Safety and Costs of Endobronchial Ultrasound-Guided Nodal Aspiration and Mediastinoscopy

Abstract

Background

There remains debate over the best invasive diagnostic modality for mediastinal nodal evaluation. Prior studies have limited generalizability and insufficient power to detect differences in rare adverse events. We compared the risks and costs of endobronchial ultrasound (EBUS)-guided nodal aspiration and mediastinoscopy performed for any indication in a large national cohort.

Methods

We conducted a retrospective study (2007-2015) with MarketScan, a claims database of individuals with employer-provided insurance in the United States. Patients who underwent multimodality mediastinal evaluation (n = 1,396) or same-day pulmonary resection (n = 2,130) were excluded. Regression models were used to evaluate associations between diagnostic modalities and risks and costs while adjusting for patient characteristics, year, concomitant bronchoscopic procedures, and lung cancer diagnosis.

Results

Among 30,570 patients, 49% underwent EBUS. Severe adverse events—pneumothorax, hemothorax, airway/vascular injuries, or death—were rare and invariant between EBUS and mediastinoscopy (0.3% vs 0.4%; P = .189). The rate of vocal cord paralysis was lower for EBUS (1.4% vs 2.2%; P < .001). EBUS was associated with a lower adjusted risk of severe adverse events (OR, 0.42; 95% CI, 0.32-0.55) and vocal cord paralysis (OR, 0.57; 95% CI, 0.54-0.60). The mean cost of EBUS was $2,211 less than mediastinoscopy ($6,816 vs $9,023; P < .001). After adjustment this difference decreased to $1,650 (95% CI, $1,525-$1,776).

Conclusions

When performed as isolated procedures, EBUS is associated with lower risks and costs compared with mediastinoscopy. Future studies comparing the effectiveness of EBUS vs mediastinoscopy in the community at large will help determine which procedure is superior or if trade-offs exist.

FOR EDITORIAL COMMENT, SEE PAGE 493

Endobronchial ultrasound (EBUS)-guided nodal aspiration and mediastinoscopy are invasive procedures used to diagnose mediastinal lymphadenopathy and to stage lung cancer. Several practice guidelines recommend EBUS over mediastinoscopy, citing equivalent diagnostic accuracy, greater safety, and lower total direct medical expenditures (heretofore referred to as “costs”).^{1, 2, 3, 4} However, the body of literature cited by guidelines has limitations, including potential publication bias, limited generalizability to practice in the community at large, and insufficient power to detect differences in rare, severe adverse events.^{5, 6, 7, 8} Varying opinions across professional societies, nations, and experts suggest uncertainty concerning the superiority of one procedure over the other.^{9, 10, 11, 12, 13, 14}

We sought to enhance the evidence base, using data from a large national cohort of individuals with employer-provided health insurance undergoing EBUS or mediastinoscopy for any indication. In accordance with the prevailing literature and several practice guidelines, we hypothesized that EBUS is associated with less risk and lower costs compared with mediastinoscopy.

Materials and Methods

Study Design, Population, and Data Source

We conducted a retrospective cohort study of adults who underwent EBUS or mediastinoscopy between January 1, 2007 and June 30, 2015. The data source was MarketScan—a database of inpatient and outpatient claims for US workers and their dependents with employer-provided commercial insurance.¹⁵ We excluded patients older than 64 years because we did not have access to claims for beneficiaries dually enrolled in commercial insurance and Medicare. In addition, we excluded individuals with less than 3 months of continuous enrollment before and after the use of EBUS or mediastinoscopy (unless disenrollment was due to death). Because we could not disentangle the effect of multiple procedures on outcomes and costs, we excluded patients who underwent both EBUS and mediastinoscopy and those who underwent EBUS or mediastinoscopy on the same day as a pulmonary resection (see e-Fig 1 for exclusions). Our institutional review board approved this study (STUDY00006405).

Mediastinoscopy and EBUS Use

The use of EBUS or mediastinoscopy was defined as the first occurrence of an International Classification of Diseases, 9th edition (ICD-9) procedure code and/or Common Procedure Terminology (CPT) code, using inpatient or outpatient claims during the study period (e-Table 1).

Covariates

Available variables within MarketScan included age, sex, region, health plan, and calendar year. Derived covariates included the following: (1) Charlson Comorbidity Index as modified by Klabunde et al¹⁶; (2) concomitant bronchoscopic procedure(s) performed on the same day as EBUS or mediastinoscopy; and (3) lung cancer diagnosis. We defined lung cancer diagnosis as either an ICD-9 diagnostic code for lung cancer 90 days before or after EBUS or mediastinoscopy, or CPT or ICD-9 diagnostic or procedure codes for lung cancer therapy 90 days after EBUS or mediastinoscopy.

Adverse Events

Adverse events were chosen on the basis of prior reports of these complications.^{17, 18, 19} Several rare complications, including pneumothorax, hemothorax, great vessel injury, airway injury, or death, were considered severe adverse events. Pneumothorax and hemothorax were defined by ICD-9 diagnostic codes and a hospitalization on the same day or an associated ICD-9 procedure or CPT code for treatment (eg, tube thoracostomy). Great vessel and airway injuries were defined by procedure codes for aortic and tracheal/bronchial repair, respectively. We measured deaths occurring on the same day as EBUS or mediastinoscopy or during a hospitalization starting on the same date as the procedure. Because vocal cord paralysis may not be recognized immediately after a procedure, we evaluated the incidence within 90 days of EBUS or mediastinoscopy, using a previously published algorithm.²⁰

Costs

Costs were calculated using professional service and facilities claims on the day of the procedure including any associated hospitalization. Costs included payments made by the insurer and patient, and therefore the economic analysis takes the perspective of both the payer and patient. Out-of-pocket costs were reported for descriptive purposes. Costs were adjusted for inflation using the Medical Consumer Price Index and reported in 2015 US dollars.²¹

Analysis

A χ² test was used to compare categorical variables, a Wilcox rank-sum test to compare medians, and a Student t test to compare means. Logistic regression was used to evaluate the relationship between diagnostic modality and adverse events while adjusting for patient characteristics (age, sex, region, health plan, and comorbidity index), year, concomitant bronchoscopy, lung cancer diagnosis, and clustering at the health plan level. A generalized linear model (γ distribution, identity link)²² was used to evaluate the relationship between diagnostic modality and costs while adjusting for potential confounding from patient characteristics, calendar year, concomitant bronchoscopy, severe adverse events, procedure-related hospitalization, and clustering at the health plan level. Variables were sequentially added to the cost model to understand the contribution of covariates to cost differences. Overall, 8.6% of patients had missing covariate data. The proportion of patients with missing data did not vary between mediastinoscopy and EBUS (e-Table 2). Adverse events did not vary significantly between patients with and without missing covariate data, although costs were significantly lower for patients with missing covariate data. We report only the case-complete results. Two-sided P values <.05 were considered statistically significant. STATA/SE version 14.2 (StataCorp LP) was used for all analyses.

Results

Among 30,570 patients undergoing mediastinal nodal evaluation, 49% underwent EBUS. The frequency of EBUS increased from 14% in 2007 to 79% by 2015 (P < .001). Patients undergoing EBUS tended to be older; to reside in the south; and to have a high-deductible health plan, higher comorbidity index, and lung cancer diagnosis (Table 1). Compared with mediastinoscopy, EBUS was associated with a higher rate of concomitant bronchoscopic procedures and a lower rate of hospitalization.

Table 1.

Cohort Characteristics

Characteristic	All	EBUS	Mediastinoscopy	P Value
	(N = 30,570)	(n = 15,097)	(n = 15,473)
Age, median (IQR), y	54 (46-59)	55 (47-60)	53 (45-59)	< .001
Men, No. (%)	15,373 (50)	7,517 (50)	7,856 (51)	.088
Region, No. (%)
Northeast	6,279 (21)	2,894 (20)	3,385 (23)	< .001
North central	8,233 (28)	4,164 (28)	4,069 (27)
South	12,176 (41)	6,405 (44)	5,771 (38)
West	3,006 (10)	1,175 (8)	1,831 (12)
Health plan type, No. (%)
HMO or capitated POS	3,484 (12)	1,697 (12)	1,787 (12)	< .001
Standard/lower deductible insurance	23,160 (80)	11,228 (79)	11,932 (82)
High-deductible insurance	2,186 (7.6)	1,292 (9.1)	894 (6.1)
Comorbidity index, No. (%)
0	8,502 (28)	3,782 (25)	4,720 (31)	< .001
1	3,778 (12)	1,978 (13)	1,800 (12)
2+	18,290 (60)	9,337 (62)	8,953 (58)
Concomitant bronchoscopic procedure, No. (%)	10,400 (34)	9,491 (63)	909 (5.9)	< .001
Lung cancer, No. (%)	13,170 (43)	6,959 (46)	6,211 (40)	< .001
Same-day hospitalization, No. (%)	2,105 (6.9)	216 (1.4)	1,889 (12)	< .001
Calendar year, No. (%)				< .001
2007-2009	9,267 (30)	2,218 (15)	7,049 (46)
2010-2012	11,746 (38)	6,073 (40)	5,673 (37)
2013-2015	9,557 (31)	6,806 (45)	2,751 (18)

EBUS = endobronchial ultrasound; HMO = health maintenance organization; IQR = interquartile range; POS = point of service.

Severe adverse events occurred in < 1% of patients (Table 2). Great vessel injury occurred more often after mediastinoscopy. Unadjusted rates of severe adverse events did not vary between EBUS and mediastinoscopy (0.3% vs 0.4%; P = .189), although the rate of vocal cord paralysis was lower with EBUS (1.4% vs 2.2%; P < .001). After adjustment for potential confounding factors (Table 3), EBUS was associated with a 58% lower risk of severe adverse events. Factors associated with a higher risk of a severe adverse event included health plans other than high-deductible plans, higher comorbidity index, concomitant bronchoscopy, and more recent calendar years. EBUS was also associated with a 43% lower adjusted risk of vocal cord paralysis. Factors associated with a higher risk of vocal cord paralysis included age, higher comorbidity index, male sex, and lung cancer diagnosis.

Table 2.

Unadjusted Risks and Costs Associated with Endobronchial Ultrasound and Mediastinoscopy

	EBUS (n = 15,097)		Mediastinoscopy (n = 15,473)		P Value
	No. of Events	% (95% CI)	No. of Events	% (95% CI)
Severe adverse events
Anya	49	0.3 (0.2-0.4)	65	0.4 (0.3-0.5)	.189
Pneumothorax	38	0.3 (0.2-0.3)	26	0.2 (0.1-0.2)	.133
Hemothorax	0	0 (0-0.02)	5	0.03 (0.01-0.08)	.065
Great vessel injury	3	0.02 (0.004-0.06)	17	0.1 (0.06-0.2)	.003
Airway injury	0	0 (0-0.02)	4	0.03 (0.007-0.07)	.125
Procedure-related death	8	0.05 (0.02-0.1)	14	0.09 (0.05-0.2)	.287
Vocal cord paralysis	214	1.4 (1.2-1.6)	345	2.2 (2.0-2.5)	< .001

Direct procedure-related expenditures, 2015 USD	95% CI	95% CI	95% CI	95% CI	P Value
Total costs
Median (IQR)	$5,409 ($3,242-$8,587)	$5,338-$5,489	$6,215 ($3,835-$10,094)	$6,119-$6,301	< .001
Mean (± SD)	$6,816 (± $6,836)	$6,707-$6,925	$9,023 (± $14,986)	$8,791-$9,263	< .001
Out-of-pocket costs
Median (IQR)	$150 ($0-$679)	$143-$164	$131 ($0-$642)	$118-$138	< .001
Mean (± SD)	$487 (± $841)	$474-$501	$462 (± $896)	$448-$476	.010

USD = US dollar. See Table 1 legend for expansion of other abbreviation.

^aA patient may have more than one adverse event.

Table 3.

Factors Associated With Procedure-Related Adverse Events

Factor	Severe Adverse Eventa OR (95% CI)	Vocal Cord Paralysis OR (95% CI)
Mediastinal evaluation method
Mediastinoscopy	Referent	Referent
EBUS	0.42 (0.32-0.55)	0.57 (0.54-0.60)
Age, y	1.02 (1.00-1.04)	1.02 (1.01-1.02)
Men	0.92 (0.55-1.56)	1.51 (1.45-1.57)
Region
Northeast	Referent	Referent
North central	0.75 (0.69-0.82)	0.86 (0.67-1.11)
South	1.18 (0.94-1.47)	0.92 (0.65-1.29)
West	1.4 (0.72-2.72)	0.86 (0.79-0.93)
Health plan type
HMO or capitated POS	Referent	Referent
Standard/lower deductible insurance	1.13 (1.07-1.18)	1.01 (0.98-1.04)
High-deductible insurance	0.57 (0.56-0.59)	0.73 (0.69-0.77)
Comorbidity index
0	Referent	Referent
1	2.16 (1.57-2.96)	1.26 (1.13-1.4)
2+	2.02 (1.87-2.17)	1.74 (1.27-2.39)
Concomitant bronchoscopic procedure	2.35 (2.06-2.68)	1.02 (0.95-1.1)
Lung cancer	0.61 (0.58-0.65)	1.28 (1.11-1.47)
Calendar year
2007-2009	Referent	Referent
2010-2012	1.35 (1.19-1.54)	0.97 (0.88-1.07)
2013-2015	1.44 (1.02-2.04)	1.06 (0.90-1.24)

See Table 1 legend for expansion of abbreviations.

^aSevere adverse events include pneumothorax, hemothorax, airway injury, great vessel injury, and death.

Unadjusted mean total costs were $2,211 (95% CI, –$1,951 to –$2,471) lower with EBUS compared with mediastinoscopy. After adjustment, costs associated with EBUS were $1,650 (95% CI, –$1,776 to –$1,525) lower. Concomitant bronchoscopic procedures and same-day hospitalizations were the two dominants factors that explained some of the cost differences between EBUS and mediastinoscopy (Table 4).

Table 4.

Potential Explanatory Factors for Differences in Cost Between Endobronchial Ultrasound and Mediastinoscopy

Model	Absolute Difference in Cost (95% CI)a: EBUS vs Mediastinoscopy
Unadjusted	–$2,211 (–$2,471 to –$1,951)
Add patient factorsb and year	–$2,639 (–$2,882 to –$2,396)
Add lung cancer diagnosis	–$2,638 (–$2,877 to –$2,398)
Add health plan type	–$2,651 (–$2,896 to –$2,405)
Add concomitant bronchoscopic procedures	–$3,465 (–$3,741 to –$3,188)
Add serious adverse eventc	–$3,373 (–$3,640 to –$3,105)
Add same-day hospitalization	–$1,650 (–$1,776 to –$1,525)

See Table 1 legend for expansion of abbreviation.

^aIn 2015 US dollars.

^bIncluding age, sex, region, and comorbidity index.

^cSevere adverse events include pneumothorax, hemothorax, airway injury, great vessel injury, and death.

All analyses were repeated in the subgroup of patients with a lung cancer diagnosis, and the findings were similar in magnitude and direction to the primary analysis (e-Table 3).

Discussion

We compared the risks and costs of EBUS with mediastinoscopy in a national cohort of patients undergoing mediastinal nodal evaluation for any indication. EBUS was associated with a lower risk of severe adverse events and vocal cord paralysis as well as costs when compared with mediastinoscopy.

The frequency of severe adverse events following EBUS and mediastinoscopy in this study were similar to prior investigations. EBUS complication rates ranged from 0.06% to 1.4%,^{17,23, 24, 25, 26} whereas mediastinoscopy complication rates ranged from 0% to 4.8%.^{6, 7, 8,27, 28, 29, 30, 31, 32} It is not possible to directly compare complication rates across studies because investigators used different definitions. The most likely reason our adverse event rates were lower than in prior studies is that our claims-based approach to measuring complications identified severe events—for instance, pneumothorax requiring a tube thoracostomy or hospitalization as opposed to an incidentally detected pneumothorax in an asymptomatic patient. A benefit of using claims to measure adverse events is that this approach is less susceptible to reporting bias because providers generally bill for procedures and hospitalizations in a fee-for-service health-care system. The comparable findings between our study and others suggest that EBUS is safer than mediastinoscopy.

Another important finding is the association between concomitant bronchoscopic procedures and a higher risk of adverse events. Others have noted a similar relationship using the American College of Chest Physicians Quality Improvement Registry, Evaluation, and Education (AQuIRE) database.¹⁷ That study showed that transbronchial biopsies occurred more often when rapid onsite cytologic evaluation (ROSE) was not available, suggesting that ROSE may curb the use of concomitant bronchoscopic procedures. For instance, a biopsy of the lung can be avoided in circumstances where ROSE facilitates the diagnosis of mediastinal nodal disease using EBUS.³³ Our database lacks sufficient granularity to determine the appropriateness of concomitant bronchoscopic procedures or the availability of ROSE. These findings suggest a potential opportunity to improve the safety of mediastinoscopy and EBUS (in particular) by avoiding unnecessary concomitant bronchoscopic procedures.

Vocal cord paralysis is the most common complication after mediastinoscopy and, to the best of our knowledge, has not been reported after EBUS. We evaluated this outcome for EBUS because we were referred a patient with a peripheral lung cancer tumor, without lymphadenopathy, and with new-onset vocal cord paralysis after EBUS via laryngeal mask airway (LMA). Reasons for vocal cord paralysis include direct, compression (mediastinal hematoma), and stretch injuries to the recurrent laryngeal nerve; direct injury to the vocal cord by the endotracheal tube (ETT) or bronchoscope in cases where an LMA was used; or rarely the LMA itself.³⁴ Other reasons for observing vocal cord paralysis in association with EBUS or mediastinoscopy include preexisting vocal cord paralysis or evolution in underlying patient disease over time leading to vocal cord dysfunction. With mediastinoscopy, a causal relationship likely underlies the observed association (ie, direct recurrent laryngeal nerve injury). The rate of vocal cord paralysis after mediastinoscopy in our study was higher than in single-institution studies,^7,27,29,32 but similar to that in ASTER (Assessment of Surgical Staging vs Endoscopic Ultrasound in Lung Cancer: a Randomized Clinical Trial).⁵ Given the potential for publication bias, it may be that vocal cord paralysis is an underreported, rare complication of EBUS. However, given the limitations of the database, we can only conclude that vocal cord paralysis is a finding in association with EBUS.

Use of claims to compare the costs of EBUS and mediastinoscopy is novel. Trial data have provided mixed results. ASTER authors curiously concluded that endosonography was less expensive than mediastinoscopy, even though differences in costs were statistically insignificant.^35,36 REMEDY (A Clinical Trial of Endobronchial Ultrasound for the Diagnosis of Mediastinal Lymphadenopathy)—a prospective single-arm study complemented with a modeled decision cost analysis—demonstrated that the mean cost of EBUS was $2,117 lower than that of mediastinoscopy—an estimate closest in magnitude to ours.³⁷ However, the Lung-BOOST (A New Pathway With Bronchoscopic or Oesophageal Ultrasound for Lung Cancer Diagnosis and Staging) trial—a multicenter, pragmatic, randomized controlled trial of EBUS vs mediastinoscopy among patients with suspected stage I-IIIA lung cancer—showed no differences in diagnostic and staging costs (or treatment-related costs in the subgroup of patients with confirmed lung cancer) despite the use of significantly fewer diagnostic tests and a shorter time-to-treatment decision.³⁸ Data from our study supplement trial findings by providing information about costs in the community at large.

Our cost analysis revealed two potentially modifiable drivers of cost—avoiding concomitant bronchoscopic procedures and hospitalization when possible. Others have shown that moderate sedation for EBUS (as opposed to monitored anesthesia care), performance of EBUS in the endoscopy suite rather than the operating room, and minimization of surgical supplies for mediastinoscopy as other strategies for cost reduction.^39,40 There are other potential ways to reduce costs. For instance, performing EBUS or mediastinoscopy on the same day as a pulmonary resection bundles the costs of both procedures, lowering costs from a payer perspective. Because of the limited granularity of MarketScan, we were unable to test whether these factors accounted for cost differences between EBUS and mediastinoscopy.

The single most important limitation of this study is that we were unable to directly compare the effectiveness of EBUS and mediastinoscopy (eg, diagnostic accuracy and yield). Although trial data suggest EBUS is superior to or at least no different than mediastinoscopy,^5,38 there is reason to believe that effectiveness may be different outside the trial setting. An American College of Surgeons study showed that only 47% of patients undergoing mediastinoscopy had lymph node material sent to pathology,⁴¹ although contemporary regional studies show a substantially lower 7% to 17% rate of mediastinoscopy without nodal sampling.^42,43 The AQuIRE database was used to evaluate the effectiveness of EBUS outside the trial setting, and it was found that diagnostic yield varied from 37% to 54% across hospitals.⁴⁴ There are also concerns that the thoroughness of mediastinal evaluation during lung cancer staging varies by diagnostic modality.⁴⁵ ASTER showed that a median of three mediastinal nodal stations were assessed in the endosonography group compared with a median of four mediastinal nodal stations in the mediastinoscopy group.⁵ A recent regional study showed a similarly differential and even lower median number of mediastinal lymph node stations sampled by mediastinoscopy and EBUS (two stations vs one station, respectively).⁴³ Differential effectiveness can adversely impact (1) safety and costs because of a need to perform additional invasive diagnostic procedures; (2) long-term patient outcomes as a result of delayed diagnosis or inappropriate treatment (resulting from inaccurate staging); and (3) patient satisfaction by burdening the patient with multiple interactions with the health-care system.

Our study has several other limitations. The findings may not be generalizable to elderly patients or those without insurance or with government insurance. Although we did not find a consistent relationship between age and risk, others have.¹⁷ It is likely that adverse events occur more frequently in the elderly population, but it is less certain that older age explains differences in outcomes between EBUS and mediastinoscopy. Another important limitation is that MarketScan does not include information on physician or hospital characteristics. Consequently, we could not explore whether provider characteristics explain differences in outcomes and costs, and we could not adjust for confounding by these variables resulting in potential bias. Unmeasured patient factors, such as prior mediastinoscopy, prior thoracic radiation, and fibrosing mediastinitis, may have confounded our analyses. We think this is unlikely because factors associated with poor outcomes would likely steer providers to preferentially perform EBUS; yet, EBUS was associated with better outcomes. Another limitation is that we were unable to identify diagnoses made by EBUS and mediastinoscopy, using claims data. ICD-9 diagnostic codes are notoriously inaccurate for diagnosis. ICD-9 codes for lung cancer are highly sensitive,⁴⁶ supporting our decision to perform a subgroup analysis for these patients. Finally, our analysis provides no information on costs along a longer time horizon.

Conclusions

We demonstrate an association between EBUS and greater safety and lower costs compared with mediastinoscopy when performed as standalone procedures for any indication. Our unique contribution to this body of evidence is (1) a data source less susceptible to reporting bias; (2) findings generalizable to a heterogeneous group of patients and providers outside the trial setting; and (3) a study with sufficient power to detect small but important differences in rare outcomes. Future studies should test the effectiveness of EBUS and mediastinoscopy in a similarly representative national cohort to determine whether there are trade-offs between these procedures in terms of effectiveness, safety, and costs.