Cost-effectiveness microsimulation of catheter-directed thrombolysis in submassive pulmonary embolism using a right ventricular function model

STRUCTURED ABSTRACT

Background:

Approximately 30–50% of hemodynamically stable patients presenting with acute pulmonary embolism (PE) have evidence of right ventricular (RV) dysfunction. These patients are classified as submassive PE and the role of reperfusion therapy remains unclear. We sought to identify the circumstances under which catheter-directed thrombolysis (CDT) would represent high-value care for submassive PE.

Methods:

We used a computer-based, individual-level, state-transition model with one million simulated patients to perform a cost-effectiveness analysis comparing the treatment of submassive PE with CDT followed by anticoagulation to treatment with anticoagulation alone. Because RV function impacts prognosis and is commonly used in PE outcomes research, our model used RV dysfunction to differentiate health states. One-way, two-way, and probabilistic sensitivity analyses were used to quantify model uncertainty.

Results:

Our base case analysis generated an incremental cost-effectiveness ratio (ICER) of $119,326 per quality adjusted life year (QALY). Sensitivity analyses resulted in ICERs consistent with high-value care when CDT conferred a reduction in the absolute probability of RV dysfunction of 3.5% or more. CDT yielded low-value ICERs if the absolute reduction was less than 1.56%.

Interpretation:

Our model suggests that catheter-directed thrombolytics represents high-value care compared to anticoagulation alone when CDT offers an absolute improvement in RV dysfunction of 3.5% or more, but there is substantial uncertainly around these results. We estimated the monetary value of clarifying the costs and consequences surrounding RV dysfunction after submassive PE to be approximately $268 million annually, suggesting further research in this area could be highly valuable.

BACKGROUND

The incidence of acute pulmonary embolism (PE) in the United States is estimated at 35/100,000 persons annually and appears to be increasing [1,2]. Patients with overt hemodynamic collapse have high mortality risk and the presence of circulatory compromise is an indication for prompt reperfusion therapy [3–5] [6,7]. As many as 30–50% of hemodynamically stable patients nevertheless have evidence of right ventricular (RV) dysfunction by imaging or serum biomarkers [8,9] and are classified as submassive PE. The role of reperfusion therapy in this population remains unclear [10].

Trials of systemic thrombolytics in submassive PE demonstrated an increased risk of major bleeding and intracranial hemorrhage (ICH). As a result, thrombolysis is recommended only as a rescue agent should hemodynamic decompensation occur [6] [11–13]. This led to interest in catheter-directed thrombolytics (CDT), which have demonstrated improvement in RV function without increased ICH risk [14–16].

We performed a cost-effectiveness analysis to evaluate the circumstances under which treatment of submassive PE with CDT represents high-value care compared to anticoagulation alone.

METHODS

Model overview

A computer-based, individual-level, state-transition (i.e. microsimulation) model was constructed to evaluate the cost-effectiveness of CDT (followed by anticoagulation) compared to anticoagulation alone from the healthcare sector perspective. As RV function impacts post-PE prognosis and is commonly used in PE outcomes research, this model primarily uses RV function to differentiate health states (Figure 1).

Figure 1. State-transition schematic.

Simulated patients enter the model 30 days after index event. Initial health states are distinguished based on the probability of ongoing ventricular (RV) dysfunction given index treatment received. With each cycle, patients may remain in their health state or transition to another.

CTEPH chronic thromboembolic pulmonary hypertension; ICH intracranial hemorrhage; PE pulmonary embolism; RV right ventricle

Each patient enters the model 30 days after their index PE event, having received either CDT or anticoagulation alone. The patient’s initial health state is determined by their probability of residual RV dysfunction, based on the treatment they received.

The cycle length, or time between transitions, is 30 days. At the end of a cycle, the patient may remain in the same health state or change health states due to an event, such as severe extracranial bleeding, ICH, recurrent PE, development of chronic thromboembolic pulmonary hypertension (CTEPH), or death. Cycles continue, with the patient accruing costs and utilities, until the patient dies or reaches 100 years of age. Utilities are measured in quality-adjusted life years (QALYs), which are a measure of disease burden.

Comparison between treatment strategies is accomplished through the use of incremental cost effectiveness ratios (ICER), which are the difference in cost divided by the difference in outcome. Per the American College of Cardiology and American Health Association practice guidelines, a high-value intervention is associated with an ICER < $50,000/QALY, intermediate-value with an ICER between $50,000–150,000/QALY, and low-value with an ICER > $150,000/QALY [17]. A strategy is dominant if, under the specified conditions, it results in more QALYs for less cost than the alternatives.

Initial patient characteristics of age and sex were generated to mimic the patients in the National Readmissions Database (NRD). [18] The proportion of patients with pre-existing conditions requiring lifelong anticoagulation was generated from a composite of the NRD, the PEITHO trial, and the prevalence of atrial fibrillation (eTable 1). [11,18,19] All-cause mortality rates based on age and sex were derived from the National Vital Statistics Report from the Centers for Disease Control [20].

Base case

Model parameters of probability, cost, and health state utility were derived from published, peer-reviewed literature on submassive PE and are listed in Table 1 and eTable 2. In our base case analysis, we simulated 1,000,000 patients in each treatment arm with a baseline probability of residual RV dysfunction of 20% and an odds ratio of residual RV dysfunction after CDT of 0.878.

Table 1.

Select model parameters. Probabilities and utilities are listed in annual terms unless otherwise specified. For a complete list of model inputs, please refer to eTable 2.

Variable	Base case value	95% CI or standard dev	Distribution	Sensitivity analysis range	Source
Probabilities
RV dysfunction after PE	20%	10.2, 32.0	beta	10–35%	[23]
OR for RV dysfunction after CDT	0.878			0.5–1	[23]
30-day mortality after PE (CDT)	2.67%	0.88, 4.57	beta	1–5%	[14,16,15]
30-day stroke risk after PE (CDT)	0.20%	.02, 0.55	beta	0.2–0.6%	[14–16]
Intracranial hemorrhage while on AC	0.28%	0.18, 0.40	beta	0.1–0.6%	[25]
Recurrent PE with normal RV function	1.86%	1.44, 2.32	beta	1.2–3.5%	[25]
OR for recurrent PE with RV dysfunction	1.74			1–2.6	[25]
Development of CTEPH	0.97%	0.90, 2.30	beta	0.5–2.3%	[39]
Health State Utilities
RV dysfunction	0.764	0.60, 0.89	beta	0.60–0.84	[40,27]
Costs (2017 USD)
Event hospitalization (CDT)	28,770	9,711, 68,577	gamma	9,700–68,600	[18]
Event hospitalization (AC)	12,842	16,529	gamma	5,000–28,000	[41]
Anticoagulant costa	369	190, 621	gamma	150–650	[42]
Recurrent PE event	38,484	58,151	gamma	1,500–100,000	[43]
RV dysfunction carea	2,606	3,106	gamma	0–5,710	[44]
CTEPH carea	2,914	3,788b	gamma	0–6,700	[45,46]

^aMonthly (per cycle) cost

^bauthor assumption

CI = confidence interval, RV = right ventricle, OR = odds ratio, CDT = catheter directed therapy, PE = pulmonary embolism, AC = anticoagulation, CTEPH = chronic thromboembolic pulmonary hypertension

Treatment assumptions

CDT has been demonstrated to reduce RV strain in the acute setting, however the data regarding the long-term impact of thrombolytics (CDT or systemic) on the RV compared to anticoagulation alone is conflicting and sparse. Echocardiographic follow-up at a median of 38 months in the PIETHO trial of systemic thrombolysis demonstrated no significant difference in RV function or pulmonary hypertension [21]. This stands in contrast to prior studies that have demonstrated greater improvement in pulmonary hypertension and RV function in patients treated with thrombolytics at 3, 6, and 28-month follow-ups [22–24]. Given the lack of consensus in the literature, we chose a base case with a small difference between the two strategies at 30 days, given by the odds ratio of 0.878 for those treated with CDT.

Improved likelihood of RV recovery after thrombolysis may be important not only for quality of life, but to reduce risk of recurrent PE. In a large venous thromboembolism anticoagulation trial (in which no patients received thrombolytics), the subgroup with RV dysfunction had 1.74 times the odds of recurrence compared to those without RV dysfunction [25]. Residual RV dysfunction may be an indication of residual, flow-limiting clot burden pre-disposes to additional PE events [26]. Treatment with thrombolytics would be expected to reduce this risk of recurrent PE; a metanalysis of thrombolytic trials found patients treated with thrombolysis had 0.40 times the risk of recurrent PE compared to those treated with anticoagulation alone [12]. Our base case used the more conservative of the two values, an odds ratio of 1.74 for recurrence in the setting of RV dysfunction.

Technically, we applied odds ratios by converting probabilities to odds, multiplying by the odds ratio, and converting the adjusted odds back to probabilities.

Patients in both strategies received anticoagulation for six months after their PE event unless they had a pre-existing condition requiring lifelong anticoagulation or they suffered a recurrent PE, in which case anticoagulation was continued indefinitely.

Costs and utilities

The utilities for RV dysfunction and CTEPH have not been well-established in the literature. As a direct determination is beyond the scope of this study, estimates were made from the range of heart failure utilities reported that would also be consistent with the World Health Organization functional class for pulmonary hypertension [27]. It was assumed that CTEPH would have lower utility than RV dysfunction.

Similarly, the costs associated with treatment of RV dysfunction are not well delineated. We used estimates from non-systolic heart failure and applied a wide standard deviation. Non-systolic heart failure was chosen over pulmonary arterial hypertension (PAH) as PAH costs would likely be inflated by expensive, disease-specific medications and, diagnostically, the presence of PAH in the setting of prior clot would overlap with CTEPH.

Costs are reported in US dollars and were adjusted to 2017 USD using the Consumer Price Index for Medical Care and rounded to the nearest whole dollar. All costs and health utilities were discounted at an annual rate of 3% and adjusted with half-cycle correction.

Uncertainty

One-way sensitivity analyses used 1,000,000 patients in the micro-simulation over the ranges noted in Table 1. Two-way sensitivity analyses used a 1,000,000 patient Markov cohort.

Overall model uncertainty was assessed with a probabilistic sensitivity analysis (PSA) in which 1000 (outer loop) iterations of the 100,000 (inner loop) patients in the microsimulation model were run, each with the model parameter randomly drawn from a pre-specified probability distribution (Table 1, e-Table 2).

Modeling and analysis were performed using TreeAge Pro 2019 (TreeAge Sofware, Williamstown, MA).

RESULTS

In the base case, simulated patients incurred a mean lifetime cost of $255,408 for CDT and $245,393 for anticoagulation alone and resulted in a mean of 10.90 QALYs and 10.81 QALYs, respectively. The incremental cost-effectiveness ratio (ICER) for CDT in the base case was $119,326/QALY, which falls in the intermediate-value range defined by the ACC-AHA.

Under the base case conditions, 326,545 people treated with CDT and 327,179 people treated with anticoagulation alone had at least one recurrent PE, which is consistent with published data on long-term outcomes for PE [28,29].

One-way sensitivity analyses conducted on model parameters are summarized in Figure 2, with the ranges for each parameter listed in Table 1.

Figure 2. One-way sensitivity analyses on key model parameters.

The incremental cost effectiveness ($/QALY) as the variable value is varied along a range of plausible values. The red bar represents the base case. Values to the left of the first black bar represent high-value care and values to the right of the second black bar represent low-value care.

AC anticoagulation; CDT catheter-directed therapy; CTEPH chronic thromboembolic pulmonary hypertension; ICH intracranial hemorrhage; OR odds ratio; PE pulmonary embolism; QALY quality adjusted life year; RV right ventricle

Probability of right ventricular dysfunction

When holding the probability of RV dysfunction with anticoagulation alone constant at 20% and performing a one-way sensitivity analysis on the odds ratio for residual RV dysfunction after CDT, the ICER for CDT remained below the low-value threshold while the odds ratio remained below 0.904, which corresponds to an absolute probability difference of 1.56%. The ICER for CDT fell into the high-value range below an odds ratio of 0.792, which corresponds to an absolute probability difference of 3.47%. Anticoagulation alone was dominated by CDT when the odds ratio fell below 0.713, which corresponds to an absolute probability difference of 4.88%.

Similarly, holding the odds ratio for RV dysfunction constant at 0.878 and performing a one-way sensitivity analysis of the probability of RV dysfunction with anticoagulation alone, the ICER for CDT remained below the low-value threshold while the probability remained below 20.75%, which corresponds to an absolute probability difference of 2.06%. The ICER for CDT fell into the high-value range at a probability of 21.88%, which corresponds to an absolute probability difference of 2.14%. Anticoagulation alone dominated CDT when the probability of RV dysfunction after anticoagulation alone fell below 14.35%, which corresponds to an absolute probability difference of 1.53%, and anticoagulation alone was dominated by CDT when the probability was greater than 23.33%, which corresponds to an absolute probability difference of 2.24%.

A two-way sensitivity analysis of these parameters taken jointly demonstrated that the absolute difference in post-event probability of residual RV dysfunction between the two strategies required to make CDT cost-effective with a high-value ICER was 3.2%. At absolute differences less than 1.43%, CDT represents a low-value intervention (eFigure 1).

Costs associated with treatment

The model was sensitive to multiple cost parameters. One-way sensitivity analyses on the cost of index hospitalization demonstrated that a hospitalization cost of at most $22,618 for CDT was associated with high-value care, while a cost of $30,778 or more was associated with low value care. Anticoagulation alone was dominated when the hospitalization cost for CDT fell below $18,538 or the hospitalization cost for anticoagulation alone rose above $23,117. A cost for index episode more than $18,954 for anticoagulation alone was associated with CDT being classified as high-value care. A two-way sensitivity analysis on cost of index event with each treatment option produced similar figures, with CDT representing high-value care when costs were between $15,782 and $38,762 depending on the cost of anticoagulation alone (eFigure 2).

Increasing monthly costs associated with RV care are associated with increasing cost-effectiveness; under the base case conditions, CDT represents high-value care when RV dysfunction had a monthly cost more than $5,086 and low-value care when RV dysfunction had a monthly cost less than $1618. Model results were robust to plausible changes in the costs of a recurrent PE event, anticoagulation, and CTEPH care.

Overall model uncertainty

In a probabilistic sensitivity analysis, the mean ICER for CDT therapy was $82,543/QALY. CDT was consistent with high value care in 36.1% of iterations and low value care in 40% of iterations. At a cost-effectiveness threshold of $104,464/QALY, the two strategies had an equal chance of being cost-effective (Figure 3). Under the base case conditions, the estimated value of perfect information is $10,065 per case.

Figure 3. Cost effectiveness acceptability frontier and expected value of perfect information from probabilistic sensitivity analysis.

Dashed lines are the probability of each strategy being cost-effective across all iterations of the probabilistic sensitivity analysis for a range of thresholds. The cost-effective acceptability frontier (CEAF) is the probability of the optimal strategy being cost-effective. The dotted line is the expected value of perfect information (EVPI) at each threshold.

AC anticoagulation; CDT catheter directed thrombolysis; QALY quality adjusted life year

DISCUSSION

We developed an individual, state-transition model to evaluate the costs and consequences of catheter-directed thrombolysis for treatment of acute submassive pulmonary embolism compared to treatment with anticoagulation alone. We found that CDT represented high-value care, that is, an ICER of $50,000/QALY or less, compared to anticoagulation alone when CDT offers an absolute risk reduction of at least 3.5% in chronic RV dysfunction.

Our model builds on prior analyses of the cost-effectiveness of the therapeutic options for submassive PE due to its use of RV function as the discriminant for health state, which more closely mirrors clinician decision making and outcomes measured in treatment efficacy trials [30,31]. It further provides a mechanism through which long-term sequalae, such as recurrent PE or CTEPH can be modeled based on acute treatment decisions.

As might be expected by the structure of the model, the results were sensitive to the costs of index event, the costs associated with ongoing RV dysfunction, and the proportion of patients experiencing ongoing RV dysfunction. The model was very sensitive to the risk of ICH, which was a concern with systemic thrombolytics in the PIETHO trial. To date, the risk does not appear similarly elevated in CDT, though the number of patients receiving CDT remains small and the ICER would increase if further experience negates this assumption [32,33]. Similarly, the risk for peri-procedural bleeding will need to be monitored as it would increase the cost of index event.

The true difference in RV outcomes and the sequelae of thrombolytics compared to anticoagulation alone require further study. The value of perfect information derived from our PSA was $10,065 per case, which translates to approximately $268 million per annum, which suggests that clarifying the impact of thrombolytic therapies on the RV is likely to be of high value (e-Table 3).

Limitations to this study include modeling only two treatment paradigms as well as the lack of explicitly modeling “post-PE syndrome” or “chronic thromboembolic disease”, which are periods following a PE event marked by functional limitation and impaired quality of life. While their prevalence and mechanism are not fully elucidated, RV dysfunction, elevated mean pulmonary artery pressures, and thrombotic burden have all been implicated [34]. This suggests that a clinical outcome, such as RV function, should be used as a health state discriminant rather than acute treatment strategy alone.

We did not model reduced-dose systemic thrombolytics, also referred to as “half-dose tPA” due to significant differences in the dosing regimens used in trials to date. Nevertheless, those trials suggest a comparably low risk of ICH events and mechanistically, would also be expected to improve right heart function over anticoagulation alone [24,35,36].

Neither did we model any mechanical clot fragmentation devices as there is a substantial paucity of data. For example, the AngioJet (Boston Scientific, Minneapolis, MN) has a black box warning regarding use in the pulmonary vasculature, the Angiovac (Angiodynamics Inc, Lantham, NY) requires veno-venous bypass, and the FlowTriever (Inari Medical, Irvine, CA) is currently under investigation in a clinical trial [37,38].

Our model assumed an anticoagulation duration of six months after index event in the absence of conditions favoring lifelong anticoagulation. Removal of anticoagulation is known to increase risk of recurrent PE, so if our estimates of the proportion of the population whom would discontinue anticoagulation were too high, this could bias our results. However, our model generated a number of recurrent events similar to published 10-year recurrence data, which suggests our estimates of the population composition were reasonable. [29,28]

Lastly, the data for the cost and utility of RV dysfunction, especially in the absence of CTEPH, has not been described in the literature. Our model is sensitive to both of these variables and thus more precise estimates of these quantities would improve our estimates.

CONCLUSIONS

In conclusion, our model suggests that catheter-directed thrombolytics for the treatment of submassive pulmonary embolism represents high-value care compared to anticoagulation alone when CDT offers an absolute improvement in RV dysfunction of 3.5%, but there is substantial uncertainly around these results. Due to this uncertainty, we estimated the monetary value of clarifying the costs and consequences surrounding RV dysfunction after submassive PE to be approximately $268 million per year, suggesting further research in this area could be highly valuable.

KEY POINTS:

• The role of reperfusion therapy in hemodynamically stable patients with evidence of right heart dysfunction after acute pulmonary embolism is unclear.

• We used a Markov microsimulation model to perform a cost-effectiveness analysis comparing the treatment of pulmonary embolism with catheter-directed thrombolytics to anticoagulation alone.

• Our model suggests that catheter-directed thrombolytics represents high-value care when they offer an absolute improvement in RV dysfunction of 3.5% or more.

• We estimated the monetary value of clarifying the costs and consequences surrounding right ventricular dysfunction after submassive pulmonary embolism to be approximately $268 million annually.

• Further research into the right ventricular consequences of lytic therapy for submassive pulmonary embolism is warranted.