Cost-Effectiveness of Imaging Strategies to Reduce Radiation-Induced Cancer Risk in Crohn's Disease

Computed tomography enterography (CTE) is an important tool for monitoring disease activity and treatment response in patients with Crohn's disease (CD) involving the small bowel.¹ Concern has been raised, however, about the associated risk of radiation-induced cancer from repeated CTE and in particular about exposure in young patients.^2,–4 The magnitude of these risks has been estimated by extrapolating from existing long-term studies of the effects of radiation exposure assuming a linear no-threshold model of risk from radiation exposure.^4,5 These models have been used to estimate the lifetime increased risk of cancer from one-time and repeated medical radiation exposure.^2,6,–10 While the linear no-threshold model is the predominant theory,^4,11,–13 there is debate in the medical physics literature as to whether application of these studies to patients receiving diagnostic radiation, in particular of its application to radiation doses of less than 10 mSv, is appropriate.^14,–17

There are several indications for diagnostic imaging in the 65% of patients with CD involving the small bowel: establish the extent and severity of disease; reassess a patient who remains symptomatic on therapy; identify complications of disease including penetrating and stricturing disease; assess response to therapy; and evaluation of acute abdominal pain for causes such as abscess or perforation.^3,18 An emerging concept is to image patients serially for the presence of bowel damage with the goal of administering treatment to prevent further disease progression.^19,20

Previous studies have shown that patients with CD have many imaging studies over their lifetime, such as small bowel follow-through and CT scans of the abdomen and pelvis.^21,–24 Patients diagnosed at an early age, those with upper gastrointestinal disease, penetrating disease, history of intravenous corticosteroid or infliximab use, and patients with multiple surgeries are at risk for higher doses of radiation accumulating over their lifetime.^22,23 The potential risk of CTE-induced cancer related to cumulative low-dose radiation is a consideration in the choice of imaging modality for the diagnosis and monitoring of CD activity. Therefore, it has been suggested that techniques that eliminate or reduce radiation exposure should be considered.^24,–26

Magnetic resonance enterography (MRE) provides detailed anatomic, functional, and real-time information without the need for ionizing radiation, making it well suited for the evaluation of small-bowel disorders.^27,–29 Several studies have noted that MRE and CTE have similar accuracy in imaging CD.^25,30,–34 For patients with CD, MRE can provide the information necessary to monitor patient disease progression and inform clinical decisions without the radiation risk; however, MRE is more costly than CTE.

Given limited resources available for healthcare, it is important to have clinical guidelines recommending practices that maximize health outcomes and value. To determine value, policy makers must compare choices based on long-term costs and benefits to patients. The purpose of this study was to evaluate whether MRE is cost-effective compared to CTE for routine disease monitoring in CD patients.

Materials and Methods

Model Overview

We developed a Markov model to simulate cohorts of 100,000 patients with CD from age 20 until death or age 100 using 1-month time steps (Fig. 1). To assess the impact of switching from CTE to MRE for routine radiology in patients with CD, we compared the following six strategies: performing all studies using CTE only; MRE until 30, 40, 50, or 60 years of age followed by use of CTE; and MRE only. We evaluated each of these screening strategies on five different patient subgroups defined by required frequency of monitoring: twice annually, once annually, once every other year, once every 5 years, and once every 10 years. We assumed routine imaging occurred until age 80. We considered the impact of CT radiation on the incidence of colorectal, kidney, liver, lung, stomach, and urinary bladder cancer, and leukemia based on the organ-specific effective dose of an abdominal/pelvic CT scan.⁷ We assumed that all of the clinical information necessary to appropriately monitor patient health and to inform clinical decisions can be gathered equally well by CTE and MRE.^25,30,–34 Full model details are available in the Supplemental Methods.

The simulated cohorts consist of patients with CD with small bowel involvement, no history of proctocolectomy, no history of cancer, and no contraindications to MRE. We did not include patients with isolated colonic CD as part of the simulated cohorts. In each month all individuals may survive or die of noncancer causes. Individuals with cancer may die from that cancer and individuals without a previous history of cancer may develop a cancer. We assumed that individuals can only develop one cancer in their lifetime. Individuals with a history of cancer remain in that cancer state until death from cancer or other causes. During each month, patients with ileocolonic CD may undergo total proctocolectomy at a rate based on the general CD population, which eliminates the risk of colorectal cancer for these patients.

We followed the recommendations of the U.S. Panel on Cost Effectiveness in the development of the model and in the analysis of results, including taking a societal perspective, considering costs and benefits over a lifetime horizon, and discounting at 3% annually.³⁵ Costs were converted to 2009 U.S. dollars using the U.S. gross domestic product deflator. Base case parameter values, ranges, and distributions used in sensitivity analysis are presented in Supplemental Table 1. Costs of imaging were based on the 2009 Medicare reimbursement schedule. Cancer incidence, mortality, direct costs, patient time costs, and quality-of-life decrement were primarily based on Surveillance, Epidemiology, and End Results (SEER) and SEER-Medicare analysis.^36,–38 Cancer incidence rates were also adjusted using hazard ratio's specific to CD patients.^39,40 We implemented the model in TreeAge Pro 2009 Suite (TreeAge Software, Williamstown, MA) and Microsoft Excel 2007 (Redmond, WA).

Radiation-induced Cancer Risk

We assumed that radiation exposure from each dose-equivalent, measured in millisievert (mSv) of exposure, results in an additional annual rate of cancer for the lifetime of the patient following the linear no-threshold model.⁴ Thus, the annual cancer rate increases additively with the number of CTE scans. Reports of the effective exposure from CTE vary widely in the literature from 12 mSv to 25 mSv; in the base case, we assumed that one CTE exam results in 16 mSv of radiation exposure.^24,25 We calculated the incremental annual rate of each cancer per mSv based on the organ-specific radiation exposure from an abdominal and pelvic CTE by calibrating the model to the estimated increase in absolute lifetime cancer risk in a population of healthy 50-year-olds from a paired CT colonography (Fig. 2).⁷ Details of the calibration procedure are presented in the Supplemental Methods.

Results

Radiation-induced Cancers Averted

For patients requiring a high frequency of monitoring (two scans per year), the model estimates that using CTE for every scan will result in an average of 20,146 radiation-induced cancers per 100,000 patients over their lifetime (Table 1). However, in patient populations requiring only one scan every 5 or 10 years, using CTE for patient monitoring resulted in 2,388 or 1,206 radiation-induced cancers, respectively. By using MRE for routine patient monitoring until the age of 40 years, the expected number of radiation-induced cancers can be reduced by half for any of the patient populations considered. Approximately 75% of expected radiation-induced cancers in CD patients can be averted by using MRE instead of CTE for all patient monitoring before age 50. The total number of cancers and the estimated number of radiation-induced cancers are presented by cancer site in Supplemental Tables 2 and 3.

Cost-effectiveness

The incremental cost-effectiveness ratio (ICER), expressed as incremental cost per life-year (LY) or incremental cost per quality-adjusted life-year (QALY) gained, varied little across the frequency of monitoring (Table 1): compared to using CTE only, using MRE until age 30 had ICERs between $37,538 and $41,031 per LY gained and between $52,969 and $57,772 per QALY-gained across the range of imaging frequencies. Using MRE until age 50 resulted in ICERs between $58,022 and $62,648 per LY gained and between $84,250 and $90,982 per QALY gained. Using MRE until age 60 resulted in ICERs between $84,599 and $94,754 per LY gained and $124,217 and $139,148 per QALY gained. Using MRE to monitor CD patients over the age of 60 resulted in ICERs greater than $100,000 per LY gained and $100,000 per QALY gained for all frequency of scanning.

Sensitivity Analysis

Extensive one-way, multi-way and probabilistic sensitivity analyses indicated the model results were robust (Fig. 3; Supplemental Results). Key model variables were the additional cancer risk per mSv of radiation exposure, the radiation dose per scan, the difference in cost between CTE and MRE, and whether the patient had a history of total protocolectomy (full results for patients with a history of total proctocolectomy in Supplementary Tables 5–7).

Through sensitivity analysis, we learned that even at an attributable cancer risk per mSv of 0.36 per 100,000 person-years (half of the base case), MRE continued to have an ICER of less than $100,000 per QALY gained in patients younger than 30 years of age. At an attributable cancer rate per mSv of greater than 1.0 per 100,000 person-years, the ICER for using MRE rather than CTE for all patients younger than 60 is less than $100,000 per QALY gained. Assuming the base case rate of cancer per mSv, we found that at radiation doses of greater than 20 mSv per CTE, imaging all patients younger than 60 years with MRE rather than CTE costs less than $100,000 per QALY gained. At doses of less than 6.0 mSv per CTE the use of MRE had an ICER of greater than $100,000 per QALY gained, indicating that if CTE could be provided at this low dose then CTE would likely be preferred to MRE for all types of patients.

The difference in cost between CTE and MRE is also important in identifying the optimal strategy (Fig. 3D). When the cost difference between CTE and MRE is less than $600, MRE for patients up to 60 years has an ICER less than $100,000 per QALY gained. However, when the cost difference is $1500, MRE for patients greater than 40 years has an ICER greater than $100,000 per QALY gained.

We performed a probabilistic sensitivity analysis using the parameter distributions presented in Supplementary Table 1. At a maximum willingness-to-pay threshold of $50,000 per QALY, we found that the probability that using CTE for patients of all ages is the optimal choice is 60% and the probability that using MRE for patients younger than 30 years of age is the optimal choice is 40%. At a maximum willingness-to-pay threshold of $100,000 per QALY, the probability that using CTE for all patients is the optimal choice is 0.1% and the probability that using MRE for all patients under the age of 50 is the optimal choice is 87% (Fig. 4).

Discussion

Imaging is an essential element in the clinical management of patients with CD in terms of assessing stricturing and penetrating disease. Serial imaging has a growing noninvasive role for assessing ongoing inflammation and response to therapy. However, the potential for an increased lifetime risk of cancer as a consequence of imaging that employs ionizing radiation is a concern of patients and clinicians. Using a mathematical model empirically calibrated to previously published estimates for the lifetime risk of radiation-induced cancer from abdominal/pelvic CT,⁷ we conducted a cost-effectiveness analysis comparing the use of MRE to that of CTE for nonemergent disease assessment in patients with ileal and ileocolonic CD. We used the linear no-threshold model to estimate the number of radiation-induced cancers for various frequencies of scanning, age of transition from MRE to CTE, radiation exposure per CTE, and risk of radiation-induced cancer per mSV radiation dose.

In one study of patients with CD, 7% of the population was exposed to greater than 50 mSv of diagnostic radiation over 5 years.²¹ We estimate that this level of exposure continued over the life of a patient would result in ≈6,900 additional cancers per 100,000 patients. The number of radiation-induced cancers can be substantially reduced by using MRE.

While there is no formal threshold for defining what is and what is not good value for money in the U.S. healthcare system, there is general acceptance that technologies and programs costing less than $50,000 per QALY gained are cost-effective.⁴¹ Higher thresholds, such as $100,000 per QALY gained have also been proposed in recent years.⁴² Using a willingness-to-pay threshold of $100,000 per QALY-gained, we found that for patients with no history of total proctocolectomy using MRE for patients younger than 50 years of age is likely the optimal strategy. For patients with a prior history of total proctocolectomy, we found that using MRE for patients younger than 30 years of age is the optimal strategy.

We also found that when the cost difference between CTE and MRE is small (less than $600), or when the expected dose per CTE exceeds 20 mSv, that using MRE instead of CTE may be cost-effective for patients up to 60 years of age. Conversely, we found that if the expected rate of radiation-induced cancers per mSv of radiation exposure is less than half of what is predicted by the linear no-threshold model, when the expected dose per CTE was less than 6.0 mSv, or when patients strongly prefer CTE to MRE, that CTE is the cost-effective imaging tool for patients at any age.

Although there are no prospective data specifically linking CT-radiation exposure to cancer mortality,^15,–17 prospective randomized trials to determine the effect of additional CT scans over a lifetime are not feasible. Given the uncertainty around the risk of radiation-induced cancers, policies that attempt to minimize radiation exposure through either MRE or CTE with lower doses of radiation would minimize the impact of this uncertainty on patient safety.

The amount of radiation exposure varies across CT technologies, machines, and with the practices of the facility.^7,24,25 The dose of radiation exposure from CTE can be reduced by reducing the quality of the image, energy filters, and new image processing algorithms that can reduce overall exposure or exposure specifically to radiosensitive organs.^43,–46 Each dose-reducing strategy comes with potential drawbacks. However, recent studies have indicated that substantial reductions in the CTE radiation dose are possible by reducing image quality and improving the processing algorithm while maintaining diagnostic accuracy.^{25,33,34,46,47} At one center the initial attempt to reduce radiation exposure with CTE resulted in a reduction from 16 mSv per exam to 12 mSv per exam.²⁵ We found that if each CTE exam results in less than 6.0 mSv of exposure, then CTE is the preferred imaging tool even for patients less than 30 years old. Our results show that “low-dose” CTE is a promising future alternative method of reducing radiation-induced cancers at a potentially lower cost than MRE.

Our analysis has several limitations. We used U.S. healthcare system costs and clinical data from U.S. patient cohorts when available, so our findings may not be generalizable to other healthcare systems. The linear no-threshold model may not account for the ability of DNA repair mechanisms to respond to damage from low or very-low levels of radiation.^14,16 Other theories of low dose radiation effects exist: the threshold model that assumes a negligible effect at low doses; the radiation hormesis model that stipulates the activation of DNA repair mechanisms by very-low dose radiation has a net-positive impact; and other nonlinear effect models that postulate harm from radiation exposure is higher when the time between exposures is short. In each of these other models, it is important to explicitly define the threshold of exposure, whether it be at one time or within a specified critical time interval, at which cancer risk occurs. It is a limitation of our current study that we only consider the linear no-threshold model. We also did not consider patient imaging frequencies of greater than two times per year and we assumed that scans were evenly spaced.

While we did incorporate baseline cancer rates in CD patients higher than those observed in the general population,^39,40 we did not include the risk of lymphoma in the model, which may be heightened in patients receiving immunosuppressant medications.^48,49 We assumed that individuals can only develop one cancer in their lifetime, but in a population at high risk for a second primary cancer because of radiation exposure this may underestimate the costs of radiation exposure and the benefits of avoiding them.

MRE may have disadvantages other than higher cost that have not been considered in this study. MRE may be more operator-dependent and thus additional training may be required due to a lack of widespread subspecialized clinical expertise. Greater interobserver variability may result in more MRE studies being ordered than would be required if CTE were used. We focused our analysis on nonemergent imaging for CD. MRE may not be readily available in the assessment of potential perforation or complete bowel obstruction to help direct therapy in the emergency room. We did not model injury due to high-speed impact of accidental ferromagnetic external objects or burns from radiofrequency coils or physiologic monitors in the MRI system room, due to the sparse data and accidental nature. A limitation to our study is that MR safety accidents may affect the cost-effectiveness of MRE; however, generally there is already a policy imperative to prevent those accidents. In addition, we did not examine abdominal sonography as an alternative method to monitor disease without ionizing radiation due its limited evaluation and localization of bowel inflammation proximal to the terminal ileum, and limited identification of penetrating complications such as abscess, fistula, and stenoses.⁵⁰ However, local availability and expertise may lead to better accuracy of one imaging modality test than another.

Patients and their clinicians, who may be weighing the potential risks of radiation exposure when deciding whether to order a CTE, may be inclined to using imaging more frequently with MRE. The additional costs and benefits of these scans have not been considered in this study. We assume that all the scans provided to CD patients are necessary for their clinical management. Eliminating unnecessary scans, regardless of which technology used, is cost-saving from a patient and societal perspective.

We assume that the clinical information and accuracy provided by MRE would result in the same clinical decisions as CTE, and therefore would not result in changes in the clinical care or outcomes. We base this assumption on previous studies and a meta-analysis that showed similar accuracy of MRE and CTE in imaging CD.^25,30,–34 Our analysis aimed to specifically examine the cost-effectiveness of MRE as an alternative to CTE in terms of potential radiation risks and costs while holding constant the accuracy between the two systems.

In conclusion, MRE is a cost-effective alternative to CTE for younger patients with CD who are likely to have multiple imaging tests over the course of their disease. Ongoing research efforts directed toward reducing radiation exposure increase the likelihood that “low-dose” CTE will be a cost-effective alternative to traditional imaging as well.

Acknowledgements

Specific author contributions: Study concept and design, developed the model, analysis and interpretation of results, drafting the article: Lauren E. Cipriano; study concept and design, analysis and interpretation of results, editing and revising the article: Barrett G. Levesque; study concept and design, analysis and interpretation of results, provided critical review, editing, and revising the article: Gregory S. Zaric; analysis and interpretation of results, provided critical review, editing and revising of the article: Edward V. Loftus, Jr.; provided critical review, editing and revising of the article: William J. Sandborn.

References

Author notes