Rates of downstream procedures and complications associated with lung cancer screening in routine clinical practice: A retrospective cohort study

Abstract

Background:

Lung cancer screening (LCS) using low-dose computed tomography (LDCT) reduces lung cancer mortality but can lead to downstream procedures, complications, and other potential harms. Estimates of these events outside of the National Lung Screening Trial (NLST) have been variable and lacked evaluation by screening result that allows for more direct comparison to trials.

Objective:

Identify rates of downstream procedures and complications associated with LCS

Design:

Retrospective cohort study

Setting:

Five U.S. healthcare systems

Patients:

Individuals who completed a baseline LDCT for LCS from 2014–2018

Measurements:

Outcomes included downstream imaging, invasive diagnostic procedures, and procedural complications. For each, we calculated absolute rates overall and stratified by screening result and by lung cancer detection, and calculated positive (PPV) and negative predictive value (NPV).

Results:

Among the 9,266 screened patients, 1,472 (15.9%) had a positive baseline LDCT, of which 140 (9.5%) were diagnosed with lung cancer within 12 months [PPV=9.5% (95% CI: 8.0, 11.0); NPV=99.8% (95% CI: 99.7, 99.9); sensitivity=92.7% (95% CI: 88.6, 96.9); specificity=84.4% (95% CI: 83.7, 85.2)]. Absolute rates of downstream imaging and invasive procedures in screened patients were 31.9% and 2.8%, respectively. In patients undergoing invasive procedures after positive results, complications were substantially higher than those observed in NLST (30.6% vs 17.7% for any complication; 20.6% vs 9.4% for major complications).

Limitations:

Assessment of outcomes are retrospective and based on procedural coding.

Conclusions:

Our results indicate substantially higher rates of downstream procedures and complications associated with LCS in practice than those observed in NLST. Assessing and improving diagnostic management is likely needed to ensure screening benefits outweigh potential harms.

Primary Funding Sources:

National Cancer Institute; Gordon and Betty Moore Foundation

INTRODUCTION

Annual lung cancer screening (LCS) with low-dose CT (LDCT) in eligible adults reduces lung cancer mortality in randomized clinical trials and increases early detection in community practice (1–4). However, in addition to potential benefits, LCS can lead to potential harms, including receipt of false-positive results, downstream imaging with ionizing radiation, invasive diagnostic procedures, and complications stemming from invasive procedures (5–9). The magnitude of such potential harms in community practice remains uncertain, as does information about how these events may vary by screening result and lung cancer detection (10–12). Without evaluation of screening harms, the value of the benefits and overall effectiveness of LCS in clinical practice cannot be fully understood.

In the United States, annual LCS was first recommended by the U.S. Preventive Services Task Force (USPSTF) in 2013 for adults aged 55–80 who have smoked within the last 15 years and have at least a 30 pack-year smoking history (13). The USPSTF expanded guidelines in 2021 to lower age eligibility to age 50 and pack-years to 20 (14,15). Initial LCS guidelines were largely supported by results from the National Lung Screening Trial (NLST), a large-scale trial conducted in the United States from 2002–2009 that showed a 20% lung cancer-specific mortality benefit associated with LDCT (1,16). More recent trials have supported NLST findings, showing 16–20% lung cancer-specific mortality benefit (17,18). Trials have also observed potential harms associated with LDCT including a high false-positive rate (73.4% specificity after initial screening in NLST) and higher rates of downstream procedures and complications in those with positive screens (1,9,19). Although trial results have informed guidelines and practice, there is growing evidence that LCS outcomes outside trials may differ in part due to differences in the underlying screening population (e.g., higher comorbidities) and variations in implementation and follow-up (10,11,20).

As such, it is key to understand the occurrence of both benefits and risks related to LCS in clinical practice to not only inform and tailor guidelines but also to develop quality measures and interventions to improve screening benefits across health systems and patient populations (21,22). Although reports of potential patient harms outside of trials are beginning to emerge, previous studies assessing LCS-related harms have not reported results stratified by screening result or subsequent cancer detection (10,11,23,24). Furthermore, such reports have not accounted for key clinical characteristics (including smoking pack-years) to enable comparison of real-world populations against trial populations.

The objective of this study was to assess downstream imaging, invasive diagnostic procedures, and procedural complications associated with LCS in a longitudinal cohort of patients screened at one of five U.S. healthcare systems (20,25,26) to help advance the evidence needed to understand LCS-related harms in relationship to benefits.

METHODS

Study Population and Setting

This retrospective cohort study was conducted within the Population-based Research to Optimize the Screening Process (PROSPR) Lung Consortium, a National Cancer Institute (NCI)-funded collaboration comprising five diverse U.S. healthcare systems: Henry Ford Health (HFH); Kaiser Permanente Colorado (KPCO); Kaiser Permanente Hawaii (KPHI); Marshfield Clinic Health System (MCHS); and University of Pennsylvania Health System (UPHS) (25). The PROSPR Lung Consortium is focused on evaluating uptake, effectiveness, and equity of LCS in real-world practice. The Consortium uses a common data model that includes harmonized patient-level data extracted from each healthcare system’s administrative, electronic health record (EHR), and/or claims data. Data elements include patient demographics, tobacco use and social history, orders and referrals, procedures, diagnoses, cancer registry data, and area-level measures geocoded to the patient’s address. The KPCO Institutional Review Board (IRB) approved this study and waived informed consent due to minimal risks to the patients whose data were analyzed, with deferral from IRBs at other performance sites. In this manuscript, we followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.

Study Cohorts

To assess the occurrence of downstream procedures and complications following LCS, we first identified an overall cohort of patients aged 55–80 that completed at least one LDCT for LCS (LCS-LDCT) between January 1, 2014, through September 30, 2018 (“screened cohort”) and then identified a second nested cohort of screened patients who underwent an invasive procedure after positive baseline LCS-LDCT (“invasive procedures cohort”). All individuals had to have at least 12 months of continuous engagement and at least one completed healthcare visit before and after their screening index date. Continuous engagement was defined by either continuous enrollment in the site’s health plan (for membership-based systems) or empanelment as a primary care patient (for non-membership-based systems). Patients who died before the end of the continuous engagement period, but otherwise met study inclusion criteria, were retained in analysis to avoid bias. For all analyses, we used a population-based approach (i.e., no a priori sample size) and included all who met study eligibility.

Screening Cohort:

We used LCS-specific procedure codes to identify all LCS-LDCTs (Current Procedural Terminology (CPT) or Healthcare Common Procedure Coding System (HCPCS) codes G0297 and S8032) and categorized the first completed LCS-LDCT as the baseline LCS-LDCT for each screened patient (27). We excluded individuals with a history of lung cancer prior to screening index date, with missing comorbidity data, and who had never smoked (See Supplemental Figure 1 for full consort chart).

Invasive Procedures Cohort:

To assess complications associated with invasive diagnostic procedures after LCS, we identified patients within the screening cohort who underwent an invasive diagnostic procedure within 12 months of their positive baseline LCS-LDCT. We restricted to only those with positive baseline LCS-LDCTs to exclude complications that likely stemmed from diagnostic procedures unrelated to lung cancer evaluation.

Supplemental Matched Cohort:

To be able to compare our results to prior papers that reported incremental rates (11,23,24), we generated a cohort of patients (“controls”) who did not have the exposure of interest (screening or subsequent invasive procedure, as appropriate) and matched on key covariates to identify background rates of events in unexposed patients. We then calculated incremental rates for all outcomes by subtracting the observed rate in the controls from the observed rate in the exposed patients (11). Detailed methods and results of the matched control analysis are reported in the Supplemental Materials (see Supplemental Figure 2, Supplemental Table 4–6).

Data Elements

Cohort Characteristics.

We used EHR data from each healthcare system to assess patient age at screening index date, sex, race and ethnicity, most recent smoking status and max pack-years on or before the screening index date, comorbidities, and cancer diagnosis using tumor registry data collected at each system. We calculated Charlson Comorbidity Index based on diagnostic codes within the year prior to assigned screening index date (28). To identify screening results, we reviewed structured and unstructured data extracted from radiology reports to ascertain assigned Lung-RADS (classification system used to standardize LCS-LDCT results (29)) from each LCS-LDCT and classified the baseline screening result as negative (if Lung-RADS 1–2) or positive (if Lung-RADS 3–4) (30). For all data elements, we included the proportion of patients with missing data in their EHR. We included all data available from January 1, 2014, through September 30, 2019, which are the beginning and end dates of the current dataset.

Outcomes

Imaging and Diagnostic Procedures:

Drawing upon existing literature, clinical expertise, and chart review within our systems (1,23,23,24,31), we identified the relevant set of CPT and International Classification of Diseases (ICD)-9 and ICD-10 diagnostic and procedure codes to identify imaging and invasive procedures that occurred within 12 months of screening index date (see Supplemental Table 1 for a list of all included procedures and codes). We only included invasive procedures commonly used in the lung cancer diagnostic process (1,11,23,24). To compare with prior literature, we grouped imaging procedures as Chest CT (excluding LDCT), LCS-LDCT, and PET/Chest MRI (24). We grouped invasive procedures as needle biopsy, bronchoscopy, mediastinoscopy/mediastinotomy, thoracoscopy, thoracotomy, and other pleural procedures (pleural drainage, thoracentesis, and pleural biopsy).

Complications:

Drawing upon existing literature (23,24), clinical expertise, and chart review within our systems, we identified CPT and ICD-10 diagnostic codes associated with relevant complications that occurred within 30 days of the date of invasive procedure or matched index date (See Supplemental Table 2 for a list of all included complications and codes). Drawing upon NLST classifications (1), we grouped complications by severity as follows: minor (e.g., ileus, pneumothorax without chest tube), intermediate (e.g., respiratory distress, pleural effusion), and major (e.g., cardiac arrest, lung collapse, acute respiratory failure). We identified complication rates overall (assigning patients to the most severe complication group experienced based on NLST categories), and by procedure [grouped as needle biopsy, bronchoscopy, thoracic surgery (including mediastinoscopy, mediastinotomy, thoracoscopy, and thoracotomy), and other pleural procedures].

Statistical Methods

We used descriptive statistics to assess and report the absolute rates of imaging, invasive procedures, and complications for all screened patients. For all rates, we calculated 95% CIs using the Wald methods. We also calculated sensitivity (defined as the proportion of patients with confirmed lung cancer who were screen-positive), specificity (defined as the proportion of patients without confirmed lung cancer who were screen-negative), positive predictive value (defined as the proportion of screen-positive patients with confirmed lung cancer), and negative predictive value (defined as the proportion of screen-negative patients without confirmed lung cancer) for all screened patients at 12 months after baseline LCS-LDCT, and at 24 months in a subset of those screened from January 2014-September 2017 (to allow for 24 month follow-up time in our current dataset). For all absolute rates, we anchored on the patient’s baseline LCS-LDCT and counted patients only once for overall rates of procedures (e.g., any invasive procedure), but allowed an individual patient to be counted in multiple subcategories (e.g., needle biopsy and thoracotomy) if they had additional procedures within 12 months. We report absolute rates overall in addition to stratified by those with confirmed lung cancer within 12 months of their index date, and by result of baseline LDCT to evaluate diagnostic work-up in each group. We also descriptively compared observed rates in PROSPR patients to those reported from the NLST overall and by screening result. However, it should be noted that criteria for positive scans are not equivalent in NLST and Lung-RADS. Lung-RADS increased the diameter threshold for a positive baseline screen from 4-mm to 6-mm, which has been shown to decrease false positives but also sensitivity in contrast to NLST criteria (32). All analyses were performed using SAS (Version 9.4), and results reported using STROBE guidelines.

Role of the Funding Source

The funding source had no role in the design, conduct, or analysis of the study or the decision to submit the manuscript for publication.

RESULTS

Table 1 describes the characteristics of the 9,266 PROSPR patients who completed a baseline LCS-LDCT and the 180 screen-positive PROSPR patients with subsequent invasive procedures within 12 months of their baseline LCS-LDCT, as compared to those of NLST participants. In comparison to all screened patients, screen-positive patients with subsequent invasive procedures were slightly older (59.4% vs 52.0% aged 65 or older), reported higher rates of current smoking status at baseline screen (72.8% vs 55.2%) as well as higher pack-years (52.6% vs 32.8% 50+ pack-years) and more comorbidities (41.1% vs. 32.4% 2+ CCI).

Table 1 :

Characteristics of PROSPR Patients Who Underwent Lung Cancer Screening and Subsequent Invasive Procedures Compared with NLST Participants

	Screened PROSPR patients	Screen-Positive PROSPR patients with subsequent invasive proceduresa	NLST participants
	n (%)	n (%)	n (%)
Total patients	9,266 (100.0)	180 (100.0)	26,309 (100.0)
Year of Baseline LCS-LDCT
2014	615 (6.6)	3 (1.7)	-
2015	1,463 (15.8)	29 (16.1)	-
2016	2,245 (24.2)	39 (21.7)	-
2017	2,723 (29.4)	57 (31.7)	-
2018	2,220 (24.0)	52 (28.9)	-
Baseline Lung-RADS
Negative Screen
1	1,995 (21.5)	-	14,709 (55.6)
2	5,248 (56.6)	-	8,145 (30.8)
Positive Screen
3	906 (9.8)	24 (13.3)	1,697 (6.4)
4	31 (0.3)	13 (7.2)	-
4A	338 (3.6)	42 (23.3)	1,107 (4.2)
4B	153 (1.7)	72 (40.0)	358 (1.4)
4X	44 (0.5)	29 (16.1)	149 (0.6)
Missing/Unclassifiedb	551 (5.9)	-	290 (1.1)
Lung Cancer Diagnosis (12-months)	164 (1.8)	120 (66.7)	292 (1.1)
Positive Baseline Screen	140 (9.5)	120 (66.7)	270 (3.8)
Age
55–59	2,040 (22.0)	25 (13.9)	11,245 (42.7)
60–64	2,402 (25.9)	48 (26.7)	8,059 (30.6)
65–69	2,562 (27.6)	38 (21.1)	4,689 (17.8)
70–74	1,649 (17.8)	45 (25.0)	2,313 (8.8)
75–80	613 (6.6)	24 (13.3)	1 (<0.1)
Sex
Female	4,324 (46.7)	76 (42.2)	10,770 (40.9)
Male	4,942 (53.3)	104 (57.8)	15,539 (59.1)
Race
American Indian/Alaska Native	63 (0.7)	1 (0.6)	92 (0.3)
Asian	327 (3.5)	10 (5.6)	559 (2.1)
Black	1,260 (13.6)	27 (15.0)	1,195 (4.5)
White	6,805 (73.4)	132 (73.3)	24,289 (90.9)
Native Hawaiian/Other Pacific Islander	151 (1.6)	2 (1.1)	91 (0.3)
Otherb	199 (2.1)	2 (1.1)	-
Unknown	461 (5.0)	6 (3.3)	163 (0.6)
Hispanic Ethnicity	393 (4.2)	4 (2.2)	479 (1.8)
Smoking status
Currently Smokes	5,117 (55.2)	131 (72.8)	12,643 (48.1)
Formerly Smoked	4,149 (44.8)	49 (27.2)	13,666 (51.9)
Pack-Years (PY) c
<30 PY	725 (9.7)	8 (5.2)	6 (0.0)
30<35 PY	1,282 (17.2)	14 (9.1)	2,879 (10.9)
35<40 PY	794 (10.7)	10 (6.5)	3,844 (14.6)
40<45 PY	1,403 (18.9)	25 (16.2)	4,155 (15.8)
45<50 PY	791 (10.6)	16 (10.4)	2,930 (11.1)
50+ PY	2,443 (32.8)	81 (52.6)	12,495 (47.5)
Charlson Comorbidity Index
0	3,399 (36.7)	49 (27.2)	-
1	2,861 (30.9)	57 (31.7)	-
2+	3,006 (32.4)	74 (41.1)	-
Comorbid conditions
COPD	3,244 (35.0)	81 (45.0)	4,674 (17.5)
Stroke	147 (1.6)	3 (1.7)	753 (2.8)
Interstitial lung disease	13 (0.1)	0 (0.0)	70 (0.3)

Abbreviations: COPD = Chronic obstructive pulmonary disease; LCS-LDCT = Lung cancer screening using low-dose computed tomography; Lung-RADS = Lung Imaging Reporting and Data System; NLST = National Lung Screening Trial; PY = Pack-years; PROSPR = Population-based Research to Optimize the Screening Process.

Notes: All PROSPR patient data were assessed at baseline LDCT unless otherwise noted. Data on NLST participants were assessed at initial screen (Church et al. 2013; N=26,723) for all variables except for LungRADS (Pinsky et al. 2014; N=26,455), race (NLST 2011; N=26,722), and comorbidities (Aberle et al. 2010; N=26,723).

^aInvasive procedures include needle biopsy, bronchoscopy, mediastinoscopy/mediastinotomy, thoracoscopy, thoracotomy, and other pleural procedures.

^bOther race is the category documented in the EHR and therefore, no other information on specific race group is available.

^cWe report the highest number of pack-years documented in the patient’s chart on or before the time of baseline LDCT. Patients without documented pack-years were excluded from column percentages to allow comparison to NLST, but not from analysis. Missing for pack-years was 19.7% (N=1,828) for screened patients and 14.4% (N=26) for screen-positive patients with subsequent procedures. Pack-years are categorized by 5-year intervals to align with NLST reporting.

As indicated in Table 2, 1,472 (15.9%) of all screened patients had a positive baseline LCS-LDCT, of whom 140 [PPV=9.5% (95% CI: 8.0, 11.0)] had confirmed lung cancer within 12 months [sensitivity=92.7% (95% CI: 88.6, 96.9); specificity=84.4% (95% CI: 83.7, 85.2)]. For the subset of screened patients screened in 2014–2017 to allow for 24 month follow-up (N=5,593), 96 [PPV=10.2%; (95% CI: 8.3, 12.1)] of those with a positive baseline screen and 33 (0.7%) of those with a negative baseline screen were diagnosed with lung cancer within 24 months [sensitivity=74.4% (95% CI: 66.9, 81.9); specificity=84.5% (95% CI: 83.6, 85.5)]. Negative predictive values at 12 and 24 months were high [99.8% (95% CI: 99.7, 99.9) and 99.3% (95% CI: 99.0, 99.5) respectively].

Table 2.

Positive and Negative Predictive Values at 12 and 24 months after Baseline LCS-LDCT in Screened PROSPR Patients

Baseline Lung-RADSa	Confirmed Lung Cancer: Within 12 months of baseline LCS-LDCTb					Confirmed Lung Cancer: Within 24 months of baseline LCS-LDCTc
Positive Predictive Value	No	Yes	Total	PPV%	PPV Range (95% CI)	No	Yes	Total	PPV%	PPV Range (95% CI)
All Positive LCS-LDCTs	1,332	140	1,472	9.5	(8.0, 11.0)	846	96	942	10.2	(8.3, 12.1)
3	896	10	906	1.1	(0.4, 1.8)	585	16	601	2.6	(1.4, 4.0)
4	24	7	31	22.6	(7.9, 37.3)	19	6	25	24.0	(7.3, 40.7)
4A	306	32	338	9.5	(6.4, 12.6)	183	22	205	10.7	(6.5, 15.0)
4B	92	61	153	39.9	(32.1, 47.6)	56	39	95	41.1	(31.2, 50.9)
4X	14	30	44	68.2	(54.4, 81.9)	3	13	16	81.3	(62.1, 100.0)
Negative Predictive Value	No	Yes	Total	NPV%	NPV Range (95% CI)	No	Yes	Total	NPV%	NPV Range (95% CI)
All Negative LCS-LDCTs	7,232	11	7,243	99.8	(99.7, 99.9)	4,618	33	4,651	99.3	(99.0, 99.5)
1	1,993	2	1,995	99.9	(99.8, 100.0)	1,191	9	1,200	99.3	(98.8, 99.7)
2	5,239	9	5,248	99.8	(99.7, 99.9)	3,427	24	3,451	99.3	(99.0, 99.6)

Abbreviations: CI = Confidence Interval; LCS-LDCT = Lung cancer screening using low-dose computed tomography; Lung-RADS = Lung Imaging Reporting and Data System; NPV = Negative Predictive Value; PPV = Positive Predictive Value; PROSPR = Population-based Research to Optimize the Screening Process.

^aAnalysis excludes patients with missing LungRADS (n=551). This exclusion resulted in 13 confirmed cases being excluded from this analysis as we could not categorize the patient’s baseline scan as positive or negative.

^bIncludes patients with a baseline LCS-LDCT between 1/1/2014–9/30/2018, to allow for adequate follow-up time (12 months) in our dataset (1/1/2014–9/30/2019).

^cIncludes patients with a baseline LCS-LDCT between 1/1/2014–9/30/2017 to allow for adequate follow-up time (24 months) in our dataset (1/1/2014–9/30/2019). Due to the change in the denominator, the crude number of cancer cases is lower in 24 months analysis than in 12 months analysis. All individuals screened between 10/1/2017–9/30/2018 were excluded from two-year calculations because our dataset ends 9/30/2019 and thus we have less than 24 months of follow-up for these individuals. All individuals who disenrolled from the health system before 24 months from their baseline screen were also excluded from two-year calculations. As a result, the PPV values may be underestimates of the true PPV. Patients that died before 24 months were retained in analysis.

Imaging and procedure rates following screening:

31.9% (95% CI: 31.0–32.9; 2,956 patients) of screened PROSPR patients had downstream imaging and 2.8% (95% CI: 2.5–3.1; 260 patients) had invasive diagnostic procedures within 12 months of baseline LCS-LDCT (Table 3). Absolute imaging rates were 22.9% (95% CI: 22.1–23.8) for chest CT, 10.2% (95% CI: 9.6–10.9) for follow-up LDCT, and 4.5% (95% CI: 4.1–4.9) for PET/Chest MRI. Absolute rates for invasive follow-up procedures were 0.8% (95% CI: 0.6–1.0) for needle biopsy, 1.3% (95% CI: 1.1–1.6) for bronchoscopy, 0.6% (95% CI: 0.5–0.8) for mediastinoscopy or mediastinotomy, 1.0% (95% CI: 0.8–1.2) for thoracoscopy, 0.4% (95% CI: 0.3–0.6) for thoracotomy, and 0.2% (95% CI: 0.1–0.3) for other pleural procedures. Imaging and invasive procedures, for all but subsequent LCS-LDCTs, were much more common in those with confirmed lung cancer than those without. Similarly, imaging and invasive procedures were more common in those with positive rather than negative screening results (Table 3).

Table 3.

Absolute Rates of Downstream Imaging and Invasive Diagnostic Procedures in Screened PROSPR Patients

	Absolute Ratesa	Observed By Baseline Screening Resultsb		Observed By Confirmed Cancer Diagnosisc
	Observed (n=9,266)	Positive Screen (n=1,472)	Negative Screen (n=7,243)	Lung Cancer (n=164)	No Lung Cancer (n=9,102)
	% (95% CI)	% (95% CI)	% (95% CI)	% (95% CI)	% (95% CI)
Any Downstream Imaging	31.9 (31.0, 32.9)	85.5 (83.7, 87.3)	20.8 (19.9, 21.7)	100.0 (100.0, 100.0)	30.7 (29.8, 31.7)
Chest CT	22.9 (22.1, 23.8)	77.2 (75.0, 79.3)	11.1 (10.4, 11.8)	92.1 (87.9, 96.2)	21.7 (20.8, 22.5)
LCS-LDCT	10.2 (9.6, 10.9)	20.7 (18.6, 22.8)	10.5 (9.8, 11.2)	0.6 (0.0, 1.8)	10.4 (9.8, 11.0)
PET or Chest MRI	4.5 (4.1, 4.9)	9.8 (8.3, 11.3)	1.1 (0.9, 1.3)	93.3 (89.5, 97.1)	2.9 (2.6, 3.2)
Any Invasive Procedure	2.8 (2.5, 3.1)	12.2 (10.6, 13.9))	0.8 (0.6, 1.0)	86.0 (80.7, 91.3)	1.3 (1.1, 1.5))
Needle Biopsy	0.8 (0.6, 1.0)	4.2 (3.2, 5.2)	0.1 (0.0, 0.2)	28.7 (21.7, 35.6)	0.3 (0.2, 0.4)
Bronchoscopy	1.3 (1.1, 1.6)	5.6 (4.4, 6.7)	0.4 (0.3, 0.6)	37.2 (29.8, 44.6)	0.7 (0.5, 0.9)
Mediastinoscopy/Mediastinotomy	0.6 (0.5, 0.8)	3.2 (2.3, 4.1)	0.1 (0.0, 0.1)	29.3 (22.3, 36.2)	0.1 (0.0, 0.2)
Thoracoscopy	1.0 (0.8, 1.2)	4.9 (3.8, 6.0)	0.1 (0.1, 0.2)	40.9 (33.3, 48.4)	0.3 (0.2, 0.4)
Thoracotomy	0.4 (0.3, 0.6)	2.0 (1.3, 2.8)	0.1 (0.0, 0.1)	18.3 (12.4, 24.2)	0.1 (0.0, 0.1)
Other Pleural Proceduresd	0.2 (0.1, 0.3)	0.6 (0.2, 1.0)	0.1 (0.0, 0.2)	5.5 (2.0, 9.0)	0.1 (0.1, 0.2)

Abbreviations: CI = Confidence Interval; CT = Computed Tomography LCS-LDCT = Lung Cancer Screening using Low-Dose Computed Tomography; MRI = Magnetic Resonance Imaging; PET = Positron Emission Tomography; PROSPR = Population-based Research to Optimize the Screening Process.

^aThe absolute rate reports the observed rate in screened patients, unadjusted for background rates. Supplemental Table 5 reports incremental rate, adjusting for background rates in matched controls.

^bAbsolute rates in screened patients stratified by baseline LDCT results (Positive=LungRADS 3 or 4; Negative=LungRADS 1 or 2). Analysis excludes patients with missing LungRADS (n=551).

^cAbsolute rates in screened patients stratified by confirmed lung cancer diagnosis within 12-months of baseline LCS-LDCT.

^dLimited to other pleural procedures related to the diagnostic evaluation of lung cancer.

Absolute complication rates associated with invasive diagnostic procedures:

The observed complication rate within 30-days of an invasive diagnostic procedure in screen-positive patients was 30.6% (95% CI: 23.8–37.3; 55 of 180 patients) and differed by severity of complication with rates of 20.6% (95% CI: 14.7–26.5) for major, 8.3% (95% CI: 4.3–12.4) for intermediate, and 1.7% (95% CI: 0.0–3.5) for minor complications (Table 4). Among procedure types, complication rates were highest following thoracic surgery [33.0% (95% CI: 23.1–42.8)] and lowest for needle biopsy [25.7% (95% CI: 11.2–40.2)]. No screen-positive patients underwent other pleural procedures. In the subset of patients with confirmed lung cancer, we observed higher procedural complication rates overall [35.0% (95% CI: 26.5–43.5) vs 21.7% (11.2–32.1)] than in those patients without confirmed lung cancer (Table 4; Figure 2). We also observed variation in complication rates (in addition to imaging and procedure rates) by healthcare system (Supplemental Table 3).

Table 4.

Absolute Rates of Complications in Screen-Positive PROSPR Patients with Subsequent Invasive Diagnostic Procedures

	Overall Ratesa		Observed By Confirmed Cancer Diagnosisb
		Observed		Lung Cancer		No Lung Cancer
	n	% (95% CI)	n	% (95% CI)	n	% (95% CI)
Total patients	180		120		60
At least one complication	55	30.6 (23.8, 37.3)	42	35.0 (26.5, 43.5)	13	21.7 (11.2, 32.1)
Most severe classified as major	37	20.6 (14.7, 26.5)	31	25.8 (18.0, 33.7)	6	10.0 (2.4, 17.6)
Most severe classified as intermediate	15	8.3 (4.3, 12.4)	11	9.2 (4.0, 14.3)	4	6.7 (0.4, 13.0)
Most severe classified as minor	3	1.7 (0.0, 3.5)	0	0.0 (0.0, 0.0)	3	5.0 (0.0, 10.5)
Needle Biopsy
Total patients	35		19		16
At least one complication	9	25.7 (11.2, 40.2)	6	31.6 (10.7, 52.5)	3	18.8 (0.0, 37.9)
Most severe classified as major	6	17.1 (4.7, 29.6)	5	26.3 (6.5, 46.1)	1	6.3 (0.0, 18.1)
Most severe classified as intermediate	3	8.6 (0.0, 17.9)	1	5.3 (0.0, 15.3)	2	12.5 (0.0, 28.7)
Most severe classified as minor	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)
Bronchoscopy
Total patients	57		29		28
At least one complication	17	29.8 (18.0, 41.7)	11	37.9 (20.3, 55.6)	6	21.4 (6.2, 36.6)
Most severe classified as major	9	15.8 (6.3, 25.3)	7	24.1 (8.6, 39.7)	2	7.1 (0.0, 16.7)
Most severe classified as intermediate	6	10.5 (2.6, 18.5)	4	13.8 (1.2, 26.3)	2	7.1 (0.0, 16.7)
Most severe classified as minor	2	3.5 (0.0, 8.3)	0	0.0 (0.0, 0.0)	2	7.1 (0.0, 16.7)
Thoracic Surgery
Total patients	88		72		16
At least one complication	29	33.0 (23.1, 42.8)	25	34.7 (23.7, 45.7)	4	25.0 (2.8, 46.2)
Most severe classified as major	22	25.0 (16.0, 34.1)	19	26.4 (16.2, 36.6)	3	18.8 (0.0, 37.9)
Most severe classified as intermediate	6	6.8 (1.6, 12.1)	6	8.3 (2.0, 14.7)	0	0.0 (0.0, 0.0)
Most severe classified as minor	1	1.1 (0.0, 3.4)	0	0.0 (0.0, 0.0)	1	6.3 (0.0, 18.1)
Other Pleural Procedures c
Total patients	0		0		0
At least one complication	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)
Most severe classified as major	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)
Most severe classified as intermediate	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)
Most severe classified as minor	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)	0	0.0 (0.0, 0.0)

Abbreviations: CI = Confidence Interval; PROSPR = Population-based Research to Optimize the Screening Process.

^aThe absolute rate reports the observed rate in screen-positive patients with subsequent invasive procedures, unadjusted for background rates. Supplemental Table 6 reports incremental rate, adjusting for background rates in matched controls. Within each procedure group, patients who experienced more than one complication were classified by the most severe complication experienced as classified by NLST (minor, intermediate, major).

^bObserved rates in screened patients with subsequent invasive procedures stratified by confirmed lung cancer diagnosis within 12 months of baseline LDCT.

^cOther pleural procedures related to diagnostic evaluation of lung cancer.

Figure 2.

Rates of Complications after Subsequent Invasive Procedures in PROSPR Patients Compared to NLST

Abbreviations; LDCT= Low-Dose Computed Tomography; NLST = National Lung Screening Trial; PROSPR = Population-based Research to Optimize the Screening Process.

Comparison to NLST:

In comparison to NLST (Table 1), screened PROSPR patients overall were older (52.0% vs 26.7% aged 65 or older), more racially diverse (73.4% vs 90.9% White race) and had a higher proportion of females (46.7% vs 40.9%) (33). PROSPR patients also reported higher rates of current smoking status (55.2% vs 48.1%) and COPD (35.0% vs 17.5%) at baseline LCS-LDCT than observed in NLST. When directly comparing observed procedure rates in PROSPR patients with positive screens to those of NLST participants with positive screens (1,16), all but thoracotomy were markedly higher in PROSPR patients (Figure 1). Specifically, screen positive PROSPR patients had higher rates of needle biopsy (4.2% vs 2.2%), bronchoscopy (5.6% vs 4.3%), mediastinoscopy or mediastinotomy (3.2% vs 0.9%), and thoracoscopy (4.9% vs 1.2%) than equivalent NLST participants. In screen positive PROSPR patients who underwent subsequent invasive procedures, observed complication rates overall were nearly 2 times greater (30.6% vs 17.7%) and observed rates of complications classified as major were over 2 times greater (20.6% vs 9.4%) than in those observed in equivalent NLST participants (Figure 2)(1).

Figure 1.

Rates of Invasive Procedures in Screened PROSPR Patients Compared to NLST

Abbreviations: LDCT= Low-Dose Computed Tomography; NLST = National Lung Screening Trial; PROSPR = Population-based Research to Optimize the Screening Process.

DISCUSSION

To our knowledge, this is the first paper to report downstream imaging, procedures, and complications in a real-world cohort of patients screened for lung cancer overall and by baseline screening result and confirmed lung cancer. Importantly, this allowed for a more direct comparison to NLST because NLST reported downstream procedures and complications in patients only after a positive LDCT (1). In our study (Figure 1), the observed rates of invasive procedures in this cohort most equivalent to those reported in NLST (i.e., patients with positive LDCT) are markedly higher than rates reported in NLST—for all procedures except thoracotomy (2.0% PROSPR vs 2.8% NLST)(1,16), which may reflect changes in clinical practice. Prior papers have compared incremental rates (i.e., observed rate in those screened – observed rate in those not screened) to those of NLST, which is informative for practice, but not directly comparable stratification by screening results. While the inclusion of rates by screening results is a strength, differences between NLST and Lung-RADS criteria for a positive screen should be considered when interpreting our results. Given the decrease in size threshold (from 4 mm to 6 mm) in Lung-RADS, diagnostic evaluation and management of positive screens likely varied from that of NLST patients. However, given the widespread implementation of Lung-RADS in the U.S., it remains important to understand results by positive result to ensure patients are being optimally managed in practice.

The absolute rate of complications associated with invasive procedures after positive screening is high in our study (30.6%). Similar to other published reports, we observed higher rates of complications following surgical procedures in contrast to non-surgical procedures (10,11,23). In the NLST, the reported rate of complications after a positive screen (including those who did not undergo invasive procedures) was 1.4% overall; however, the complication rate in NLST limited to those undergoing invasive procedures after a positive screen was substantially higher (17.7%)(1,16). The observed rate of overall complications in our equivalent population was 1.7 times higher than observed in NLST, a difference that could shift the balance of screening benefits and harms in community practice. The observed rate of complications in our study might be driven by differences in the underlying patient population, as PROSPR patients were older, more likely to be currently smoking, and had a higher rate of underlying comorbidities at baseline screening than patients in NLST(33). It is important for future research to understand what factors may be driving any differences within and across healthcare systems and how they may contribute to overall harms and benefits from screening.

The rate of positive screens in this cohort is lower than those observed in the first round of NLST (15.9% vs 27.3%), which is likely attributed to differences between the definition of positive scans used by the NLST and those used by Lung-RADS (1,16). However, although lower than the NLST false-positive rate (26.6%), the false-positive rate (defined as 1-specificity) observed in our study remains high (15.6%) and is higher than was estimated in NLST using Lung-RADS (12.8%) (32). Furthermore, while lower than that of those with confirmed lung cancer, the observed procedural complication rate in patients without confirmed cancer is quite high (21.7%; 95% CI: 11.2–32.1) and higher than those with false-positive results reported in NLST (9.4%)(34). While the rate of invasive procedures should never be zero in patients without confirmed cancer because these procedures are necessary for diagnosis, the rates of procedures and resulting complications observed in PROSPR patients indicates that patients may be managed more aggressively than optimal in practice. While observed differences by healthcare system (Supplemental Table 3) suggests that implementation, reporting, and diagnostic evaluation might differ by system, these results should be interpreted cautiously given the wide and overlapping confidence intervals. However, these results indicate that future research is needed to understand how variations in diagnostic management might shape the benefits and harms of LCS.

Practice Implications:

The complication rate observed in NLST may often not be obtainable in settings outside of a highly controlled trial (10), and thus, monitoring and measuring complications in practice is essential to guideline development, quality measurement, and ensuring the balance of screening benefits and harms is maintained. Additionally, similar to those of other studies, our findings indicate that those patients screened in routine practice are different than participants in trials and are experiencing LCS-related harms at a greater rate. As such, incorporating life-years-gained or other tailored approaches that incorporate comorbidities assessment may be necessary to identify patients most likely to receive net benefits from lung cancer screening in practice (5,9,35).

Study limitations:

As noted in the methodssection, NLST did not use the same criteria for positive screens as Lung-RADS and thus cannot be directly compared to screen-positive patients in routine clinical practice. Due to challenges in identifying indication based on procedural codes alone, we were unable to identify if the indication for procedures and thus it is possible that some may have been for diagnostic evaluation of other lung diseases (e.g., inter) or for lung cancer treatment. Additionally, assessment of procedures and complications is retrospective, and could be underestimated due to coding practices within healthcare systems. Furthermore, while providing information regarding incremental rates is important to avoid overestimating harms associated with LCS (see Supplemental Materials), methods used cannot establish causal attribution to LCS. There are also additional potential screening harms such as cumulative radiation exposure, psychological effects, financial and opportunity costs, and incidental findings that we were unable to consider within the scope of the current study (7).

CONCLUSION

In conclusion, our study provides robust evidence on the rate of downstream imaging, procedures, and complications in patients screened for lung cancer at five diverse healthcare systems across the United States. We observed higher rates of both invasive procedures and complications than those observed in NLST, highlighting the need for practice-based strategies to assess variations in the quality of care and to prioritize LCS among those patients most likely to receive a net benefit from screening in relationship to potential complications and other harms.

ABBREVIATIONS

EHR

Electronic health record

LCS

Lung cancer screening

LDCT

Low-dose computed tomography

NLST

National Lung Screening Trial

PROSPR

Population-based Research to Optimize the Screening Process

USPSTF

U.S. Preventive Services Task Force