Radial probe EBUS versus CT-guided needle biopsy for evaluation of peripheral pulmonary lesions: an economic analysis

Study site

The Royal Melbourne Hospital in Melbourne, Australia, is a tertiary referral centre for the diagnosis, staging and management of lung cancer, with substantial experience in both EBUS-TBLB and CT-PNB. The hospital serves a catchment area of over 600,000 people. Patients with suspected/known lung cancer are managed by a multidisciplinary team comprising respiratory physicians, thoracic radiologists, thoracic surgeons, medical oncologists and radiation oncologists. The multidisciplinary team manages approximately 300 patients with lung cancer per year.

Modelling approach

Decision analysis, using specialised software (TreeAge Pro 2009, Excel module; TreeAge Software Inc., Williamstown, MA, USA), was applied to compare the downstream costs of EBUS-TBLB and CT-PNB (fig. 1) [11]. The pathway demonstrating lower healthcare costs (i.e. cost minimisation) is identified as the more cost-beneficial pathway. The analysis accounted for costs of each procedure, as well as costs incurred as a result of extra procedures required in the event of a negative result from either modality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1–

Decision tree illustrating possible clinical pathways following selection of either diagnostic approach. EBUS: endobronchial ultrasound; PPL; peripheral pulmonary lesion; CT: computed tomography. Square: decision node, i.e. the clinician may choose any clinical pathway for an individual patient. Circle: chance node, i.e. patients may experience either outcome, based on chance. The proportion of patients following each pathway from a chance node is dependent on pre-defined clinical parameters (table 3). Triangle: terminal node in the decision pathway, i.e. an individual patient has reached a definitive outcome in their diagnostic pathway.

An advantage of decision tree analysis is its capacity to simulate even complex clinical algorithms, such as that for the evaluation of PPL. Furthermore, it can explicitly capture the uncertainty that is inherent in modelling of any type [12].

Model population

The modelled population comprised hypothetical patients referred to a multidisciplinary team for evaluation of PPL, for whom the team felt investigation was warranted and that either CT-PNB or EBUS would be acceptable modes of initial investigation of the lesion. Therefore, this excluded patients with the following features: 1) a clinical condition precluding investigation; 2) a lesion <1 cm diameter anywhere in lung fields; 3) evidence on CT scan of central (endobronchially visible) lesion; and 4) other clinical site of disease more amenable to tissue diagnosis.

Healthcare costs

Unit cost estimates, in Australian dollars (AU$), were based on recorded hospital costs for patients undergoing the above-mentioned procedures at the Royal Melbourne Hospital between February 7, 2008 and January 22, 2010. All patients had provided written consent for inclusion in a randomised pragmatic trial comparing EBUS-TBLB with CT-PNB [10].

EBUS-TBLB and CT-PNB are performed on an outpatient basis at the Royal Melbourne Hospital. EBUS-TBLB is performed in a day procedure unit, with sedation administered by resident staff from the Respiratory Unit, as previously described [13]. The procedure itself has previously been described [14], using a 20-MHz radial EBUS probe (UM-BS20–26R; Olympus, Tokyo, Japan) and guide sheath. CT-PNB is performed using a coaxial needle (Bard TruGuide needle; Bard Biopsy Systems, Tempe, AZ, USA) and core biopsy instrument (Bard Biopy-Cut needle and Bard Magnum biopsy instrument; Bard Biopsy Systems).

Costs were derived from actual patient data at the Royal Melbourne Hospital, and includes both direct care costs (e.g. physician, nursing, radiology and pathology costs) as well as indirect costs such as equipment sterilisation and repair costs, and non-clinician staff costs (e.g. clerical or cleaning staff). Costing data for each patient admission was obtained from cost weight analysis compiled according to guidelines from the Clinical Costing Standards Association of Australia [15]. Hospital and median costs for EBUS-TBLB and CT-PNB were calculated based upon all patients included in a recently published randomised pragmatic trial [10]. Summary data for all uncomplicated procedures are shown in table 1. Costs for patients in whom complications occurred are shown in table 2. Costs for thoracoscopic resection were established following an audit of all patients undergoing thoracoscopy/thoracotomy for resection of lung lesions at the Royal Melbourne Hospital from July 1, 2007 to June 30, 2008. All costs were updated to 2010/2011 levels according to the locally recorded Health Price Index, which reported an increase of 3% per year [16].

Other input parameters

Other input parameters applied to the decision tree analysis are described in table 3. Sensitivity and specificity of EBUS-TBLB for evaluation of PPL was based on our own experience and a published meta-analysis [10, 17], while data for CT-PNB was based on our reported experience and published guidelines [2, 10].

Sensitivity analysis

Calculations based on the above data constituted a “base-case” analysis, as defined by National Institute for Health and Clinical Excellence guidelines [18]. We recognised that model input values may vary significantly across different institutions. For example, diagnostic sensitivity of EBUS-TBLB differs considerably between institutions [17], and there is significant discrepancy in reported complication rates following CT-PNB [2]. Therefore, a series of one-way sensitivity analyses were undertaken within the range of each parameter recorded in table 3, based on data from recent pooled analyses. The values of these key inputs were varied one at a time, while maintaining the other inputs at “base-case” values. Subsequent analysis was undertaken to determine the threshold above which the most cost-beneficial approach remained in comparison to other diagnostic modalities.

Cost may also alter, depending on the severity of the condition, and institutional approaches to management (e.g. in- versus outpatient care and frequency of intercostal catheter insertion). Sensitivity analysis was performed to determine if a threshold cost for complications existed, above which the alternate investigation modality proved more cost-beneficial.

In order to assess the impact of uncertainty more accurately, probabilistic sensitivity analysis was performed using Monte Carlo simulations [19]. With this method, input parameters are assigned a distribution to reflect the nature of uncertainty. Multiple model simulations are then run. Monte Carlo simulation was performed using triangular distributions of values (lowest, likeliest, highest) as recorded in table 4. With each simulation, one value from every input range is randomly sampled from within a specified data range according to its probability distribution. Thus, multiple outputs are generated, and uncertainty ranges are derived from the distributions of these. In our analysis, 10,000 simulations were undertaken.

Cost-effectiveness

The above methodology is used to assess the comparative cost-benefit of competing diagnostic strategies for assessment of PPL. Cost-effectiveness requires consideration of quality-of-life measures. Patient preferences with regard to the impact of procedural complications or anxiety related to waiting for test results have been shown to influence cost-effectiveness analyses for patients with PPL [20].

Cost-effectiveness outcomes are expressed in cost per quality-adjusted life-year (QALY), with utility being the measure on which quality adjustment is based. Utility allows adjustment of life-years gained by an intervention when those gained years would be lived in less than perfect health. Extra life-years are given a utility value of between 0 and 1 to account for this. This method is suitable for assessment of chronic health/disease states, although it is not able to assess the cost impact of short-term disease states, such as pain or complications arising from a diagnostic procedure, or the anxiety resulting from a non-diagnostic procedure [21].

Multiple methods for assessment of the impact of transient disease states have been described. With the time-trade-off (TTO) technique, a patient decides between a longer period of time in less optimal health versus a shorter period in good health. A variation, the wait-trade-off technique, quantifies patients’ preference for undergoing a particular test or treatment that has associated discomfort or restrictions that the patient may dislike. The patient is asked to trade-off extended time with the condition being diagnosed or treated in order to avoid the noxious effects of the test or treatment in favour of a similarly effective test or treatment but one not having side-effects [22]. A QALY toll is reflected in the wait-trade-off by an individual's willingness to wait longer to avoid more noxious experiences [23] and may be measured by disutility, being the fraction of a year of perfect health a patient would be willing to give up to avoid having to undergo a diagnostic test and to avoid its short-term morbidity [24]. This tool was originally designed for use in states related to diagnostic screening and testing [25].

Sensitivity analysis was performed for disutility, starting at a theoretical disutility of 0 for both the procedure itself (that is no utility penalty), as well as disutility attributable to complications arising from the procedure. One-way sensitivity analysis was performed to identify theoretical thresholds that may influence cost-effectiveness outcomes.

Assumptions

As sensitivity analysis is based on theoretical patients, we were required to make some specific assumptions regarding the theoretical model population. Key assumptions in the analysis were as follows. 1) There was a well-defined outcome in each arm of our decision model, i.e. pathologic diagnosis of PPL. 2) The long-term outcomes (measures of effectiveness) were equivalent in each model arm, i.e. treatment and outcomes of all patients was similar regardless of how the diagnosis was determined. As previously recognised [26], a cost-benefit analysis that assumes competing diagnostic strategies has equivalent outcomes and focuses, thereafter, only on cost outcomes is the most appropriate form of economic analysis to use in this setting. 3) Once a diagnosis has been made, the downstream costs of medical care were the same, regardless of how the diagnosis was achieved. 4) Thoracotomy/thoracoscopy had a diagnostic accuracy of 100% in the evaluation of PPL. 5) Pathology costs were identical regardless of the method of tissue acquisition.