Abstract
Background:
Colorectal cancer is the second leading cause of cancer death in the United States. Approximately 3% of colorectal cancers are associated with Lynch Syndrome. Controversy exists regarding the optimal screening strategy for Lynch Syndrome.
Methods:
Using an individual level microsimulation of a population affected by Lynch syndrome over several years, effectiveness and cost-effectiveness of 21 screening strategies were compared. Modeling assumptions were based upon published literature, and sensitivity analyses were performed for key assumptions. In a two-step process, the number of Lynch syndrome diagnoses (Step 1) and life-years gained as a result of foreknowledge of Lynch syndrome in otherwise healthy carriers (Step 2) were measured.
Results:
The optimal strategy was sequential screening for probands starting with a predictive model, then immunohistochemistry for mismatch repair protein expression (IHC), followed by germline mutation testing (incremental cost-effectiveness ratio [ICER] of $35 143 per life-year gained). The strategies of IHC + BRAF, germline testing and universal germline testing of colon cancer probands had ICERs of $144 117 and $996 878, respectively.
Conclusions:
This analysis suggests that the initial step in screening for Lynch Syndrome should be the use of predictive models in probands. Universal tumor testing and general population screening strategies are not cost-effective. When family history is unavailable, alternate strategies are appropriate. Documentation of family history and screening for Lynch Syndrome using a predictive model may be considered a quality-of-care measure for patients with colorectal cancer.
Colorectal cancer (CRC) is the second leading cause of cancer death in the United States, with an incidence of over 142 820 new cases and 50 830 deaths per year (1). Up to six percent of these cancers are hereditary and potentially preventable. Lynch Syndrome (LS) is the most common hereditary colorectal cancer syndrome (2), accounting for approximately 3% of all colorectal cancers. Detection of Lynch Syndrome allows for personalization of medical care for the affected individual and provides an opportunity for preventive cancer care in family members. In the case of Lynch Syndrome, this is particularly important given that there is increased risk for a variety of cancers (3–7).
Recently, two distinct approaches to screening new CRC patients (probands) and screening the general population for LS have been recommended based on separate cost-effectiveness analyses (8–12). These approaches are based on the recognition that LS is caused by mutations in one of several DNA mismatch repair genes, leading to loss of expression of the specific protein product and the phenotype of microsatellite instability (MSI).
In 2010, Mvundura et al. found that it was cost-effective with an Incremental Cost Effectiveness Ratio (ICER) of less than $45 000 per life-years gained (LYG) to perform immunohistochemistry (IHC) studies for mismatch repair (MMR) protein expression in all newly diagnosed CRC cases followed by genotyping in patients with loss of MMR protein expression by IHC (8). All strategies in this study started with laboratory testing of the pathologic specimen.
Dinh et al. subsequently concluded that screening of the general population for LS was also cost-effective (10). Various strategies using PREMM1,2,6—a predictive model for assessment of risk for LS based on history—assigned risk levels to subjects in the general population. A risk of 5% and age cutoff of 25 to 35 years were found to be the most cost-effective strategies. However, PREMM1,2,6 was not intended for use in the general population and has not been validated in this population. When applied to the general population setting even using the sensitivity and specificity found by the validation studies in high-risk populations, PREMM1,2,6 will result in a low positive predictive value and a substantial number of false positives (13,14). Furthermore, the costs of using PREMM1,2,6 were not included in the model (14).
In 2011, Ladabaum et al. published a study of proband screening for LS and included strategies based on patient history before initiating laboratory tumor-based testing. The authors concluded that IHC + BRAF was the most cost effective strategy, when strategies based on history were excluded from analysis (12). Notably, the number of relatives for each proband was high (8), introducing a potential bias in favor of more expensive screening strategies.
A comparison of proband vs general population screening has not been done. In an effort to clarify the optimal approach to screening for LS, we conducted a comparison of all published algorithms and compared their effectiveness and cost-effectiveness.
Methods
Study Design and Setting
The modeling paradigm for comparative effectiveness analysis of LS was built around a cost-effectiveness endpoint. With respect to the assessment of cost, this study took a societal perspective and included two steps:
Step 1: The process by which healthy individuals with LS were identified. These were either first-degree relatives of probands, or members of the general population who were identified as a result of primary testing.
Step 2: The healthy individuals affected by LS who were followed over time with lifetime screening to detect colon cancer at earlier stages and prophylactic interventions for gynecologic cancers to reduce mortality and morbidity from LS.
Step 1 included 20 strategies: four strategies that started with clinical criteria (Amsterdam [15–18] or revised Bethesda [15,19]) and six strategies that started with a predictive model (PREMM1,2,6 [13], MMRpro [17], or MMRpredict [20,21]), all of which were followed by either IHC then germline testing or directly followed by germline testing. As with clinical criteria, predictive models used a subject’s medical history to calculate a predicted risk for LS. A cutoff value of 5% or higher was used to identify subjects in whom further testing is warranted (13,20,22,23). An additional five strategies started with tumor testing (MSI, IHC, BRAF testing, or combination) followed by germline testing.
Four strategies started with PREMM1,2,6 in the general population (GP), followed by germline testing. It is important to note that there is no rationale for the use of any of the predictive models or clinical criteria for screening the general population. These models were designed and validated in specific populations with personal or family history of colorectal cancer. Therefore, MMRpredict, MMRpro, Amsterdam Criteria, and revised Bethesda Criteria were excluded from the general population screening setting. However, given the published data by Dinh et al. we chose to include PREMM1,2,6-based general population screening strategies in the model to allow for a head-to-head comparison with proband screening strategies.
Lastly, universal germline testing for all newly diagnosed CRCs was also modeled. All twenty strategies were compared with a referent strategy of no additional screening to diagnose LS in the proband setting.
Cost-effectiveness
The benefits (effects) were measured in LYG accrued as a result of the foreknowledge of LS diagnosis and reduced overall mortality. Incremental cost-effectiveness ratios (ICERs, ie, the ratio of the increase in cost to the increase in the effect) using LYG as effect were the primary measure of cost-effectiveness. If a strategy is not cost-effective compared with the next more effective strategy, it is designated as dominated. Absolute dominance (AD) indicates that the next strategy is more effective and less costly, and extended dominance (ED) indicates that the next strategy is more effective and more costly, but has a lower ICER.
All of the strategies were modeled using TreeAge Pro 2012 by TreeAge Software, Inc. Costs structures used in previous studies were updated where appropriate based on the literature. Sensitivity and specificity for various tests were obtained from the validation studies, and stage distribution and disease incidence were obtained from SEER database (24) (see the Supplementary Materials, available online).
Study Models
The Step 1 model was a decision tree that represented all strategies using a cohort (n = 100 000) of CRC probands or healthy general population and calculated the number of LS diagnoses using each strategy (primary diagnoses). Using the assumptions about the first-degree relatives, this number was translated to a number for healthy individuals affected with LS (secondary diagnoses) who could be offered preventive measures described in Step 2. Similarly, the number of missed but detectable LS diagnoses by other strategies were identified and were used to divide the LS population in the Step 2 model into aware (detectable and diagnosed) and not aware (detectable but missed) cases of LS.
The Step 2 model was a Markov model that represented the natural history of CRC and other LS-related cancers in the population affected with LS. Colonoscopy every year between ages 20 and 80 years was used as the screening modality (25). The study population represented the same characteristics of the LS individuals in Step 1 in the much larger cohort of 100 000 subjects.
Effectiveness of the Screening Strategies, Simulation Scenarios
The simulated individuals in the Step 2 model had preferences in terms of compliance with screening recommendations given the status of the foreknowledge of LS (aware or not aware of carrying LS mutations). Since young otherwise healthy adults who carry LS mutations are the ones who benefit most from a screening, an age distribution for the simulated population conforming to a normal distribution with a median age of 30 years and standard deviation of three years was selected, with a male: female ratio of 1:1, similar to a previous study (12). Simulation terminated upon subjects’ death, age of 80 years, or 50 years of follow up, whichever came first.
We assumed that the diagnostic strategy used in Step 1 would influence the outcome of LS in the cohort. However, alternatives to this assumption are also discussed and examined in the Supplementary Materials (available online).
We used the sensitivity and specificity values for predictive models that were reported in the validation studies for each model. However, a study by Green et al. demonstrated that performance of predictive models such as MMRpro can be optimized. These optimized values are discussed and examined in the Supplementary Materials (available online).
Sensitivity Analysis
One-way sensitivity analyses using the range for cost of germline testing ($100-$880 per gene, varied in five equal steps), prevalence of LS in proband population (0.01–0.06 in six equal steps), and the number of first-degree relatives per primary diagnosis of LS (one to six in six equal steps) were performed and are reported here. One-way sensitivity analysis was also performed for the test parameters of the predictive models. In this analysis, the sensitivity values for the predictive models were decreased by a factor of 50% at increments of 10%. Further sensitivity analyses involving improved test parameters for predictive models are reported in the Supplementary Materials (available online).
Results
The results demonstrated an absolute risk reduction for CRC and CRC deaths by 2% to 11% and 1.5% to 8.5%, respectively. These numbers for gynecological cancers and cancer deaths were 1% to 5.6% and 0.7% to 3.7%, respectively. The numbers for other cancers did not change, as there was no standard screening or prophylactic treatment offered to prevent them. As expected, the most effective results were achieved by germline testing of all probands (Table 1). In contrast, the most cost-effective strategy was a three-step approach beginning with a predictive model: MMRpro, IHC, germline, with an ICER of $35 143 per LYG. The remaining undominated strategies were IHC + BRAF, germline with an ICER of $144 117, MMRpro, germline with an ICER of $223 988, and universal germline testing with an ICER of $996 878. All other strategies including all general population screening strategies were dominated (Figure 1). Using a threshold of $50 000 per life-year gained, only MMRpro, IHC, and germline testing met the cost effectiveness threshold (see Table 2).
Table 1.
Summary of the results*
| Strategy | Diagnoses of LS Primary (P) Secondary (S) |
Cost per diagnosis of LS Primary (P) Secondary (S) |
Incremental costs per primary diagnosis of LS (ICER) | Rank and status of the strategy | Probability of being aware | Sensitivity† | Colorectal cancer Cases Deaths |
Endometrial and ovarian cancer Cases Deaths |
Other cancer Cases Deaths |
Death because of Natural causes
Colonoscopy |
|---|---|---|---|---|---|---|---|---|---|---|
| No screening (referent strategy) | 0 | $0 | $0 | 1 | 0 | 0 | 33 357 | 12 815 | 7590 | 28 224 |
| 0 | $0 | 19 245 | 8464 | 6496 | 6 | |||||
| Amsterdam, IHC, germline | 493 | $11 618 | $ - | ED | 0.18 | 0.17 | 31 374 | 11 834 | 7683 | 28 955 |
| 641 | $11 202 | 17 698 | 7775 | 6565 | 42 | |||||
| Amsterdam, germline | 594 | $22 284 | $ - | AD | 0.22 | 0.2 | 30 932 | 11 621 | 7695 | 29 127 |
| 772 | $19 406 | 17 388 | 7643 | 6580 | 48 | |||||
| MMRpredict, IHC, germline | 1546 | $5365 | $ - | ED | 0.57 | 0.55 | 27 069 | 9683 | 7777 | 30 597 |
| 2010 | $6391 | 14 338 | 6440 | 6627 | 101 | |||||
| MMRpredict, germline | 1863 | $22 699 | $ - | AD | 0.69 | 0.67 | 25 810 | 9018 | 7834 | 31 060 |
| 2422 | $19 725 | 13 353 | 5997 | 6644 | 115 | |||||
| MMRpro, IHC, germline | 1994 | $4924 | $6053 | 2 | 0.74 | 0.72 | 25 226 | 8724 | 7874 | 31 318 |
| 2593 | $6053 | 12 940 | 5805 | 6672 | 131 | |||||
| MMRpro, germline | 2403 | $24 783 | $127 603 | 4 | 0.89 | 0.88 | 23 564 | 7802 | 7991 | 32 002 |
| 3124 | $21 328 | 11 706 | 5226 | 6770 | 158 | |||||
| PREMM, IHC, germline | 2017 | $7298 | $ - | ED | 0.75 | 0.73 | 25 157 | 8679 | 7872 | 31 359 |
| 2622 | $7878 | 12 882 | 5780 | 6673 | 131 | |||||
| PREMM, germline | 2430 | $47 309 | $ - | AD | 0.9 | 0.89 | 23 455 | 7751 | 7999 | 32 042 |
| 3159 | $38 656 | 11 605 | 5187 | 6779 | 161 | |||||
| Bethesda, IHC, germline | 1838 | $6495 | $ - | AD | 0.68 | 0.66 | 25 904 | 9059 | 7840 | 31 037 |
| 2389 | $7261 | 13 423 | 6029 | 6648 | 113 | |||||
| Bethesda, germline | 2214 | $37 702 | $ - | AD | 0.82 | 0.8 | 24 367 | 8197 | 7914 | 31 670 |
| 2878 | $31 266 | 12 272 | 5480 | 6707 | 142 | |||||
| IHC, germline | 2241 | $17 145 | $ - | AD | 0.83 | 0.82 | 24 247 | 8141 | 7934 | 31 749 |
| 2913 | $15 453 | 12 185 | 5442 | 6725 | 144 | |||||
| IHC + BRAF, germline | 2241 | $14 796 | $75 082 | 3 | 0.83 | 0.82 | 24 247 | 8141 | 7934 | 31 749 |
| 2913 | $13 646 | 12 185 | 5442 | 6725 | 144 | |||||
| MSI, germline | 2295 | $38 261 | $ - | AD | 0.85 | 0.84 | 24 056 | 8033 | 7971 | 31 803 |
| 2984 | $31 696 | 12 048 | 5380 | 6757 | 148 | |||||
| MSI + IHC, germline | 2430 | $36 045 | $ - | ED | 0.9 | 0.89 | 23 455 | 7751 | 7999 | 32 042 |
| 3159 | $29 991 | 11 605 | 5187 | 6779 | 161 | |||||
| MSI + IHC + BRAF, germline | 2430 | $36 181 | $ - | AD | 0.9 | 0.89 | 23 455 | 7751 | 7999 | 32 042 |
| 3159 | $30 096 | 11 605 | 5187 | 6779 | 161 | |||||
| Up-front germline | 2700 | $117 333 | $668 535 | 5 | 1 | 1 | 22 328 | 7166 | 8024 | 32 493 |
| 3510 | $92 521 | 10 760 | 4803 | 6817 | 177 | |||||
| PREMM GP screening 1, | 138 | $597 283 | $ - | AD | 1 | 0.89 | 22 328 | 7166 | 8024 | 32 493 |
| germline age > 20 y | 179 | $461 713 | 10 760 | 4803 | 6817 | 177 | ||||
| PREMM GP screening 2, | 129 | $597 283 | $ - | AD | 0.93 | 0.89 | 23 086 | 7554 | 8008 | 32 205 |
| germline age > 25 y | 168 | $461 713 | 11 328 | 5055 | 6788 | 164 | ||||
| PREMM GP screening 3, | 120 | $597 283 | $ - | AD | 0.87 | 0.89 | 23 841 | 7941 | 8000 | 31 891 |
| germline age > 30 y | 156 | $461 713 | 11 893 | 5309 | 6780 | 152 | ||||
| PREMM GP screening 4, | 110 | $597 283 | 0.8 | 0.89 | 24 594 | 8325 | 7912 | 31 562 | ||
| germline age > 35 y | 144 | $461 713 | $ - | AD | 12 432 | 5551 | 6707 | 137 |
* The first six columns summarize the performance results of each strategy in Step 1. The last four columns provide frequencies of cancer diagnoses and mortalities because of cancer and noncancer causes. ICER = incremental cost-effectiveness ratio; IHC = immunohistochemistry; LS = Lynch Syndrome; MSI = microsatellite instability.
† Specificity of all strategies other than the referent strategy is considered to be 1, as all strategies include germline testing as a final step.
Figure 1.
Cost-effectiveness of 21 screening strategies. The line connects the undominated strategies in the analysis. Undominated strategies progressively become more expensive, and the slope of the line between any two undominated strategies is the incremental cost-effectiveness ratio (ICER) for the more expensive strategy. The four outlier strategies in the top right corner represent the general population screening strategies. Numbers in parentheses indicate the rank of the strategy with respect to the ICERs. AD = absolute dominance; ED = extended dominance; IHC = immunohistochemistry; MSI = microsatellite instability.
Table 2.
Summary of the results of the microsimulation for Step 2*
| Strategy | Discounted life-years per LS diagnosis |
Discounted costs per LS diagnosis |
Discounted incremental cost per life-year gained (ICER) |
Rank and status of the strategy | Discounted incremental cost per life-year gained (ICER) |
Rank and status of the strategy |
|---|---|---|---|---|---|---|
| No screening (referent strategy) | 21.58 | $30 409 | $0 | 1 | $0 | 1 |
| Amsterdam, IHC, germline | 21.71 | $44 798 | $ - | ED | ||
| Amsterdam, germline | 21.74 | $53 668 | $ - | AD | ||
| MMRpredict, IHC, germline | 21.99 | $46 445 | $ - | ED | ||
| MMRpredict, germline | 22.07 | $61 742 | $ - | AD | ||
| MMRpro, IHC, germline | 22.1 | $48 909 | $35 143 | 2 | ||
| MMRpro, germline | 22.21 | $66 764 | $223 988 | 4 | ||
| PREMM, IHC, germline | 22.11 | $50 902 | $ - | ED | ||
| PREMM, germline | 22.21 | $84 241 | $ - | AD | ||
| Bethesda, IHC, germline | 22.06 | $49 128 | $ - | AD | ||
| Bethesda, germline | 22.16 | $75 553 | $ - | AD | ||
| IHC, germline | 22.17 | $59 913 | $ - | AD | $ - | AD |
| IHC + BRAF, germline | 22.17 | $58 106 | $144 117 | 3 | $46 925 | 2 |
| MSI, germline | 22.18 | $76 520 | $ - | AD | $ - | AD |
| MSI + IHC, germline | 22.21 | $75 576 | $ - | ED | $400 728 | 3 |
| MSI + IHC + BRAF, germline | 22.21 | $75 681 | $ - | AD | $ - | AD |
| Up-front germline | 22.28 | $139 811 | $996 878 | 5 | $940 024 | 4 |
| PREMM GP screening 1, germline age > 20 y | 22.28 | $308 259 | $ - | AD | $ - | AD |
| PREMM GP screening 2, germline age > 25 y | 22.23 | $307 101 | $ - | AD | $ - | AD |
| PREMM GP screening 3, germline age > 30 y | 22.19 | $306 064 | $ - | AD | $ - | AD |
| PREMM GP screening 4, germline age > 35 y | 22.15 | $304 920 | $ - | AD | $ - | AD |
* The numbers in columns “Rank and status” show the cost effectiveness rank of the undominated strategies. The last two columns show the results of analysis when strategies that use clinical criteria or predictive models were excluded. AD = absolute dominance; ED = extended dominance; ICER = incremental cost-effectiveness ratio; IHC = immunohistochemistry; LS = Lynch Syndrome; MSI = microsatellite instability.
If the predictive models and clinical criteria were removed from the ICER calculations, IHC + BRAF (germline with an ICER of $46 925), MSI, IHC (germline with an ICER of $400 728), and universal germline testing (ICER of $940 024) emerged as undominated. Of these, only IHC + BRAF and germline met the cost-effectiveness criteria at a threshold of $50 000.
The results of Step 1, using the number of diagnoses of LS for each strategy as the “effect” also produced a table of ICERs. The strategies identified as undominated and their rankings matched the results of the simulation in Step 2.
The yield of strategies in Step 1 in terms of the number of LS diagnoses is markedly lower for general population strategies. This would have created a bias against these strategies. In order to remove this bias, we used the general population strategy with the highest number of LS diagnoses (including all diagnoses, both primary and secondary) as the gold standard and used this number to determine the aware and not aware breakdown of the population for Step 2.
Clinical criteria and general population screening strategies did not emerge as cost-effective strategies for screening for LS (Table 2).
Sensitivity Analysis
Decreasing the prevalence of LS among the CRC patient population to 1% or increasing it to 6% did not affect the ranking of cost-effective strategies, although the ICER did change (Table 3). As expected, the results were most sensitive to the costs of germline testing. Decreasing the cost of germline testing to $685 per gene resulted in domination of IHC + BRAF, germline by MMRpro, germline with an ICER of $136 482. Decreasing the cost of germline testing to $295 per gene resulted in MMRpro, germline becoming marginally cost-effective with an ICER of $60 953 per LYG. Decreasing the cost of germline testing further improved the ICER for MMRpro, germline until it reached $33 195 when the cost of germline testing was set at $100 per gene, dominating MMRpro, IHC, germline testing. Interestingly, the ICER for universal germline testing decreased only to $116 355, still notably higher than using MMRpro, germline testing (Table 3).
Table 3.
The results of sensitivity analysis*
| Strategy | Germline testing costs Step 1 ICERsStep 2 ICERs | LS prevalence in CRCStep 1 ICERsStep 2 ICERs | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| $100 | $295 | $490 | $685 | $880 | 0.01 | 0.02 | 0.03 | 0.04 | 0.05 | 0.06 | |
| No screening | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 |
| $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | |
| Amsterdam, IHC, germline | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED |
| AD | ED | ED | ED | ED | AD | AD | ED | ED | ED | ED | |
| Amsterdam, germline | ED | ED | ED | AD | AD | AD | AD | AD | AD | AD | AD |
| ED | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| MMRpredict, IHC, germline | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED |
| ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | |
| MMRpredict, germline | ED | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| ED | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| MMRpro, IHC, germline | ED | $6053 | $6053 | $6053 | $6053 | $13 149 | $7827 | $6053 | $5166 | $4633 | $4278 |
| ED | $35 143 | $35 143 | $35 143 | $35 143 | $48 624 | $38 513 | $35 143 | $33 458 | $32 447 | $31 773 | |
| MMRpro, germline | $5850 | $27 621 | $50 384 | $74 703 | $127 603 | $329 540 | $178 087 | $127 603 | $102 360 | $87 215 | $77 118 |
| $33 195 | $60 953 | $98 717 | $136 482 | $223 988 | $538 757 | $302 680 | $223 988 | $184 642 | $161 034 | $145 295 | |
| PREMM, IHC, germline | AD | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED |
| ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | |
| PREMM, germline | ED | ED | ED | AD | AD | AD | AD | AD | AD | AD | AD |
| ED | ED | ED | AD | AD | AD | AD | AD | AD | AD | AD | |
| Bethesda, IHC, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | ED | ED | ED | |
| Bethesda, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| IHC, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| IHC + BRAF, germline | AD | AD | ED | $72 124 | $75 082 | $212 732 | $109 494 | $75 082 | $57 875 | $47 552 | $40 669 |
| AD | AD | ED | ED | $144 117 | $369 166 | $200 379 | $144 117 | $115 986 | $99 108 | $87 855 | |
| MSI, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| MSI + IHC, germline | AD | AD | AD | ED | ED | ED | ED | ED | ED | ED | ED |
| AD | AD | AD | ED | ED | ED | ED | ED | ED | ED | ED | |
| MSI + IHC + BRAF, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| Up-front germline | $66 499 | $217 008 | $367 517 | $518 026 | $668 535 | $2 037 507 | $1 010 778 | $668 535 | $497 414 | $394 741 | $326 292 |
| $116 355 | $336 486 | $556 616 | $776 747 | $996 878 | $3 001 459 | $1 498 023 | $996 878 | $746 305 | $595 961 | $495 732 | |
| PREMM GP screening 1, germline age > 20 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| PREMM GP screening 2, germline age > 25 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| PREMM GP screening 3, germline age > 30 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| PREMM GP screening 4, germline age > 35 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
* Prevalence of Lynch Syndrome with a range of 0.01 to 0.06. Cost of Germline Testing with a range of $100 to $880 per gene with five equally spaced values. AD = absolute dominance; CRC = colorectal cancer; ED = extended dominance; ICER = incremental cost-effectiveness ratio; IHC = immunohistochemistry; LS = Lynch Syndrome; MSI = microsatellite instability.
Decreasing the number of first-degree relatives to one while increasing the ICERs did not change the status or ranking of strategies. Increasing this number to six resulted in improved ICERs while maintaining the status and ranking of cost-effective strategies (Table 4). Reducing the sensitivity of the predictive models simultaneously to 0.6 and 0.5 of the original values resulted in MMRpro, IHC, germline being dominated by IHC + BRAF, germline with an ICER of $48 339 and $49 976, respectively. The results are summarized in Table 4.
Table 4.
The results of sensitivity analysis*
| Strategy | Number of family members per primary LS diagnosis Step 1 ICERs Step 2 ICERs |
Adjustment to the sensitivity of predictive models Step 1 ICERs Step 2 ICERs |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | 0.5 | 0.6 | 0.7 | 0.8 | 0.9 | 1.0 | |
| No screening | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 |
| $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | $0 | |
| Amsterdam, IHC, germline | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED |
| AD | AD | AD | ED | ED | ED | ED | ED | ED | ED | ED | ED | |
| Amsterdam, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| MMRpredict, IHC, germline | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED |
| ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | |
| MMRpredict, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| MMRpro, IHC, germline | $21 204 | $11 735 | $8578 | $7000 | $6053 | $5421 | $9552 | $8386 | $6928 | $6928 | $6441 | $6053 |
| $63 749 | $46 549 | $39 441 | $35 230 | $35 143 | $35 466 | ED | ED | $41 739 | $41 896 | $38 897 | $35 143 | |
| MMRpro, germline | $628 954 | $315 609 | $211 161 | $158 937 | $127 603 | $106 713 | AD | AD | AD | AD | AD | $127 603 |
| $908 321 | $512 973 | $303 489 | $281 703 | $223 988 | $181 282 | ED | AD | AD | AD | AD | $223 988 | |
| PREMM, IHC, germline | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED |
| ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | ED | |
| PREMM, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | ED |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| Bethesda, IHC, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| Bethesda, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| IHC, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| IHC + BRAF, germline | $366 365 | $184 300 | $123 625 | $93 286 | $75 082 | $62 945 | $16 928 | $19 673 | $30 255 | $30 255 | $42 644 | $75 082 |
| $688 261 | $327 204 | $213 951 | $204 997 | $144 117 | $114 427 | $49 976 | $48 339 | $51 365 | $61 906 | $82 590 | $144 117 | |
| MSI, germline | AD | AD | AD | AD | AD | AD | ED | ED | ED | ED | ED | AD |
| AD | AD | AD | AD | AD | AD | ED | AD | AD | AD | AD | AD | |
| MSI + IHC, germline | ED | ED | ED | ED | ED | ED | $223 804 | $223 804 | $223 804 | $223 804 | $223 804 | ED |
| ED | ED | ED | ED | ED | ED | $470 700 | $342 546 | $394 976 | $345 684 | $342 390 | ED | |
| MSI + IHC + BRAF, germline | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| Up-front germline | $3 333 617 | $1 667 941 | $1 112 715 | $835 103 | $668 535 | $557 490 | $655 287 | $655 287 | $655 287 | $655 287 | $655 287 | $668 535 |
| $4 263 540 | $2 321 786 | $1 625 661 | $1 112 200 | $996 878 | $913 274 | $959 041 | $984 980 | $914 742 | $954 918 | $877 994 | $996 878 | |
| PREMM GP screening 1, germline age > 20 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| PREMM GP screening 2, germline age > 25 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| PREMM GP screening 3, germline age > 30 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
| PREMM GP screening 4, germline age > 35 y | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD |
| AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | AD | |
* The number of relatives was varied from one to six per diagnosis of Lynch Syndrome. The sensitivities of all predictive models were subjected to an adjustment factor to determine when tumor testing could dominate the predictive models. AD = absolute dominance; ED = extended dominance; ICER = incremental cost-effectiveness ratio; IHC = immunohistochemistry; LS = Lynch Syndrome; MSI = microsatellite instability.
Discussion
Our results suggest that screening algorithms that begin with predictive models are the most cost-effective strategies for screening for LS. This evidence suggests that universal tumor testing for all colon cancer patients for LS is not cost-effective. In the event that a CRC patient was adopted or the patient does not have knowledge of these cancers in his/her family and clinical suspicion exists, then proceeding to tumor-based testing would be appropriate. Given that routine screening of all colon cancers with IHC has been recently endorsed (26,27), our analysis suggests that using a predictive model first has comparable sensitivity (effectiveness) and can spare considerable societal resources. General population screening starting with PREMM1,2,6 failed to result in sufficient numbers of LS diagnoses to make it a contender compared with proband screening. The prevalence of LS is much lower in the general population compared with colon cancer population, and the costs of identifying LS subjects are much higher compared with proband screening strategies. Given the similarities in performance characteristics of other predictive models, if they were to be evaluated in the general population setting, their performances would suffer from similar shortcomings as PREMM1,2,6 in this context.
Our results remained stable over a host of sensitivity analyses—some reported here and some listed in the Supplementary Materials (available online). While there is no predefined threshold for willingness to pay to aid the decision-making based on the ICERs in Step 1, it can be inferred from our results that the strategies that cost less per diagnosis of LS will perform better overall.
As demonstrated in this analysis, the sensitivity of MMRpro had to be decreased by 40% for MMRpro, IHC, germline to be dominated by tumor-based testing. Furthermore, even if the cost of universal germline testing decreased to $100 per gene, MMRpro remained more cost-effective than universal germline testing, which dominated all other tumor based testing.
Up to 6% of CRC is hereditary, with LS accounting for only 50% of these. This fact points out that even if universal tumor testing is adopted as the strategy of choice for LS, there are still 50% of hereditary cases that require family history and investigation for a substantial number of CRC patients. Thus, a careful family history should be the first step, not only in LS diagnosis but also for identification of other syndromes that might be otherwise missed by universal tumor testing for LS.
MMRPro is data intensive and requires information about the affected and unaffected family members, which may lead to incomplete input in routine practice (23). MMRPredict is modeled based on a cohort of colon cancer patients who underwent evaluation for the identification of MLH1, MSH2, and MSH6. Therefore, it is best suited for patients with colon cancer and has limitations when used in families with less common Lynch-related cancers (20). PREMM1,2,6 is easy to use and accounts for many Lynch-associated tumors beyond colorectal and endometrial cancers. It is modeled based on the data from a large population of probands with known mismatch repair mutation (13). As expected, performance of these models is directly associated with the quality of information fed into them. With high-quality information, a negative screening result (as defined by a risk for Lynch Syndrome of <5%) can eliminate the need for molecular and/or genetic testing for Lynch Syndrome.
With the exception of the Amsterdam Criteria, the revised Bethesda Criteria as well as all predictive models are designed to be very sensitive and use the same information about the personal and family history of the CRC proband. Thus, it is difficult to recommend a single model for initial screening and it is best to consider pros and cons of each model in selecting one.
The results of this study are limited by the fact that it models reality and uses today’s information to project events that happen decades into the future. A sensitivity analysis that considers all eventualities is not feasible and would still remain vulnerable to concurrent changes in the assumptions. Because of the lack of validation data for the use of predictive models in screening the general population for LS, all such models, with the exception of PREMM1,2,6, were excluded from analysis. Unforeseen advances in diagnosis or management of CRC or LS can also alter these findings. Our conclusions also assume that information about family history is available and accurate. In clinical situations where these assumptions are not met or where pedigree size is small, routine screening with IHC is appropriate.
These findings highlight the importance of obtaining a careful and accurate family history in the evaluation of all patients with colon cancer. This information remains an effective as well as cost-effective first step in the evaluation of hereditary risk and is necessary for application of predictive models to screening for Lynch Syndrome. We recommend that such assessment using predictive models should be considered as a quality-of-care measure.
Funding
This study was funded in part by an institutional grant from the Cleveland Clinic Foundation (RPC-2011-1002) and Cancer Center Support Grant National Cancer Institute P30 CA43703.
Supplementary Material
The authors do not have any financial disclosures to make. All authors have had access to the data and certify the integrity of the analysis.
References
- 1. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer J Clin. 2013;63(1):11–30. [DOI] [PubMed] [Google Scholar]
- 2. Cunningham D, Atkin W, Lenz HJ, et al. Colorectal cancer. Lancet. 2010;375(9719):1030–1047. [DOI] [PubMed] [Google Scholar]
- 3. Aarnio M, Mecklin JP, Aaltonen LA, Nystrom-Lahti M, Jarvinen HJ. Life-time risk of different cancers in hereditary non-polyposis colorectal cancer (HNPCC) syndrome. Int J Cancer. 1995;64(6):430–433. [DOI] [PubMed] [Google Scholar]
- 4. Stoffel E, Mukherjee B, Raymond VM, et al. Calculation of risk of colorectal and endometrial cancer among patients with Lynch syndrome. Gastroenterology. 2009;137(5):1621–1627. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Engel C, Loeffler M, Steinke V, et al. Risks of less common cancers in proven mutation carriers with lynch syndrome. J Clin Oncol. 2012;30(35):4409–4415. [DOI] [PubMed] [Google Scholar]
- 6. Win AK, Lindor NM, Winship I, et al. Risks of colorectal and other cancers after endometrial cancer for women with Lynch syndrome. J Natl Cancer Inst. 2013;105(4):274–279. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Watson P, Vasen HF, Mecklin JP, et al. The risk of extra-colonic, extra-endometrial cancer in the Lynch syndrome. Int J Cancer. 2008;123(2):444–449. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Mvundura M, Grosse SD, Hampel H, Palomaki GE. The cost-effectiveness of genetic testing strategies for Lynch syndrome among newly diagnosed patients with colorectal cancer. Genet Med. 2010;12(2):93–104. [DOI] [PubMed] [Google Scholar]
- 9. Hampel H. Point: justification for Lynch syndrome screening among all patients with newly diagnosed colorectal cancer. J Natl Compr Canc Netw. 2010;8(5):597–601. [DOI] [PubMed] [Google Scholar]
- 10. Dinh TA, Rosner BI, Atwood JC, et al. Health Benefits and Cost-Effectiveness of Primary Genetic Screening for Lynch Syndrome in the General Population. Cancer Prev Res (Phila). 2011;4(1):9–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Palomaki GE, McClain MR, Melillo S, Hampel HL, Thibodeau SN. EGAPP Supplementary Materials evidence review: DNA testing strategies aimed at reducing morbidity and mortality from Lynch syndrome. Genet Med. 2009;11(1):42–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Ladabaum U, Wang G, Terdiman J, et al. Strategies to identify the Lynch syndrome among patients with colorectal cancer: a cost-effectiveness analysis. Ann Intern Med. 2011;155(2):69–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Kastrinos F, Steyerberg EW, Mercado R, et al. The PREMM(1,2,6) Model Predicts Risk of MLH1, MSH2, and MSH6 Germline Mutations Based on Cancer History. Gastroenterology. 2010;140(1):73–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Sadeghi S, Barzi A, Kattan MW, Meropol NJ. Screening for Lynch syndrome in the general population-letter. Cancer Prev Res (Phila). 2011;4(3):471; author reply 472. [DOI] [PubMed] [Google Scholar]
- 15. Hampel H, Frankel WL, Martin E, et al. Screening for the Lynch syndrome (hereditary nonpolyposis colorectal cancer). N Engl J Med. 2005;352(18):1851–1860. [DOI] [PubMed] [Google Scholar]
- 16. Salovaara R, Loukola A, Kristo P, et al. Population-based molecular detection of hereditary nonpolyposis colorectal cancer. J Clin Oncol. 2000;18(11):2193–2200. [DOI] [PubMed] [Google Scholar]
- 17. Green RC, Parfrey PS, Woods MO, Younghusband HB. Prediction of Lynch syndrome in consecutive patients with colorectal cancer. J Natl Cancer Inst. 2009;101(5):331–340. [DOI] [PubMed] [Google Scholar]
- 18. Sjursen W, Haukanes BI, Grindedal EM, et al. Current clinical criteria for Lynch syndrome are not sensitive enough to identify MSH6 mutation carriers. J Med Genet. 2010;47(9):579–585. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Pinol V, Castells A, Andreu M, et al. Accuracy of revised Bethesda guidelines, microsatellite instability, and immunohistochemistry for the identification of patients with hereditary nonpolyposis colorectal cancer. JAMA. 2005;293(16):1986–1994. [DOI] [PubMed] [Google Scholar]
- 20. Barnetson RA, Tenesa A, Farrington SM, et al. Identification and survival of carriers of mutations in DNA mismatch-repair genes in colon cancer. N Engl J Med. 2006;354(26):2751–2763. [DOI] [PubMed] [Google Scholar]
- 21. Balmana J, Balaguer F, Castellvi-Bel S, et al. Comparison of predictive models, clinical criteria and molecular tumour screening for the identification of patients with Lynch syndrome in a population-based cohort of colorectal cancer patients. J Med Genet. 2008;45(9):557–563. [DOI] [PubMed] [Google Scholar]
- 22. Balaguer F, Balmana J, Castellvi-Bel S, et al. Validation and extension of the PREMM1,2 model in a population-based cohort of colorectal cancer patients. Gastroenterology. 2008;134(1):39–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Chen S, Wang W, Lee S, et al. Prediction of germline mutations and cancer risk in the Lynch syndrome. JAMA. 2006;296(12):1479–1487. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Fast Stats: An interactive tool for access to SEER cancer statistics. 2013; http://seer.cancer.gov/faststats Accessed May 23, 2013.
- 25. Winawer S, Fletcher R, Rex D, et al. Colorectal cancer screening and surveillance: clinical guidelines and rationale-Update based on new evidence. Gastroenterology. 2003;124(2):544–560. [DOI] [PubMed] [Google Scholar]
- 26. Beamer LC, Grant ML, Espenschied CR, et al. Reflex immunohistochemistry and microsatellite instability testing of colorectal tumors for Lynch syndrome among US cancer programs and follow-up of abnormal results. J Clin Oncol. 2012;30(10):1058–1063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. (NCCN) NCCN. Colorectal Cancer Screening. Practice Guidlines 2013 version 1. 2013.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.