Simulation-based estimates of effectiveness and cost-effectiveness of smoking cessation in patients with chronic obstructive pulmonary disease

Abstract

Background: The medico-economic impact of smoking cessation considering a smoking patient with chronic obstructive pulmonary disease (COPD) is poorly documented.

Objective: Here, considering a COPD smoking patient, the specific burden of continuous smoking was estimated, as well as the effectiveness and the cost-effectiveness of smoking cessation.

Methods: A multi-state Markov model adopting society's perspective was developed. Simulated cohorts of English COPD patients who are active smokers (all severity stages combined or patients with the same initial severity stage) were compared to identical cohorts of patients who quit smoking at cohort initialization. Life expectancy, quality adjusted life-years (QALY), disease-related costs, and incremental cost-effectiveness ratio (ICER: £/QALY) were estimated, considering smoking cessation programs with various possible scenarios of success rates and costs. Sensitivity analyses included the variation of model key parameters.

Principal findings: At the horizon of a smoking COPD patient's remaining lifetime, smoking cessation at cohort intitialization, relapses being allowed as observed in practice, would result in gains (mean) of 1.27 life-years and 0.68 QALY, and induce savings of -1824 £/patient in the disease-related costs. The corresponding ICER was -2686 £/QALY. Smoking cessation resulted in 0.72, 0.69, 0.64 and 0.42 QALY respectively gained per mild, moderate, severe, and very severe COPD patient, but was nevertheless cost-effective for mild to severe COPD patients in most scenarios, even when hypothesizing expensive smoking cessation intervention programmes associated with low success rates. Considering a ten-year time horizon, the burden of continuous smoking in English COPD patients was estimated to cost a total of 1657 M£ while 452516 QALY would be simultaneously lost.

Conclusions: The study results are a useful support for the setting of smoking cessation programmes specifically targeted to COPD patients.

Figure 1. Flowchart describing the Markov multi-state model used.

X value is 1, 2, 3, or 4, respectively corresponding to GOLD 1, GOLD 2, GOLD 3 and GOLD 4 stages. Each of these healthcare states is associated with a corresponding utility and cost. ^*Value “X+1” does not exist for X = 4 (stage GOLD4). Nodes marked with an “M” represent Markov process chance nodes, while full square, full circle, and full triangle nodes correspond to decision, chance, and terminal nodes, respectively. Evolution of the cohort is made with one-year iteration step. Each patient is followed until death (all causes of death in COPD patients). At each iteration (Markov node), a given patient in a given X severity stage, is first subjected to a potential change in his/her smoking status, reflecting the background turnover observed in COPD patients (top). As indicated in Table 1, turnover probabilities were constant over age and COPD severity stages. Then (bottom), he might experience exacerbations (that only depend on patient's COPD current severity stage, as indicated in Table 1). In the end, the patient may 1) stay in the same severity stage, 2) pass to the next severity stage (X+1), 3) die. Transition probabilities from one stage to the next depend on age, severity stage, and smoking status (see parameter values in Table S2 in supporting information). Transition probabilities to death depend on the same parameters and in addition, on exacerbation status (see parameter values in in Table S3 in supporting information). As compared to current smokers, ex-smokers had a lower disease progression, and a lower probability of death.

Figure 2. Sensitivity analysis.

In each of the A to K explored scenarios (top to bottom), the value of a key parameter was changed as compared to the reference case scenario. For each of these scenarios, the figure indicates how simulation outputs (i.e. difference Δ between intervention and non intervention in terms of QALY, Cost, ICER) change, as compared to the simulation outputs of the reference scenario (reference case for which ΔQALY, ΔCost, ΔICER were 0.679 QALY, −1824 £ and −2686 £/QALY, respectively, see Table 1 for parameter values and Table 3 for more detailed simulation outputs). For example, in scenario A, ΔQALY, ΔCost, ΔICER were 0.817 QALY, −1262 £ and −1544 £/QALY, therefore representing respectively a 20% ((0,817−0,679)/0,679), a −31%, and a −42% change, as compared to the reference case scenario. Scenarios A to K correspond to the following modifications of parameter values as compared to those used in the reference case: A, the proportion of exacerbation-free patients among ex-smokers was raised by 30%; B, the increased risks of death in COPD patients (as compared to individuals of the standard population were set to the upper limits reported by Mannino et al ; C, the probability of death was increased by 30%; D, the increased risks of death in COPD patients (as compared to individuals of the standard population) were set to the lower limit reported by Mannino et al ; E, the proportion of exacerbation-free patients among ex-smokers was raised by 15%; F, health utilities and costs were not discounted; G, no disease management costs for GOLD1 patients; H, the proportion of exacerbation-free patients was raised by 15%; I, health utilities and costs were discounted at the rate of 5%; J, disease management costs increased by 15% for each severity stage; K, the transition rate from one stage to the next was increased by 30%.