Cost-effectiveness of Follow-up CT for Incidental Ascending Aortic Dilatation

Abstract

Purpose

To evaluate the cost-effectiveness of CT follow-up strategies for incidental aortic dilatation.

Materials and Methods

In this cost-effectiveness analysis, a simulation model was developed with 1 000 000 adult patients aged 55–75 years with incidentally detected dilated aortas measuring 40–50 mm. Follow-up CT strategies were evaluated for various patient age– and aortic size–based cutoffs. Follow-up frequency ranged from 1 to 3 years, as well as a single follow-up CT examination at 1 year. Patient survival was determined by risk of aortic dissection or rupture and surgical- and age-based mortality. Costs and quality-adjusted life-years (QALYs) were calculated for each strategy within the simulated cohort. A probabilistic sensitivity analysis was performed by varying model parameters.

Results

The cost-effective strategy with the highest QALYs under a willingness-to-pay threshold of $100 000 per QALY was follow-up CT for patients younger than 60 years with aortas measuring at least 40 mm in diameter every 3 years (incremental cost-effectiveness ratio, $62 511; 95% CI: $52 168, $77 739). With this strategy, follow-up imaging was needed for only 17% of dilated aortas in the cohort. Probabilistic sensitivity analysis demonstrated that the cost-effective strategies at $100 000 per QALY threshold included the following: no follow-up for patients with aortas smaller than 50 mm (39% of simulations), follow-up every 3 years for patients younger than 55 years with aortas measuring at least 45 mm (21%), and follow-up every 3 years for patients older than 65 years with aortas measuring at least 40 mm (14%).

Conclusion

Follow-up CT for an incidentally detected dilated ascending aorta smaller than 50 mm is likely not cost-effective in patients older than 60–65 years.

Keywords: CT, Thorax, Vascular, Aorta, Cost-Effectiveness, Cost-Benefit Analysis

Supplemental material is available for this article.

© RSNA, 2023

See also commentary by Shen and Fleischmann in this issue.

Summary

In a simulated patient cohort with incidentally detected dilated aortas measuring 40–50 mm in diameter, follow-up CT in patients older than 60 years was not cost-effective.

Key Points

• ■ The most cost-effective strategy with the highest quality-adjusted life-years (QALYs) under a willingness-to-pay threshold of $100 000 per QALY was follow-up CT for patients younger than 60 years of age with aortas measuring at least 40 mm in diameter every 3 years (incremental cost-effectiveness ratio, $62 511 per QALY).

• ■ Follow-up for patients older than 60 years with aortas smaller than 50 mm was not cost-effective at a willingness-to-pay threshold of $100 000 per QALY.

• ■ With an age cutoff of 60 years, follow-up CT imaging was needed for only 17% of dilated aortas in the simulated cohort.

Introduction

Dilatation of the ascending aorta (defined as a diameter ≥40 mm) is a common incidental finding in patients undergoing chest or cardiac CT, with prevalence ranging from 2.7% to 23% of patients (1,2). Radiologists and cardiologists disagree regarding the threshold that defines an “aneurysm” of the ascending aorta, with some using 40 mm and others, such as the American College of Radiology, using 50 mm (3). Per guidelines from the American Heart Association, aortas larger than 55 mm, those growing at a rate of more than 5 mm per year, and those larger than 45 mm in patients undergoing aortic valve surgery are generally considered for surgical repair (4). (Patients with genetic or bicuspid valve aortopathy generally have a lower threshold for surgery as well.) Patients with aneurysmal ascending aortas above these thresholds are at higher risk of developing complications, most notably aortic dissection with potential for rupture, which is associated with high mortality (5–7).

The risk of acute complications in patients with aortas between 40 mm and 55 mm is lower, and patients are generally not referred to surgical repair given potential morbidity from the procedure. However, dilated ascending aortas tend to grow slowly over time (5), and patients may eventually reach the surgical thresholds later. Many patients with dilated aortas undergo follow-up imaging, often with CT, to monitor aortic growth. Unfortunately, neither the American College of Radiology nor the American Heart Association guidelines specifically address whether and how often such patients should be followed up.

Simulation modeling is a method by which physicians can recreate the natural history of diseases in silico, using data from the literature to inform the model and evaluate potential management strategies (8). This method is particularly important in settings where randomized controlled trials are difficult to perform because of ethical (eg, nontreatment of a potentially fatal condition) or feasibility (eg, a condition with a low event rate requiring a very large trial) concerns. By testing several management strategies in a simulation, we can estimate the costs and benefits of each. Cost-effectiveness refers to the ratio of benefit of a management strategy in terms of quality-adjusted life-years (QALYs) versus its cost to the patient and health care system. Strategies can then be compared by calculation of an incremental cost-effectiveness ratio (ICER) that represents the cost in dollars per each QALY gained. In this study, we created a simulation model of patients with incidentally detected dilated ascending thoracic aortas between 40 and 50 mm in diameter and evaluated the cost-effectiveness of multiple CT follow-up strategies using age- and size-based cutoffs.

Materials and Methods

This study was a cost-effectiveness analysis using data from a simulated patient cohort and was therefore exempt from institutional review board review.

Model Structure

The simulation model was developed in Java (version 11.0) for computational efficiency. Output was analyzed with several custom Perl scripts (version 5.30) to generate the efficient frontier by implementing strong and weak dominance. Data were summarized and graphed in Excel (version 16; Microsoft). The simulation consisted of a microsimulation with a yearly cycle, with 1 000 000 adult patients per run and a lifetime horizon. Each patient was assumed to have an incidentally detected dilated ascending aorta measuring 40–50 mm at the time of the baseline examination. Patients were assumed to not have any aortopathy or known valvular disease.

The model structure, including aortic events and follow-up imaging, is outlined in Figure 1. The model parameters for the base-case analysis are given in Table 1. Costs were assumed for a U.S. federal payor.

Figure 1:

Simulation model structure.

Table 1:

Parameters for Simulation Model of 1 Million Patients with Incidentally Detected Ascending Aortic Dilatation

At baseline, a patient is first assigned an aortic size within a range of 40–50 mm; an exponential distribution and a mean of 43 mm were chosen based on distributions in the literature (9,10). A histogram of the baseline aortic size distribution is given in Figure S1. Patient age is then assigned using distributions that depend on aortic size, with larger aortas tending to occur in older patients (10). Age range was 18–90 years, and sex was randomly assigned, resulting in a final sample with 70% men (9,11). Aortic growth rate was also randomly assigned from an exponential distribution with a mean of 0.2 mm per year in the base-case scenario (9–11). A histogram of aortic growth rate is given in Figure S2. Growth rate was assumed to be constant during follow-up based on findings from Gagné-Loranger et al (12).

In the first cycle of the simulation, CT depicts the dilated ascending aorta. After 1 year, the following events may occur: the patient may experience nonaortic mortality (risk depending on age and sex); the patient may experience aortic dissection, rupture, or both (risk depending on aortic size); or the patient may experience neither of these outcomes. If an aortic dissection or rupture occurs, the patient may experience mortality at a rate of 56.6% (6,7); otherwise, patient life expectancy depends on age and sex. This high mortality rate accounts for out-of-hospital and in-hospital mortality, not merely surgical mortality. There is also a small risk of stroke from emergency surgery, leading to reduced quality of life (Table 1).

If the patient has not died or had emergency aortic surgery, the simulation then accounts for aneurysm annual growth. If the patient meets criteria for follow-up CT based on the previous aortic size, age, and time interval (described later), CT is then performed. If the aortic size is 55 mm or greater or if the aorta has grown at least 5 mm, then the patient will be sent for elective surgery. Surgical mortality and morbidity from stroke depend on patient age (Table 1). If the patient remains alive, life expectancy will depend on age and sex. If the patient does not meet criteria for surgery, the yearly cycle repeats, as shown in Figure 1.

Follow-up strategies were formulated as follow-up CT if the following conditions are met: patient age less than or equal to age cutoff and aortic size greater than or equal to size cutoff every N years, where N was either 1, 2, or 3. Age cutoffs tested were 55, 60, 65, 70, and 75 years; aortic size cutoffs were 40, 45, and 50 mm. For example, one tested strategy would be the following: perform follow-up CT if the patient was aged 60 or more years and aortic size was 45 mm or greater every 3 years. We also tested a strategy in which a single follow-up CT examination was performed 1 year after diagnosis. In total, five age ranges × three size cutoffs × four frequency options (once and every 1, 2, or 3 years) yields 60 strategies. Patients found to have an aorta measuring 50 mm or greater at baseline or follow-up were followed annually regardless of the management strategy or age.

The model was validated by measuring cumulative rates of aortic dissection or rupture at 5 years in untreated patients.

Model Runs

A set of 1 million simulated patients was generated as described. For each management strategy tested (described earlier), the simulation was run, and total QALYs and costs were tabulated. Costs and QALYs were discounted by the commonly used value of 3% per year. Management strategies were compared by calculating the ICER, and those along the efficient frontier were identified to determine the most cost-effective strategies.

Statistical Analysis

The CIs for the ICERs were generated by bootstrapping, with 500 runs on the base-case analysis. A $100 000 per QALY willingness-to-pay threshold was used, which is standard for U.S. health care (13).

One-Way Sensitivity Analysis

In addition to using the base-case model parameter values from Table 1, the simulation was also run after adjusting several parameters: mean aortic growth rate (range, 0.1–0.4 mm per year), aortic dissection or rupture rate (range, 50%–200% of base value), mortality rate for elective surgery (range, 50%–200% of base value), and mortality rate for emergent surgery (range, 25%–60% absolute value). The efficient frontier was calculated, and at each parameter level, we identified the optimal strategy at a willingness-to-pay threshold of $100 000 per QALY.

Probabilistic Sensitivity Analysis

Five hundred separate runs of each strategy were performed while varying a subset of parameters, as specified in Table 1. For each run, the value of each parameter was sampled from the probabilistic sensitivity analysis distribution specified in Table 1. The efficient frontier was then determined for each set of parameter values. The results of the runs were tabulated, and the fraction of runs in which a strategy was optimal at willingness-to-pay thresholds of $10 000–$100 000 was calculated.

Results

Model Validation

For the 1 000 000 simulated patients, the mean age was 68 years ± 9 (SD), and 700 189 (70%) patients were men. The mean and median aortic sizes were 43 and 42 mm, respectively, with an IQR of 41–44 mm. Mean and median aortic growth rates were 0.2 and 0.1 mm per year, respectively, with an IQR of 0.06–0.3 mm per year. At 5 years, the rate of aortic dissection or rupture was 3882 of 797 692 (0.5%) for patients with initial aortic size of 40–44 mm, 3074 of 183 164 (1.7%) for patients with initial aortic size of 45–49 mm, and 357 of 12 188 (2.9%) for patients with initial aortic size of 50 mm. These event rates are in line with the parameter targets in Table 1.

Efficient Frontier

A total of 60 strategies were tested. Nine were located along the efficient frontier, as shown in Table 2 and Figure 2. The strategy associated with the highest QALYs per patient was follow-up for all patients younger than 75 years and all dilated aortas with diameter of 40 mm or greater every year; however, this was expensive, with an ICER greater than $40 million. A follow-up strategy for patients younger than age 65 years and all dilated aortas 40 mm or greater every 3 years yielded an ICER of $106 686 (95% CI: $92 868, $124 725), just above the common $100 000 per QALY threshold. The strategy with greatest QALYs just below this threshold was follow-up for patients younger than 60 years with dilated aortas 40 mm or greater every 3 years, with an ICER of $62 511 (95% CI: $52 168, $77 739). Additional strategies below the $100 000 per QALY threshold, in order of decreasing QALYs, include the following: patients younger than 55 years with dilated aortas 40 mm or greater every 3 years, patients younger than 60 years with dilated aortas 45 mm or greater one time only, patients younger than 55 years with dilated aortas 45 mm or greater one time only, and only patients with dilated aortas 50 mm or greater annually.

Table 2:

Efficient Frontier of CT Follow-up Strategies Based on Patient Age and Aortic Size

Figure 2:

Graph shows cost and quality-adjusted life-years (QALYs) of tested strategies, with efficient frontier in black. Strategies on the efficient frontier are labeled as follows: age cutoff (in years)/aortic size cutoff (in millimeters)/frequency (in years). The 55/50/once strategy is equivalent to following patients with aortas 50 mm or greater annually.

Of note, for the follow-up strategy of all patients younger than 75 years with aortic size of 40 mm or greater, 74% (741 279 of 1 000 000) of the simulation cohort required follow-up CT. For the strategy of patients younger than 60 years with aortic size of 40 mm or greater, this number decreased to 17% of the cohort. Follow-up of only patients with dilated aortas 50 mm or greater resulted in only 1% of the cohort requiring follow-up CT.

One-Way Sensitivity Analysis

Figure 3 shows the results of multiple one-way sensitivity analyses, giving the optimal follow-up strategy at a willingness-to-pay threshold of $100 000 per QALY. As noted earlier, the optimal strategy in the base-case analysis was follow-up for patients younger than 60 years with dilated aortas 40 mm or greater every 3 years. If mean aortic growth rate was lowered, follow-up was less cost-effective. At a growth rate of 0.1 mm per year, the cost-effective strategy was follow-up for only patients with dilated aortas 50 mm or greater. If the risk of aortic dissection or rupture was increased, the cost-effective strategy was changed from an age cutoff of younger than 60 years in the base case to younger than 65 years. If mortality from elective surgery was increased, the cost-effective strategy changed from an age cutoff of younger than 60 years in the base case to younger than 55 years. The same age cutoff change was observed when mortality from aortic dissection or rupture was decreased.

Figure 3:

One-way sensitivity analysis. Graphs show the optimal strategy at a willingness-to-pay threshold of $100 000 per quality-adjusted life-year as several parameters were varied: mean aortic growth rate, aortic dissection rate (as a percentage of base-case value), mortality of elective surgery (as a percentage of base-case value), and mortality of acute aortic dissection or rupture.

Probabilistic Sensitivity Analysis

In addition to varying each parameter individually, we also performed probabilistic sensitivity analysis, which varies multiple parameters simultaneously (Table 1). The results of the probabilistic sensitivity analysis are shown in Figure 4, which displays the fraction of simulation in which a given strategy is optimal at a specific willingness-to-pay (dollars per QALY) threshold. No follow-up for patients with aortas smaller than 50 mm was the optimal strategy in more than 50% of simulations up to a willingness-to-pay threshold of $60 000 per QALY. At $100 000 per QALY, the three most frequent optimal strategies were as follows: no follow-up for patients with aortas smaller than 50 mm (39% of simulations), follow-up for patients younger than 55 years with dilated aortas 45 mm or greater every 3 years (21% of simulations), and follow-up for patients younger than 65 years with dilated aortas 40 mm or greater every 3 years (14% of simulations).

Figure 4:

Probabilistic sensitivity analysis. The fraction of simulations for which each strategy is optimal at a given willingness-to-pay (incremental cost-effectiveness ratio) threshold. Strategies that were optimal in less than 1% of runs are suppressed. Strategies are labeled as in Figure 2 (ie, age cutoff [in years]/aortic size cutoff [in millimeters]/frequency [in years]). No follow-up for patients with aortas less than 50 mm is the cost-effective strategy for most simulations under a willingness-to-pay threshold of $60 000. At $100 000, the cost-effective strategies included no follow-up for patients with aortas less than 50 mm (39% of simulations), follow-up for patients younger than 55 years with aortic size of 45 mm or greater every 3 years (21% of simulations), and follow-up for patients younger than 65 years with aortic size of 40 mm or greater every 3 years (14% of simulations).

Discussion

We performed a simulation-based cost-effectiveness analysis of follow-up CT for incidentally detected dilated ascending aortas measuring 40–50 mm. With the base-case parameter values, the cost-effective strategy with the highest QALYs under a $100 000 per QALY threshold was follow-up for patients younger than 60 years with dilated aortas 40 mm or greater every 3 years (ICER of $62 511; 95% CI: $52 168, $77 739). The optimal strategy depended on model parameter values; for example, if mean aortic growth rate was decreased to 0.1 mm per year from 0.2 mm per year, then the optimal strategy changed to no follow-up for aortas smaller than 50 mm. Probabilistic sensitivity analysis demonstrated that, under a range of parameter values, the cost-effective strategies at a $100 000 per QALY threshold included the following: no follow-up for aortas smaller than 50 mm (39% of simulations), age younger than 55 years with aorta size 45 mm or greater every 3 years (21%), and age younger than 65 years with aorta size of 40 mm or greater every 3 years (14%). These results show that, given parameter uncertainty, follow-up for dilated ascending aortas less than 50 mm may not be cost-effective; even if follow-up CT is cost-effective in general, follow-up for older patients (>60–65 years of age) with aortas smaller than 50 mm is not cost-effective. By restricting follow-up to only younger patients, follow-up imaging can be substantially reduced, from 74% of the simulated cohort to 17%.

The limited cost-effectiveness of follow-up for dilated aortas in older patients (>60–65 years) likely reflects the fact that dilated aortas in patients without aortopathy tend to grow slowly, with a range of 0.1–0.3 mm per year in several large studies (9–11). Indeed, in many patients, this aortic size may simply reflect the patient’s normal baseline rather than an active process. Mildly enlarged aortas do not reach sizes at high risk for dissection or rupture until many years later, if ever, and older patients are likely to die of other causes in the interim. The importance of age-based cutoffs for indolent processes has been shown for other incidental findings, such as thyroid nodules (14); this approach aligns with the goal of personalized medicine.

Of note, our analysis assumed that the dilated aorta was an incidental finding, and there was no evidence of aortopathy or known aortic valvular disease. Patients with genetic (eg, Marfan syndrome) or bicuspid aortopathy tend to have faster growing aortas (9) and higher risk of rupture at any given aortic size. Thus, the model parameters used in this simulation would not apply. Moreover, patients with aortopathy generally merit surgical intervention at lower size thresholds than assumed in this simulation (4) and likely have a higher rate of postoperative morbidity and mortality. If a radiologist or cardiologist identifies a dilated aorta at CT with imaging features suggesting an aortopathy, that finding would merit immediate further work-up and close follow-up. Similarly, the presence of aortic valvular calcifications or severe coronary artery calcifications would be clues to aortic stenosis or coronary artery disease that may merit aneurysm repair at a lower threshold.

Our study had several limitations, most notably that it was based on a simulation model and was therefore dependent on parameter inputs. Uncertainty about these parameters leads to some uncertainty about the study conclusions. However, by using sensitivity analyses, particularly probabilistic sensitivity analysis, we showed that the results did not deviate too far from our main conclusions (ie, that follow-up for older patients is not cost-effective) despite varying parameter values within a reasonable range. Other limitations include some model assumptions, most notably that the aortic growth rate was assumed to be constant for the patient’s lifetime. This assumption is supported by at least one study showing linear growth over the course of approximately 8 years (12), although it may be less true for larger aortas and longer time periods. We also did not generate sex- or body size–specific intervention thresholds, given that current surgical guidelines mainly focus on absolute aortic size. Finally, the costs and willingness-to-pay threshold used in this analysis are for U.S. health care and may differ for other countries.

In conclusion, our simulation model demonstrates that follow-up CT for incidentally discovered dilated ascending aortas measuring 40–50 mm is likely cost-effective for patients younger than 60 years of age, with an optimal imaging interval of 3 years, while follow-up CT in older patients (>60 to 65 years) is likely not cost-effective. Future society guidelines and recommendations should consider the potentially limited benefit of follow-up for older patients, which may contribute to reduced unnecessary imaging and patient anxiety. Prospective trials are needed to confirm these results.