Genetic testing for predisposition to cancer is now possible for many hereditary syndromes, including hereditary breast and ovarian cancer, multiple endocrine neoplasia, von Hippel-Lindau disease, familial adenomatous polyposis, and hereditary nonpolyposis colorectal cancer. Hereditary nonpolyposis colorectal cancer, the most common hereditary colon cancer syndrome [4], is a dominantly inherited disease that is estimated to account for approximately 2% to 5% of colorectal cancers [5, 6]. The isolation of several DNA mismatch repair genes (hMSH2, hMLH1, hPM2, and hMSH6) associated with hereditary nonpolyposis colorectal cancer has made it possible to identify carriers of a mutated gene within a family [7-11].

On the basis of literature review and expert opinion, investigators from the National Human Genome Research Institute Cancer Genetics Studies Consortium recently recommended that carriers of a mutation for hereditary nonpolyposis colorectal cancer undergo regular colonoscopic surveillance every 1 to 3 years [12]. However, because of the high risk for cancer and metachronous tumors and the limitations of surveillance, the panel recommended that subtotal colectomy with ileorectal anastomosis be considered in patients with colon cancer or adenomas and total proctocolectomy be considered in those with rectal cancer [12].

A definitive answer to the optimal management approach for carriers of a mutation will require clinical studies of large numbers of patients followed prospectively over a long time. However, with the increasing use of genetic predisposition testing, patients and health care providers must decide on cancer risk-reduction strategies now. This decision involves weighing several factors, including the high risk for cancer; surgical and procedure-associated risks; and the health-related quality of life associated with prophylactic colectomy, endoscopic surveillance, and having cancer. We used a decision analysis to assess the outcomes of various colorectal cancer preventive strategies, including colonoscopic surveillance, prophylactic colectomy, and delayed colectomy for patients carrying a genetic mutation associated with hereditary nonpolyposis colorectal cancer.

Methods

Top Methods Results Discussion Author & Article Info References

Model Structure

We constructed a Markov model by using the decision analysis program DATA (TreeAge Software, Inc., Williamstown, Massachusetts) to compare different approaches to colorectal cancer prevention in patients with a mutation for hereditary nonpolyposis colorectal cancer. The model follows a hypothetical cohort of patients over time and tracks the annual incidence of polyps and colorectal cancer (by stage), progression of colorectal cancer, and mortality (Figure 1). Short-term mortality associated with surgery and endoscopic surveillance was also incorporated.

View larger version (15K): [in this window] [in a new window]   Figure 1. General model framework. Schematic overview of Markov model. A cohort of 25-year-old carriers of a mutation for hereditary nonpolyposis colorectal cancer who are cancer-free enter the model. Each circle represents states of health within the model. Each of the four main categories, depicted in bold, contains several subsets that depend on whether the patient is undergoing surveillance (surveillance status), undergoes colectomy (colon status), or has a polyp diagnosed and on the patient's stage of cancer (if colorectal cancer has occurred). The arrows represent potential transitions between health states and transition probabilities that vary depending on the strategy. Each year, a patient may remain in the same health state, die, or move to another health state, as shown in the flow diagrams.

Standard therapy for average-risk patients with colorectal cancer is segmental resection of the involved colon and appropriate stage-specific therapy for advanced disease. More aggressive prophylactic colonic resection as a cancer prevention method is the only management option for patients with familial adenomatous polyposis, another group at high risk for colorectal cancer. We evaluated two types of colectomy for carriers of a mutation for hereditary nonpolyposis colorectal cancer: proctocolectomy with ileoanal anastomosis, which was assumed to eliminate all risk for colorectal cancer and the need for postoperative surveillance, and subtotal colectomy with ileorectal anastomosis, which required continued surveillance of the remaining rectal segment.

It has been recommended that at-risk patients from families with hereditary nonpolyposis colorectal cancer begin surveillance at 25 years of age or when they are 5 years younger than the youngest age at which cancer was diagnosed in a family member [5]. Because our analysis deals with mutation carriers, we adopted the more aggressive surveillance approach and assumed that surveillance began at 25 years of age. Surveillance was defined as colonoscopy every 3 years if no surgical intervention had been performed [13] and flexible sigmoidoscopy of the remaining rectal segment every 3 years after subtotal colectomy. Surveillance was assumed to stop at 85 years of age. All colorectal cancer was assumed to arise from colorectal adenomas [14-18], and adenomas detected by surveillance were assumed to be removed at the time of colonoscopy. The incidence of adenomas was assumed to be the same as that of cancer except that adenomas could be detected by colonoscopy 5 years before cancer was detected [19].

We examined the following 12 strategies for a cohort of 25-year-old patients who are cancer-free and have a known mutation for hereditary nonpolyposis colorectal cancer: immediate prophylactic proctocolectomy, immediate prophylactic subtotal colectomy, surveillance until 40 years of age followed by prophylactic proctocolectomy, surveillance until 40 years of age followed by subtotal colectomy, surveillance until 50 years of age followed by prophylactic proctocolectomy, surveillance until 50 years of age followed by subtotal colectomy, surveillance and proctocolectomy if an adenomatous polyp is found, surveillance and subtotal colectomy if an adenomatous polyp is found, surveillance and proctocolectomy if invasive cancer is diagnosed, surveillance and subtotal colectomy if invasive cancer is diagnosed, surveillance and segmental resection if invasive cancer is diagnosed, and no surveillance and segmental resection if invasive cancer is diagnosed. Figure 2 shows schematic representations of 2 of these strategies. Model outcomes were life expectancy and quality-adjusted life expectancy.

View larger version (24K): [in this window] [in a new window]   Figure 2. Schematic representation of surveillance and colectomy if adenoma is found. The patient enters the Markov tree (the encircled letter M), which depicts the clinical events that may occur during each 1-year period as a patient is followed until death. Circles represent chance nodes. Because surveillance occurs every 3 years, a patient may or may not undergo surveillance in a given year. Colorectal cancer may be diagnosed on the basis of symptoms or at the time of colonoscopy. Death may occur as a result of colorectal cancer, surgical or surveillance procedures, or causes not related to colorectal cancer. Top. Surveillance. If colorectal cancer develops, segmental resection of the cancer is performed; the patient must continue to undergo surveillance and is at risk for a second primary tumor. Bottom. Surveillance and colectomy if adenoma is found. If colorectal cancer or an adenoma develops, colectomy is performed. If subtotal colectomy is performed, the patient must continue to undergo surveillance and is at risk for a second primary tumor. If total proctocolectomy with ileoanal anastomosis is performed, the risk for a second primary cancer is eliminated and continued surveillance is not necessary. Other colectomy strategies examined follow similar paths, with variations of timing of colectomy based on the particular strategy being evaluated.

Data Sources and Assumptions

The probabilities and sources used in the model are listed in Table 1. The risk for colorectal cancer in mutation carriers was based on observed risks of colorectal cancer in a study of 210 carriers of an hMSH2 or hMLH1 mutation [20]. Mean risks for colorectal cancer among men and women in this study and in our baseline analysis were 31.5% by 40 years of age, 54.5% by 50 years of age, 63.0% by 60 years of age, and 87.5% by 75 years of age [20]. Age-specific incidence rates beyond 75 years of age were based on data from the National Cancer Institute Surveillance, Epidemiology and End Results (SEER) Program [21], resulting in a lifetime risk for colorectal cancer of 88.2%. The lifetime risk for a second primary colorectal cancer was 45% [38]. Colorectal cancers in patients with hereditary nonpolyposis colorectal cancer who do not undergo regular surveillance were assumed to follow the same stage distribution as population rates, obtained from the SEER Program: 39% for localized cancer, 40% for regional cancer, and 21% for distant cancer [21].

Because no long-term follow-up studies of patients with hereditary nonpolyposis colorectal cancer undergoing colectomy have been published, we relied on studies of patients with familial adenomatous polyposis. Long-term follow-up of patients with familial adenomatous polyposis who have undergone proctocolectomy has shown that colorectal cancer in persons who were cancer-free at the time of surgery is rare [39-41]. In theory, prophylactic proctocolectomy eradicates all large-bowel mucosa at risk for colorectal cancer [42]. A prophylactic proctocolectomy was thus assumed to be 100% protective against colorectal cancer in the baseline analysis. The risk for colorectal cancer after subtotal colectomy with ileorectal anastomosis in patients with familial adenomatous polyposis ranges from 3.6% to 32% [22-25]. On the basis of these studies, prophylactic subtotal colectomy was assumed to result in an 80% reduction in lifetime risk for colorectal cancer in carriers of a mutation for hereditary nonpolyposis colorectal cancer (resulting in a lifetime risk for colorectal cancer of 17.5%). The stage-specific mortality rate associated with colorectal cancer was equal to that associated with resection alone or colectomy at the time of cancer diagnosis [39]; however, the risk for a second primary colorectal cancer was reduced by 80% with subtotal colectomy and by 100% with proctocolectomy. We assumed a perioperative mortality rate of 0.4% to 5.6% (depending on age) for persons undergoing prophylactic colectomy [26-28] and 0.8% to 11.2% (depending on age) for surgical resection of colorectal cancer [28, 29].

The largest prospective, controlled study of hereditary nonpolyposis colorectal cancer found that surveillance resulted in a 62% decrease in risk for colorectal cancer [30]. On the basis of this information, we used a constant percentage reduction each year so that the predicted lifetime risk was reduced by 62%. All colorectal cancer was assumed to be diagnosed at a localized stage (Dukes stage A or B1), as was the case in the largest trial of colonoscopic surveillance in the general population (the National Polyp Study [13, 31]) and in patients with hereditary nonpolyposis colorectal cancer who undergo surveillance [30, 33]. The mortality rate associated with endoscopic surveillance was 0.02%, based on several large studies of complications associated with endoscopic procedures [34-37].

Age- and sex-specific mortality estimates were obtained from U.S. life tables [43]. In our baseline analysis, we assumed that stage-specific colorectal cancer mortality was the same for hereditary nonpolyposis colorectal cancer and for sporadic colorectal cancer. Five-year mortality rates, obtained from the SEER Program, for localized, regional, and distant colorectal cancer were 8%, 36%, and 93%, respectively [21]. The annual mortality rate after 5 years was assumed to be the same as the stage-specific mortality rate of the fifth year. Reports suggest that carriers of mutations for hereditary nonpolyposis colorectal cancer may have improved stage-for-stage survival compared with patients who have sporadic colorectal cancer [44, 45]. We elected to perform the baseline analysis by using SEER data and address the possibility of improved survival in sensitivity analyses.

To examine the effects of health-related quality of life on the decision between prophylactic colectomy and surveillance, we incorporated adjustments for quality of life associated with various states of health in the analysis. Utilities are measures of the strengths of preferences of a patient for various health outcomes. To get a range of reasonable utility estimates for the major health states in question, we interviewed a sample of 10 health care providers (7 gastroenterologists and 3 genetic counselors) familiar with the health states relevant to the decision facing patients with hereditary nonpolyposis colorectal cancer: subtotal colectomy; proctocolectomy; and localized, regional, or distant colorectal cancer. To derive these estimates, we used the standard gamble approach [46], in which the respondent was asked to choose between life in a given clinical state and a gamble between death (assigned a value of 0) and perfect health (assigned a value of 1). The standard gamble elicits a measure of the respondent's preference for that health state, ranging from 0 to 1.

The mean quality-adjustment weights obtained from the survey were used in the baseline analysis. These weights were 0.95 for subtotal colectomy, 0.94 for localized or regional cancer, 0.89 for proctocolectomy, and 0.56 for distant colorectal cancer. Undergoing endoscopic surveillance was assigned a disutility of 3 days: That is, because of the procedure and preparation for or anxiety about the procedure, the patient was assumed to have 0 quality of life for 3 days.

Sensitivity Analyses

Sensitivity analysis is a test of the stability of the conclusions of an analysis over a range of structural assumptions, probability estimates, or value judgments [47]. Sensitivity analyses were performed on all variables in the model over plausible ranges (Table 1) to examine how changing one or several of the baseline values might influence the results.

Role of the Funding Source

The funding source had no role in the collection, analysis, or interpretation of data or in the decision to submit the manuscript for publication.

Results

Top Methods Results Discussion Author & Article Info References

Benefits of Colorectal Cancer Prevention Strategies

Immediate prophylactic colectomy results in the greatest life expectancy for a 25-year-old mutation carrier. Compared with no surveillance, immediate prophylactic proctocolectomy leads to a life expectancy benefit of 15.6 years and immediate subtotal colectomy leads to a life expectancy benefit of 15.3 years (Table 2). Compared with surveillance, the life expectancy benefits derived from prophylactic colectomy are 2.1 years with proctocolectomy and 1.8 years with prophylactic subtotal colectomy. For a 25-year-old person who would not consider colectomy as a preventive option, surveillance leads to a life expectancy gain of 13.5 years compared with no surveillance.

The benefit of colectomy compared with surveillance decreases as the procedure is delayed to later in life (life expectancy gains of 0.7 and 0.4 years for proctocolectomy at 40 years of age and 50 years of age, respectively). The life expectancy gain of colectomy at diagnosis of colorectal cancer compared with continued surveillance is 6 days for proctocolectomy and 4 days for subtotal colectomy.

If surgery is performed at 25 years of age, proctocolectomy provides a life expectancy gain of 0.3 years compared with subtotal colectomy. If colectomy is performed when colorectal cancer is diagnosed, the life expectancy gain of performing proctocolectomy compared with subtotal colectomy is 2 days.

Discounting is a method to account for the fact that health benefits that occur today may be valued more than those that occur in the future [47]. The application of a 3% annual discount rate [48] led to a decrease in the magnitude of benefits gained from all strategies. Immediate prophylactic colectomy strategies led to 0.5 to 0.6 years of life gained compared with surveillance; however, the benefit for all delayed colectomy strategies compared with surveillance was less than 5 weeks when future years were discounted.

A comparison of the strategies in terms of quality-adjusted life expectancy is shown in Table 2. Surveillance leads to the greatest quality-adjusted life expectancy compared with all colectomy strategies. Surveillance led to a gain of 14.0 quality-adjusted life-years (QALYs) compared with no surveillance, 3.1 QALYs compared with immediate proctocolectomy, and 0.3 QALYs compared with immediate subtotal colectomy. Incorporation of quality adjustments resulted in greater quality-adjusted life expectancies for all subtotal colectomy strategies compared with proctocolectomy strategies, with benefit ranging from 0.3 QALYs if colectomy was performed when colorectal cancer was diagnosed to 2.8 QALYs if colectomy was performed at 25 years of age.

Sensitivity Analyses

The model was most sensitive to assumptions about the efficacy of surveillance in terms of reducing colorectal cancer incidence or changing stage distribution and utilities associated with colectomy and surveillance. The effects of variation of the most influential probabilities in the baseline and quality-adjusted analyses are shown in Table 3.

In the baseline analysis, surveillance plus polypectomy was assumed to prevent 62% of colorectal cancer cases. We varied the estimates of efficacy of surveillance in reducing the incidence of colorectal cancer from 50% to 85% and found that the benefit of prophylactic proctocolectomy at 25 years of age ranged from 2.8 years to 0.7 years. The life expectancy benefit of prophylactic proctocolectomy decreased with increasing age and was 0.4 years if this procedure was performed at 55 years of age. Performing colectomy at the time of colorectal cancer diagnosis never led to a significant life expectancy benefit over continued surveillance.

Results of the quality-adjusted life expectancy analysis were also sensitive to assumptions about surveillance effectiveness. Surveillance leads to the greatest quality-adjusted life expectancy unless it is less than 57% effective at decreasing cancer incidence or colorectal cancer is diagnosed at a localized stage in less than 85% of cases. If so, the greatest quality-adjusted life expectancy is achieved by undergoing subtotal colectomy at 25 years of age.

Results of the quality-adjusted analysis were also sensitive to the utility associated with colectomy (Table 4). Surveillance was the preferred strategy unless the utility for subtotal colectomy was 0.96 or more or the utility of proctocolectomy was 0.95 or more. If utilities for either procedure exceed these thresholds, the greatest quality-adjusted life expectancy is achieved by prophylactic surgery at 25 years of age. To examine the decision between the two procedures in more detail, we varied the utility associated with proctocolectomy and subtotal colectomy simultaneously. Results of this analysis are shown in Figure 3.

View larger version (10K): [in this window] [in a new window]   Figure 3. Two-way sensitivity analysis varying the utilities associated with total and subtotal colectomy and their effects on quality-adjusted life expectancy. The lines represent thresholds at which the strategy leading to the greatest quality-adjusted life expectancy changes.

The results of the model were insensitive to changes in the risk for metachronous colorectal cancer and assumptions about colorectal cancer mortality after 5 years, surgical mortality, or endoscopy-associated mortality.

Discussion

Top Methods Results Discussion Author & Article Info References

Using decision analysis, we compared various management strategies for patients with mutations associated with hereditary nonpolyposis colorectal cancer and found that prophylactic colectomy at 25 years of age led to the greatest life expectancy, with a gain of 2.1 years with proctocolectomy and 1.8 years with subtotal colectomy compared with surveillance. The benefits of colectomy decreased if surgery was performed at an older age or at identification of an adenomatous polyp or colorectal cancer. For example, performing colectomy at the time of colorectal cancer diagnosis led to a survival benefit of less than 1 week compared with continued surveillance. In contrast to familial adenomatous polyposis, in which proctocolectomy is increasingly becoming the procedure of choice, our analysis demonstrated that minimal benefit is derived from performing proctocolectomy rather than subtotal colectomy in carriers of mutations for hereditary nonpolyposis colorectal cancer. When estimates of health-related quality of life were incorporated, surveillance became the preferred cancer prevention approach. In addition, subtotal colectomy led to greater quality-adjusted life expectancy than proctocolectomy for all colectomy strategies. The results were most sensitive to assumptions about the effectiveness of surveillance and quality of life associated with prophylactic surgery and undergoing surveillance.

In another published decision analysis in patients with hereditary nonpolyposis colorectal cancer [20], prophylactic colectomy at 40 years of age compared with surveillance showed a life expectancy benefit of 8 to 18 months. Using our model for a 40-year-old person, we found a life expectancy benefit of subtotal colectomy of 9.6 months compared with surveillance, which is similar to these previous results. However, the other analysis did not address the extent or timing of colectomy or quality of life associated with the different strategies and assumed a constant risk for colorectal cancer.

It is important to note that our results are based on a mathematical model and therefore cannot provide definitive answers. Decision analysis is a useful analytical approach that provides a framework for a problem and identifies key factors to target for future research. For example, our analysis considered overall effectiveness of surveillance and could not assess the optimal surveillance method or evaluate appropriate surveillance intervals, which are unclear in this population. Prospective, controlled studies of the efficacy of various preventive options should ultimately determine the optimal management strategy for mutation carriers.

To place our results in perspective, it is useful to compare the life expectancy gains of surveillance and prophylactic surgery (ranging from 13.5 to 15.6 years) in hereditary nonpolyposis colorectal cancer with those achieved with preventive interventions for colorectal cancer in other settings. The currently recommended screening strategy for colorectal cancer-annual fecal occult blood testing and flexible sigmoidoscopy every 5 years for persons at average risk for colorectal cancer-results in a life expectancy gain of approximately 3 weeks [19]. However, our results apply to persons who are known carriers of mutations for hereditary nonpolyposis colorectal cancer, not to the majority of patients with a family history of colon cancer whose families lack the pedigree characteristics of autosomal dominant inheritance of cancer predisposition or to families with hereditary nonpolyposis colorectal cancer in which no underlying mutation can be identified.

The effectiveness of surveillance and prophylactic surgery for other cancers associated with hereditary nonpolyposis colorectal cancer is not as well established as the effectiveness of these strategies for colorectal cancer; therefore, we did not consider these issues. Competing mortality from other tumors (such as endometrial, pancreatic, biliary, and ovarian tumors) must be considered before prophylactic colectomy can be recommended, because protection from colorectal cancer may not necessarily lead to prolonged life expectancy if the alternative is succumbing to another cancer. For example, patients with familial adenomatous polyposis who have undergone subtotal colectomy and ileorectal anastomosis still have a mortality rate that is threefold higher than that in the general population, primarily because of malignant conditions of the upper gastrointestinal tract and desmoid disease [49]. Applying this factor to our baseline analysis resulted in a decrease in life expectancy benefit of prophylactic surgery compared with surveillance (from 2.0 to 0.8 years and 1.7 to 0.6 years for total and subtotal prophylactic colectomy at 25 years of age, respectively).

Our study did not address the cost-effectiveness of various strategies. The analysis was intended to study individual patient decisions, not address societal allocation of health care resources. The cost-effectiveness of various preventive strategies and genetic testing will be an important factor to consider as more data on the prevalence of mutations and disease penetrance become available.

Our method of obtaining utilities relied on a small sample of health care providers and was intended only to provide reasonable utility estimates for the major health states under consideration. Many psychological factors (such as fear of developing cancer or a personal history of cancer) may play a role in determining personal preferences, but these are too complex to be fully addressed in our study. In addition, utilities may vary depending on the type of person being asked; for example, patients in a given health state generally associate that state with a higher utility than a health care provider, who in turn ranks it higher than someone in the general population. An illustration of the dependence of these decisions on personal values is demonstrated by applying the individual utilities of the health care providers interviewed in our survey to the model; doing this indicated that the greatest quality-adjusted life expectancy would be achieved by providing surveillance for five persons, proctocolectomy with ileoanal anastomosis at 25 years of age for four persons, and subtotal colectomy with ileorectal anastomosis at 25 years of age for one person.

Therefore, our study is useful in emphasizing the importance of incorporating personal preferences when making recommendations for individual patients and the need for more research on the effect of prophylactic surgery on quality of life. Available data on prophylactic colectomy primarily relate to patients who have undergone surgery for ulcerative colitis [50-52], who clearly form a different population because patients generally experience an improvement in quality of life after colectomy. The effect of postoperative morbidity associated with colectomy, which includes diarrhea and fecal incontinence [53, 54], on patients who were previously healthy needs to be studied further.

As genetic testing for cancer predisposition becomes more widely available, prophylactic surgery to prevent colorectal, breast, ovarian, and other cancers is likely to become more widely accepted as a management option. Our results show that is important to be cautious in advocating prophylactic surgical procedures in high-risk populations when effective alternatives for cancer prevention exist. Surveillance provides a reasonable alternative to prophylactic colectomy for carriers of mutations for hereditary nonpolyposis colorectal cancer. Prophylactic surgery is clearly the best option for some persons who are unwilling to undergo endoscopic surveillance. However, in most cases, the choice between prophylactic surgery and surveillance is a complex decision in which individual, personal preferences weigh heavily.

From Brigham and Women's Hospital; Dana-Farber Cancer Institute, Harvard Medical School; and Harvard School of Public Health, Boston, Massachusetts.

Dr. Kuntz: Center for Risk Analysis, Harvard School of Public Health, 718 Huntington Avenue, 2nd Floor, Boston, MA 02115-5924.