neither age nor sex is used, despite the fact that CRC risk nearly doubles each decade between ages 50 and 80, and that a man’s CRC risk is nearly twice that of a woman’s at any age. Among the reasons given for this non-use are absence of a way to consider multiple risk factors simultaneously, lack of time for provider patient discussion about the tradeoffs among screening strategies, and the fear of further “complicating” screening recommendations, confusing providers and patients.
Current screening recommendations in the U.S. are anything but complicated. They may be distilled to this: in the absence of a highrisk family history, men and women should be screened starting at age 50 with either a colonoscopy every 10 years or a high-sensitivity stool occult blood test annually. These two screening strategies are recommended equally; consideration of patient preference and availability is advised. And while other tests and strategies are “recommended” by guideline organizations, these are rarely used in practice because of poor availability and reimbursement. In the U.S., colonoscopy is so dominant that much of the average-risk population is unaware of any other option. The population and most providers are unaware of the equipoise between stool occult blood and colonoscopy strategies. Perhaps it is this “colonoscopy-only” mindset that keeps U.S. screening adherence at 60 to 65 percent rather than a higher rate. Giving people the option of non-invasive screening increases uptake.2While a patient-provider discussion of screening options makes good sense, that discussion would be more informative and more “personalized” if patient-specific risk for CRC or advanced neoplasia were included.
In addition to already-in-place computer-based technologies for access and efficiency, there are tools available that consider several factors simultaneously to estimate current risk for advanced neoplasia3-12 or future risk for CRC.13 These multivariable models were developed and validated with comparable methods and on large numbers of average-risk persons from eastern Asia,3 China,4-5, Korea, 6 Germany,7 Spain,8 Poland 9 and the U.S.,10-13 the majority of whom had no previous screening. It is important to realize that these tools are best applied to initial, not subsequent, screening, as prior test results – either positive or negative – affect subsequent risk. All models include age and sex; many of them include cigarette smoking, family history of CRC, and a physical measure (body mass index or waist circumference), suggesting that these variables are both valid and generalizable. Model performance — calibration, discrimination, and the degree of clinically-significant risk stratification — varies and requires clinical judgment about whether the risk estimates are clinically important and distinct enough to affect the choice of screening strategy. For example, one model’s low risk group had a current risk for advanced neoplasia of 1.9 percent in derivation (1.65 percent in validation), while the high-risk group had advanced neoplasia risks of 24.9 percent (derivation) and 22.3 percent (validation).11 The low-risk estimate would seem low enough to recommend non-invasive screening, and to do so supportively. On the other hand, the high risk estimate seems high enough to recommend colonoscopy. Many patients and providers might agree with this application of the use of risk score results while others may not.
Personalizing screening recommendations by including risk of advanced neoplasia is potentially useful for several reasons. First, it provides information that can help guide a discussion about which strategy may be more appropriate for an individual patient, given no prior preference. Second, it improves the balance between benefits and risks. A program of occult blood testing for low-risk persons can help them either avoid colonoscopy completely (provided that testing remains negative) or defer colonoscopy until non-invasive testing is positive. And while programmatic noninvasive screening is recommended for average-risk persons, it perhaps could be recommended with greater support or enthusiasm for those “average-risk” persons who are actually low risk. A recommendation for non-invasive screening could increase uptake among previously unscreened low-risk persons. For high-risk persons, personalizing their risk for advanced neoplasia might convince them of the need for colonoscopy sooner rather than later, avoiding a delay in detection due to the misperception that they are “average-risk” or possibly due to one or more rounds of falsely negative non-invasive testing. Third, it improves the yield and efficiency of screening colonoscopy, optimizing use of an expensive and limited resource that is not without risk.
Personalized risk information allows providers to make a supportive recommendation for noninvasive screening or a more data-driven one for colonoscopy.
Models for CRC screening are meant to aid clinical judgment and decision making, not substitute for them. Like all tools in gastroenterology, none of the models are perfect, but all have shown some degree of scientific rigor. The best performing ones — determined by both model metrics and clinical judgment — can and should be used in persons/ populations comparable to those used in model development and validation. Personalized risk information may engage patients and providers into a discussion that considers risk and preferences, allows providers to make a supportive recommendation for noninvasive screening or a more data-driven one for colonoscopy. After all, we are in the era of precision medicine, where several domains of risk factors will soon require consideration. If we can’t handle individualizing screening decisions based on a few phenotypic and behavioral features (along with patient preferences), how will we ever handle fully implemented precision medicine?
Dr. Imperiale has no conflicts to disclose.
1.Qaseem A, Denberg TD, Hopkins RH, et al. Screening for colorectal cancer: a guidance statement from the American College of Physicians. Annals of internal medicine. 2012;156(5):378-386.
2. Inadomi JM, Vijan S, Janz NK, et al. Adherence to colorectal cancer screening: a randomized clinical trial of competing strategies. Archives of internal medicine. 2012;172(7):575-582.
3. Yeoh KG, Ho KY, Chiu HM, et al. The Asia-Pacific Colorectal Screening score: a validated tool that stratifies risk for colorectal advanced neoplasia in asymptomatic Asian subjects. Gut.2011;60(9):1236-1241.
4. Cai QC, Yu ED, Xiao Y, et al. Derivation and validation of a prediction rule for estimating advanced colorectal neoplasm risk in averagerisk Chinese. American journal of epidemiology.2012;175(6):584-593.
5. Wong MC, Lam TY, Tsoi KK, et al. A validated tool to predict colorectal neoplasia and inform screening choice for asymptomatic subjects. Gut. 2014;63(7):1130-1136.
6. Kim DH, Cha JM, Shin HP, Joo KR, Lee JI, Park DI. Development and validation of a risk stratification-based screening model for predicting colorectal advanced neoplasia in Korea. Journal of clinical gastroenterology. 2015;49(1):41-49.
7. Tao S, Hoffmeister M, Brenner H. Development and validation of a scoring system to identify individuals at high risk for advanced colorectal neoplasms who should undergo colonoscopy screening. Clinical Gastroenterology and Hepatology: the official clinical practice journal of the American Gastroenterological Association. 2013.
8.Betés M, Muñoz-Navas MA, Duque JM, et al. Use of colonoscopy as a primary screening test for colorectal cancer in average risk people. Am J Gastroenterol. 2003;98(12):2648-2654.
9. Kaminski MF, Polkowski M, Kraszewska E, Rupinski M, Butruk E, Regula J. A score to estimate the likelihood of detecting advanced colorectal neoplasia at colonoscopy. Gut. 2014;63(7):1112-1119.
10. Lin OS, Kozarek RA, Schembre DB, et al. Risk stratification for colon neoplasia: screening strategies using colonoscopy and computerized tomographic colonography. Gastroenterology. 2006;131(4):1011-1019.
11. Schroy PC, Wong JB, O’Brien MJ, Chen CA, Griffith JL. A Risk Prediction Index for Advanced Colorectal Neoplasia at Screening Colonoscopy. The American journal of gastroenterology. 2015.
12. Imperiale TF, Monahan PO, Stump TE, Glowinski EA, Ransohoff DF. Derivation and Validation of a Scoring System to Stratify Risk for Advanced Colorectal Neoplasia in Asymptomatic Adults: A Cross-sectional Study. Ann Intern Med. 2015;163(5):339-346.
13. Freedman AN, Slattery ML, Ballard-Barbash R, et al. Colorectal cancer risk prediction tool for white men and women without known susceptibility. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2009;27(5):686-693.