Development and validation of risk models to select ever-smokers for CT lung-cancer screening

Abstract

Importance

The US Preventive Services Task Force (USPSTF) recommends computed-tomography (CT) lung-cancer screening for ever-smokers ages 55-80 years who smoked at least 30 pack-years with no more than 15 years since quitting. However, selecting ever-smokers for screening using individualized lung-cancer risk calculations may be more effective and efficient than current USPSTF recommendations.

Objective

Comparison of modeled outcomes from risk-based CT lung-screening strategies versus USPSTF recommendations.

Design/Setting/Participants

Empirical risk models for lung-cancer incidence and death in the absence of CT screening using data on ever-smokers from the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO; 1993-2009) control group. Covariates included age, education, sex, race, smoking intensity/duration/quit-years, Body Mass Index, family history of lung-cancer, and self-reported emphysema. Model validation in the chest radiography groups of the PLCO and the National Lung Screening Trial (NLST; 2002-2009), with additional validation of the death model in the National Health Interview Survey (NHIS; 1997-2001), a representative sample of the US. Models applied to US ever-smokers ages 50-80 (NHIS 2010-2012) to estimate outcomes of risk-based selection for CT lung-screening, assuming screening for all ever-smokers yields the percent changes in lung-cancer detection and death observed in the NLST.

Exposure

Annual CT lung-screening for 3 years.

Main Outcomes and Measures

Model validity: calibration (number of model-predicted cases divided by number of observed cases (Estimated/Observed)) and discrimination (Area-Under-Curve (AUC)). Modeled screening outcomes: estimated number of screen-avertable lung-cancer deaths, estimated screening effectiveness (number needed to screen (NNS) to prevent 1 lung-cancer death).

Results

Lung-cancer incidence and death risk models were well-calibrated in PLCO and NLST. The lung-cancer death model calibrated and discriminated well for US ever-smokers ages 50-80 (NHIS 1997-2001: Estimated/Observed=0.94, 95%CI=0.84-1.05; AUC=0.78, 95%CI=0.76-0.80). Under USPSTF recommendations, the models estimated 9.0 million US ever-smokers would qualify for lung-cancer screening and 46,488 (95%CI=43,924-49,053) lung-cancer deaths were estimated as screen-avertable over 5 years (estimated NNS=194, 95%CI=187-201). In contrast, risk-based selection screening the same number of ever-smokers (9.0 million) at highest 5-year lung-cancer risk (≥1.9%), was estimated to avert 20% more deaths (55,717; 95%CI=53,033-58,400) and was estimated to reduce the estimated NNS by 17% (NNS=162, 95%CI=157-166).

Conclusions and Relevance

Among a cohort of US ever-smokers age 50-80 years, application of a risk-based model for CT screening for lung cancer compared with a model based on USPSTF recommendations was estimated to be associated with a greater number of lung-cancer deaths prevented over 5 years along with a lower NNS to prevent 1 lung-cancer death.

INTRODUCTION

Lung-cancer is the most common cause of cancer death in the United States.¹ The National Lung Screening Trial (NLST) demonstrated a 20% reduction in lung-cancer mortality with 3 rounds of low-dose computed tomography (CT) screening as compared with chest radiography, over a mean follow-up of 6.4 years.² Consequently, the US Preventive Services Task Force (USPSTF) and the US Centers for Medicare and Medicaid Services (CMS) now recommend annual CT screening for a risk-factor-based subgroup of smokers—current and former smokers ages 55-80 years and 55-77 years, respectively, with at least 30 pack-years of smoking and, for former smokers, no more than 15 years since quitting.^3,4 These were largely based on the entry criteria for the NLST as well as microsimulation models that considered subgroups defined by age/pack-year/quit-year criteria.^5,6

Selecting individuals at highest lung-cancer risk, as determined by individual risk calculations (i.e. risk-based selection) rather than by risk-factor-based subgroups, might lead to more efficient screening.^3,7-10 In the NLST, 88% of CT-prevented lung-cancer deaths occurred in the 60% of participants at highest risk, while the 20% of participants at lowest risk accounted for only 1% of CT-prevented lung-cancer deaths.¹¹ The cost-effectiveness of CT screening also increased with lung-cancer risk in the NLST.¹² Risk-based selection more precisely delineates the benefits and harms of screening by accommodating detailed information on all lung-cancer risk factors.³ Risk-based selection also enforces consistency of screening recommendations by accommodating “equal management of people at equal risk”.¹³ However, to our knowledge, there are currently no risk tools for lung-cancer that have been validated in representative samples of the US population. Likewise, empirical evidence is lacking for the superiority of risk-based lung-cancer screening in the US.

In this study, we sought to develop and validate empirical lung-cancer incidence and death risk models generalizable to US smokers, as well as an empirical model for risk of false-positive CT screen. Models were applied to a contemporary cohort of US ever-smokers to investigate estimated outcomes from various risk-based selection strategies versus current USPSTF recommendations, for “NLST-like” screening (3 yearly CT screens) with 5-years follow-up.

METHODS

Data sources

Data was used from two lung-cancer screening trials in the US—The Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial and the NLST— as well as data from the NHIS, a representative sample of the US population. From 1993-2001, the PLCO trial randomized 154,901 US men and women ages 55-74 years to receive four annual posterior-anterior chest radiographs (three in never-smokers) or the standard of care, and concluded that chest radiography screening did not reduce lung-cancer mortality.¹⁴ The most recent follow-up data for PLCO was available through December 2009. From 2002-2004, the NLST randomized 53,454 US smokers ages 55-74, with at least 30 pack-years of smoking and no more than 15 years since smoking cessation to receive three annual rounds of low-dose CT or posterior-anterior chest radiography.² The NLST dataset included outcomes accrued through January 15, 2009, the latest date for censoring lung-cancer death for the primary analysis. The NHIS is an annual cross-sectional, multi-stage probability sample of approximately 87,500 individuals representing the non-institutionalized civilian US population.¹⁵ NHIS data collected through 2004 have been linked with the National Death Index (NDI), with follow-up through December 31, 2006.¹⁶ The National Institutes of Health Office of Human Subjects Research deemed this study exempt from IRB review.

Statistical Analyses

Development and validation of risk models

Absolute risk models were developed to predict five-year cumulative risk of lung-cancer incidence and lung-cancer death using data on ever-smokers within the control group of the PLCO trial. PLCO data allowed us to develop valid models for both USPSTF-eligible and –ineligible smokers. Cox hazard-ratio models on time since interview were used, and accounted for life expectancy by fitting a hazard-ratio model for competing causes of death.¹⁷ Compared to previous work¹¹, each submodel (lung-cancer incidence, lung-cancer death, and death by other causes) now includes more self-reported demographic (age, gender, race, education, body-mass index(BMI)), self-reported clinical (history of emphysema and lung-cancer family-history), and self-reported smoking (cigarettes per day, smoking duration, and smoking pack-years and quit-years) variables. Variables and parameterization for continuous variables were selected using the Akaike Information Criterion. See Supplemental Methods for details.

Validation of the lung-cancer death model for US ever-smokers ages 50-80 years used NHIS surveys 1997-2001 because age at smoking initiation was not systematically collected before 1997. Via NDI linkage through 2006, each NHIS participant had at least five years of follow-up for lung-cancer death. Multiple-imputation was used to account for the <2.5% missing information on BMI, race, education, or quit-years (Table S6). Because only 5.5% of participants reported family history of lung cancer, it can be conservatively assumed that those with missing family history information had no family history. For former smokers, the number of cigarettes smoked per day was systematically missing, for which a special imputation model was developed (Supplemental Methods).

Because there is no US-representative data including both lung-cancer incidence and risk factors, validation of the lung-cancer incidence model used data on ever-smokers in the chest radiography groups of the PLCO and the NLST. The PLCO chest radiography group experienced only slightly increased lung-cancer detection.¹⁴

Model validity was assessed by calibration (the ratio of number of model-predicted cases to the number of observed cases (Estimated/Observed)) and discrimination (the Area-Under-Curve (AUC) statistic) (Supplemental Methods).

A quadrinomial logistic regression model was developed to estimate an individual’s probability of 0, 1, 2, or 3 false-positive screens over 3 rounds of CT screening as a function of their modeled lung-cancer risk. The false-positive probability is the joint probability of having a positive CT screen and not being diagnosed with lung cancer within 1 year after diagnostic follow-up initiated within 1 year after the positive screen (see Supplemental Methods). This model for false-positive risk, based on the NLST data, was assumed to hold for lung-screening programs in the US.

All p-values are 2-sided and p<0.05 was declared significant.

US-representative estimates of CT lung-screening outcomes

The empirical models were applied to a contemporary US population of ever-smokers ages 50-80 years (NHIS 2010-2012), to estimate potential modeled outcomes associated with different strategies for selecting ever-smokers for NLST-like CT screening. 5-year risks of lung-cancer and lung-cancer death (in the absence and presence of NLST-like CT screening) and risk of false-positive CT were estimated for each NHIS participant. These estimates used a key assumption that NLST observations (i.e. 3 yearly CT screens would reduce lung-cancer mortality by 20.4% and increase lung-cancer detection by 12.4% over 5 years) are applicable to all US ever-smokers, regardless of their smoking history (Supplemental Methods).

For each selection strategy, the models were used to estimate numbers of smokers screened, lung-cancer deaths averted, lung-cancers detected, and false-positive CTs. Screening program metrics were estimated: “screening effectiveness” (defined as the number needed to screen (NNS) to prevent one lung-cancer death), “screening efficiency” (defined as the number of false-positive CTs per prevented lung-cancer death), and the number of extra lung-cancers diagnosed per prevented lung-cancer death.

Outcomes were compared based on USPSTF-eligibility versus different strategies for selecting individuals for CT screening based on lung-cancer risk thresholds. For illustrative purposes, two risk-based strategies were considered at length: (1) “fixed population-size”—choosing the lung-cancer risk threshold such that the number screened matches the number of USPSTF-eligible smokers in the US, and (2) “modeled fixed effectiveness”— choosing the lung-cancer risk threshold such that the NNS matches the NNS based on screening all USPSTF-eligible smokers in the US. All analyses were conducted in R¹⁸, and used the survey package¹⁹ to account for survey weights (see Supplemental Methods).

RESULTS

Risk models

Table 1 summarizes the data sources used to fit, validate, and apply each risk model. Tables 2 shows characteristics of all cohorts. Table 3 shows risk factors in the hazard-ratio submodels for lung-cancer incidence, lung-cancer death, and competing mortality. Predictors included age, race, gender, education, BMI, family history of lung cancer, self-reported emphysema, pack-years of smoking, duration of smoking, years since smoking cessation, and packs smoked per day (Table 3). Risk of false-positive CT screens increased with lung-cancer risk (Table S1).

Table 1. Schematic of data sources used to fit, validate, and apply the lung-cancer death, lung-cancer incidence, and false-positive CT screen models.

Model	Data source	Population
Lung Cancer Incidence Model
Model development	PLCO control group	Ever-smokers ages 55-74
Validation #1	PLCO chest radiography group	Ever-smokers ages 55-74
Validation #2	NLST chest radiography	NLST-eligible ever-smokers ages 55-74
Lung Cancer Death Model
Model development	PLCO control group	Ever-smokers ages 55-74
Validation #1	PLCO chest radiography group	Ever-smokers ages 55-74
Validation #2	NLST chest radiography group	NLST-eligible ever-smokers ages 55-74
Validation #3	NHIS 1997-2001	All US ever-smokers ages 50-80
False-Positive Lung Screen Model
Model development	NLST CT group	NLST-eligible ever-smokers ages 55-74
Application of all models to Projected outcomes from NLST-like CT screening	NHIS 2010-2012	All US ever-smokers ages 50-80

Abbreviations: PLCO= Prostate, Lung, Colorectal, and Ovarian cancer screening trial; NLST=National Lung Screening Trial; NHIS= National Health Interview Survey; CT= computed tomography ; US= United States

Note: The models are for lung -cancer incidence, lung-cancer death, and false-positive CT lung screen. In the presence of NLST-like CT screening, we assume that lung-cancer incidence risk is increased by 12.4%, and lung-cancer death risk is decreased by 20.4%, as observed in the NLST (see Methods and Supplemental Methods), and are applicable to all US ever-smokers, regardless of their smoking history. All models are applied to a contemporary cohort of US ever-smokers ages 50-80 from the NHIS 2010-2012 to estimate model-based outcomes from various strategies for selecting NLST-like CT lung screening (3 yearly CT lung screens, 5 years follow-up).

Table 2.

Characteristics for all cohorts utilized for model development, model validation, and model-based estimation of screening outcomes

		PLCO control		PLCO x-ray		NLST x-ray		NLST CT		NHIS 1997-2001		NHIS 2010-2012
	Categories	N	%	N	%	N	%	N	%	N	%	N	%
	Sample Size	39,180	100.00	39,822	100.00	26,554	100.00	26,604	100.00	29,091	100.00	18,643	100.00
Age	<50	0	0.0	1	0.0	2	0.0	0	0.0	0	0.0	0	0.0
	50-54	3588	9.2	3709	9.3	2685	10.1	2653	10.0	6722	23.1	3914	21.0
	55-59	12635	32.2	12856	32.3	10657	40.1	10679	40.1	5604	19.3	3767	20.2
	60-64	11288	28.8	11542	29	7380	27.8	7412	27.9	4734	16.3	3662	19.6
	65-69	8079	20.6	8230	20.7	4136	15.6	4181	15.7	4351	15.0	3075	16.5
	70-74	3588	9.2	3484	8.7	1693	6.4	1678	6.3	4150	14.3	2226	11.9
	75-80	2	0	0	0.0	1	0.0	0	0.0	3530	12.1	1999	10.7
Gender	Male	22694	57.9	23266	58.4	15664	59.0	15701	59.0	15375	52.9	9777	52.4
	Female	16486	42.1	16556	41.6	10890	41.0	10903	41.0	13716	47.1	8866	47.6
Race	White, Non-hispanic	34673	88.5	35188	88.4	23902	90.0	23920	89.9	22350	77.6	13165	71.8
	Black, Non-hispanic	2180	5.6	2239	5.6	1158	4.4	1169	4.4	3439	11.9	2797	15.2
	Hispanic	797	2	803	2	456	1.7	478	1.8	2540	8.8	1725	9.4
	Other	1530	3.9	1569	4.0	1017	3.8	1013	3.8	464	1.6	658	3.5
	Missing	19	0.0	23	0.1	21	0.1	24	0.1	298	1.0	0	0.0
Education	<12 grade	3417	8.7	3462	8.7	1597	6.0	1632	6.1	7825	27.0	3423	18.4
	HS graduate	8666	22.2	8798	22.1	6427	24.2	6258	23.5	8225	28.4	4905	26.3
	Post HS, no college	5375	13.8	5337	13.4	3694	13.9	3728	14.0	2204	7.6	745	4.0
	Some college	9107	23.3	9269	23.3	6075	22.9	6180	23.2	4867	16.8	3630	19.5
	Bachelor’s degree	6352	16.3	6597	16.6	4432	16.7	4498	16.9	3792	13.1	4359	23.4
	Postgraduate	6135	15.7	6300	15.8	3812	14.4	3776	14.2	2046	7.1	1581	8.5
	Missing	128	0.3	59	0.1	517	1.9	532	2.0	132	0.5	0	0.0
BMI	18.5 or less	280	0.7	296	0.8	231	0.9	227	0.9	552	1.9	317	1.7
	18.6-25	12085	31.4	12286	31.2	7303	27.6	7501	28.3	9982	35.1	5298	29.2
	25.1-30	16801	43.7	16983	43.2	11470	43.4	11252	42.4	11143	39.2	6773	37.3
	>30	9310	24.2	9779	24.9	7450	28.2	7535	28.4	6748	23.7	5783	31.8
	Missing	704	1.8	478	1.2	100	0.4	89	0.3	666	2.3	472	2.5
Pack-Yearsa	<10	6995	18	7225	18.3	0	0.0	0	0.0	1570	15.4	2473	26.1
	10-19.9	7185	18.5	7623	19.3	3	0.0	1	0.0	1625	15.9	1956	20.6
	20-29.9	5857	15.1	5750	14.5	679	2.6	660	2.5	1498	14.7	1404	14.8
	30-39.9	4894	12.6	5005	12.7	7218	27.2	7194	27.0	1774	17.4	1344	14.2
	40+	13961	35.9	13939	35.3	18654	70.2	18749	70.5	3737	36.6	2309	24.3
	Missing	288	0.7	280	0.7	0	0.0	0	0.0	18887	64.9	9157	49.1
Smoking Status	Current	7925	20.2	8022	20.1	12796	48.2	12789	48.1	10475	36.0	6647	35.7
	Former	31255	79.8	31800	79.9	13758	51.8	13815	51.9	18616	64.0	11996	64.3
Quit Years	<5	4185	13.4	4087	12.9	5604	41.2	5697	41.7	2641	14.2	1542	12.9
	5-9.9	3908	12.5	4077	12.8	3900	28.7	3832	28.0	2127	11.4	1131	9.5
	10-14.9	4361	14	4281	13.5	3951	29.0	3983	29.1	2605	14.0	1323	11.1
	15+	18801	60.2	19355	60.9	151	1.1	154	1.1	11239	60.4	7954	66.6
	Missing	0	0.0	0	0.0	152	1.1	149	1.1	4	0.0	46	0.4
Years Smoked	10 or less	5472	14	5650	14.3	1	0.0	1	0.0	3410	11.7	2707	14.6
	10.1-20	7682	19.7	7961	20.1	145	0.5	164	0.6	3774	13.0	2724	14.6
	20.1-30	8008	20.5	8068	20.4	2673	10.1	2725	10.2	5123	17.6	3102	16.7
	30.1-40	9473	24.3	9728	24.6	11567	43.6	11563	43.5	8175	28.1	5094	27.4
	>40	8338	21.4	8199	20.7	12168	45.8	12151	45.7	8605	29.6	4970	26.7
	Missing	207	0.5	216	0.5	0	0.0	0	0.0	4	0.0	46	0.2
Cigarettes per daya	<10	9938	25.4	10197	25.7	6	0.0	7	0.0	2260	22.1	2768	29.2
	10-19	14111	36.1	14674	36.9	12654	47.7	12594	47.3	2619	25.7	2682	28.3
	20-29	7916	20.2	7815	19.7	7216	27.2	7276	27.3	3516	34.5	2823	29.7
	30-39	4346	11.1	4328	10.9	4811	18.1	4814	18.1	878	8.6	564	5.9
	40-59	2255	5.8	2228	5.6	1666	6.3	1691	6.4	824	8.1	519	5.5
	60-79	403	1	423	1.1	161	0.6	188	0.7	88	0.9	111	1.2
	80+	125	0.3	89	0.2	40	0.2	34	0.1	19	0.2	23	0.2
	Missing	86	0.2	68	0.2	0	0.0	0	0.0	18887	64.9	9153	49.1
Emphysema	No	37172	95.7	37932	95.7	24447	92.3	24491	92.3	27379	94.1	17362	93.1
	Yes	1676	4.3	1715	4.3	2031	7.7	2054	7.7	1712	5.9	1281	6.9
	Missing	332	0.8	175	0.4	76	0.3	59	0.2	0	0.0	0	0.0
Family Historyb	None	34490	88.8	34978	88.6	22325	84.1	22348	84.0	28991	99.7	18214	97.7
	1	3963	10.2	4092	10.4	3929	14.8	3958	14.9	91	0.3	414	2.2
	2	406	1	420	1.1	300	1.1	298	1.1	9	0.0	15	0.1
	NA	321	0.8	332	0.8	0	0.0	0	0.0	0	0.0	0	0.0
Years of Follow-upc	Median, IQR	11.78	10.5,12.9	11.92	10.5,12.9	5.58	5.2,5.9	5.58	5.2,5.9	7.13	5.8,8.6	NA	NA
Lung cancersd	number, ratese	1505	35.4	1604	37.2	964	58.7	1083	65.9	NA		NA
Lung cancer deaths	number, ratese	1092	25.2	1120	25.4	442	30.9	354	24.6	684	31.6	NA

^aInformation on cigarettes per day is not collected for former smokers in the NHIS.

^bDefinition: No First Degree Relatives (FDRs) with lung cancer = 0, 1 FDR with lung cancer = 1, Two or more FDRs with lung cancer = 2.

^cYears of follow-up with cause of death data. For NLST, the median years of follow-up with lung cancer incidence data is 6.48 (IQR: 6.03,6.83) and 6.48 (IQR: 6.04,6.83) for the x-ray and CT groups respectively.

^dInformation on lung cancer incidence is not available on the NHIS 1997-2001. Neither incidence nor mortality follow-up data is available for NHIS 2010-2012.

^eRates are expressed as events per 10,000 person-years.

PLCO = Prostate Lung Colorectal Ovarian Cancer Screening Trial

NLST - National Lung Screening Trial

NHIS – National Health Interview Survey, restricted to smokers age 50-80

CT – computed tomography

Table 3. Cause-specific hazards models for prediction of lung-cancer death, lung-cancer incidence, and competing mortality, based on data from the control group of the PLCO Cancer Screening Trial.

Factor	Coding	Lung-Cancer Incidence		Lung-Cancer Death		Competing Mortality
		Hazard Ratio	95% CI	Hazard Ratio	95% CI	Hazard Ratio	95% CI
Age	Log term	80.388	(35.904,179.985)	431.812	(185.0591,1007.5777)	—	—
	Squared	—	—	— a	—	1.001	(1.001,1.001)
Gender (female)	Binary	0.923	(0.829,1.027)	0.837	(0.736,0.950)	0.566	(0.534,0.599)
Race	Categorical
	White, Non- Hispanic	1.000	Reference	1.000	Reference	1.000	Reference
	Black, Non- Hispanic	1.244	(1.006,1.537)	1.482	(1.176,1.869)	1.468	(1.330,1.621)
	Hispanic	0.648	(0.401,1.0467)	0.687	(0.397,1.190)	1.001	(0827,1.211)
	Asian or Other	0.673	(0.485,0.935)	0.657	(0.447,0.964)	0.849	(0.740,0.974)
Education b	Trend	0.931	(0.900,0.963)	0.908	(0.873,0.944)	0.957	(0.941,0.972)
BMI≤18.5 BMI	Binary	1.063	(0.682.1.657)	1.428	(0.899,2.268)	2.006	(1.628,2.472)
	Log term	0.485	(0.343,0.686)	0.447	(0.296,0.675)	—	—
	Squarede	—	—	—	—	1.004	(1.003,1.004)
Pack-years	Categorical
	0-29.9	1.000	Reference	1.000	Reference	1.000	Reference
	30-39.9	1.634	(1.328,2.011)	1.743	(1.354,2.244)	1.080	(0.985,1.185)
	40-49.9	1.755	(1.445,2.131)	2.112	(1.679,2.657)	1.131	(1.029,1.244)
	50+	2.046	(1.608,2.603)	2.446	(1.861,3.214)	1.206	(1.079,1.349)
Quit years	Log termf	0.726	(0.676,0.779)	0.686	(0.640,0.735)	0.830	(0.802,0.859)
Years smoked	Log term	—	—	1.395	(1.099,1.771)	—	—
	Linear	1.024	(1.014,1.034)	—	—	1.002	(0.998,1.006)
>1 pack/day	Binary	1.364	(1.154,1.613)	1.273	(1.054,1.539)	1.129	(1.043,1.221)
Emphysema	Binary	1.757	(1.503,2.054)	1.741	(1.450,2.090)	1.918	(1.754,2.096)
Lung-cancer family history c	Trend d	1.519	(1.326,1.741)	1.525	(1.300,1.789)	—	—

Abbreviations: BMI= Body mass index; CI= Confidence interval.

^a“—“ means the specified risk factor, or parameterization of the risk factor, was not included.

^b<12 grade=1, high-school graduate=2, post high-school but no college=3, some College=4, Bachelor’s degree=5, graduate school=6.

^cFDR = First-degree relatives (siblings, parents, children) with history of lung cancer.

^dDefinition: No FDRs with lung cancer = 0, 1 FDR with lung cancer = 1, Two or more FDRs with lung cancer = 2.

^eThis is the square of (BMI-25). BMI is modeled as a binary category for being underweight (BMI≤18.5) and continuously for BMI>18.5.

^fThis is natural logarithm of the sum of one and quit-years. All other log terms are natural logarithms.

Note: For lung cancer incidence and death models, an increase of 1 year higher age results in the hazards increasing by HR^{{log(age+1)-log(age)}}. If all other factors are the same, a 61 year old has 431.812^{{log(61)-log(60)}}=1.11 times greater hazards of lung cancer death and a 80.388^{{log(61)-log(60)}}=1.08 times greater hazards of lung cancer diagnosis than a 60 year old. For competing mortality models, an increase of 1 year higher age results in hazards increasing by HR^{{(age+1)^2-age^2}} If all other factors are the same, a 61 year old has 1.001^{{61^2-60^2}}=1.13 times greater hazards of death from other causes than a 60 year old.

The lung-cancer incidence model was validated in the chest radiography group of the NLST (Estimated/Observed=1.06, 95%CI=0.98-1.13; AUC=0.70, 95%CI=0.69-0.72), in the PLCO radiography group ever-smokers (Estimated/Observed=0.94, 95%CI=0.87-1.02; AUC=0.80, 95%CI=0.78-0.81), and within subgroups (Table S2).

The lung-cancer death model was validated for US ever-smokers ages 50-80 in the 1997-2001 NHIS, both overall (Estimated/Observed=0.94, 95%CI=0.84-1.05; AUC=0.78, 95%CI=0.76-0.80) and within subgroups (Table S3). The model was also validated in PLCO radiography group ever-smokers, both overall (Estimated/Observed=1.08, 95%CI=0.97-1.20; AUC=0.81, 95%CI=0.79-0.83) and within subgroups (Table S4).

Lung-cancer mortality in the NLST radiography group appears to be 24% lower than that expected from the lung-cancer death model that calibrated well to PLCO and to nationally-representative NHIS data. Although this does not affect internal validity of the NLST, modeling lung-cancer screening outcomes for the US was affected (See Supplemental Results, Table S4, and Discussion).

Modeled outcomes associated with 5-year effect of risk-based NLST-like CT screening in the US

Based on NHIS 2010-2012, there were an estimated 43.4 million ever-smokers ages 50-80 years in the US. Assuming that the NLST mortality reduction of 20.4% applies also to all ever-smokers independently of exposure level, an NLST-like CT screening program was modeled to prevent an estimated 82,245 lung-cancer deaths (95%CI=79,255-85,235) over 5 years (Table 4). Screening only the 9.0 million individuals (21%) eligible by USPSTF recommendations was modeled to prevent an estimated 46,488 (95%CI=43,924-49,053) lung-cancer deaths over 5 years (57% of estimated CT-preventable deaths). Instead, the risk-based fixed population-size strategy of screening the 9.0 million smokers ages 50-80 at highest 5-year risk of lung-cancer (≥1.9%) was modeled to prevent an estimated 55,717 (95%CI=53,033-58,400) lung-cancer deaths (68% of estimated CT-preventable deaths). This was a 20% relative increase in estimated CT-preventable deaths versus USPSTF recommendations (11% absolute increase; p<0.0001), yet screening the same number of smokers. Furthermore, compared to USPSTF recommendations, the risk-based fixed USPSTF sample-size strategy was modeled to have greater estimated screening effectiveness (NNS=194 (95%CI=187-201) vs. 162 (95%CI=157-166), p<0.0001) and estimated screening efficiency (fewer false-positive screens per prevented death: 133 (95%CI=128-137) vs. 116 (95%CI=113-119), p<0.0001), while maintaining the estimated ratio of extra lung-cancers diagnosed per prevented death (0.93 (95%CI=0.93-0.94) vs. 0.91 (95%CI=0.91-0.92), p<0.0001) (Table 4).

Table 4. Projected outcomes of an NLST-like CT lung screening program (3 yearly CT screens, 5-years follow-up) in the U.S., for different strategies for selecting smokers.

			Risk-based screening strategies
	All US ever- smokers ages 50-80a Estimate (95% CIs)	USPSTF- eligible smokersb Estimate (95% CIs)	Fixed- USPSTF population sizec Estimate (95% CIs)	Modeled Fixed-USPSTF effectivenessd Estimate (95% CIs)
Estimated number of eligible smokers	43,413,257 (42,289.285- 44,537,230)	9,018,130e (8,618,900- 9,417,361)	9,018,693 e (8,645,982- 9,391,405)	12,101,749 (11,605,039- 12,598,460)
Estimated 5-year Lung-cancer risk threshold	-	-	1.9%	1.7%
Estimated 5-year Lung-cancer death risk threshold	-	-	1.2%	0.9%
Estimated percentage of eligible smokers	100%	20.8% (20.0-21.5)	20.8% (20.1-21.5)	28% (26.9-28.8)
Total estimated number of lung-cancer deaths in the absence of screening	403,161 (388,505- 417,818)	227,891 (215,313- 240,457)	273,127 (259,966- 286,276)	305,808 (291,966- 319,588)
Estimated number of preventable lung-cancer deaths from CT screening	82,245 (79,255- 85,235)	46,488 (43,924- 49,053)	55,717 (53,033- 58,400)	62,382 (59,567- 65,196)
Estimated percentage of preventable lung-cancer deaths	100%	56.5% (55.0-58.1)	67.7% (66.6-68.9)	75.8% (74.9-76.8)
Estimated number needed to screen (NNS) to prevent 1 lung-cancer death	528 (513-543)	194 (187-201)	74 (157-166)	194 (188-200)
Estimated number of false-positive screens per prevented lung-cancer death	294 (287-302)	133 (128-137)	116 (113-119)	134 (131-138)
Estimated Number of extra diagnosed lung-cancers per prevented lung-cancer death	0.96 (0.95-0.96)	0.93 (0.93-0.94)	0.91 (0.91-0.92)	0.92 (0.92-0.93)
Estimated mean 5-year lung-cancer risk	1.50% (1.42-1.50)	3.9% (3.75-4.02)	4.5% (4.42-4.67)	3.8% (3.72-3.93)
Estimated mean 5-year lung-cancer death risk	0.9% (0.90-0.96)	2.5% (2.43-2.62)	3.0% (2.94-3.11)	2.5% (2.45-2.60)

Abbreviations: NLST= National Lung Screening Trial; CT= computed tomography; US= United States; USPSTF= United States Preventive Services Task Force.

^aEstimates for US ever-smokers age 50-80 years indicate weighted estimates from the National Health Interview Survey (NHIS) 2010-2012

^bWeighted estimated number of smokers age 55-80 years in the NHIS 2010-2012 who meet the USPSTF criteria for lung- cancer screening: at least 30 pack-years of smoking and no more than 15 years since quitting smoking.

^cThe fixed USPSTF population-size was selected by choosing the lung-cancer risk threshold such that the number screened matches the number of USPSTF-eligible smokers in the US.

^dThe fixed USPSTF effectiveness population-size was selected by choosing the lung-cancer risk threshold such that the NNS matches the NNS based on screening all USPSTF-eligible smokers in the US.

^eTotals are not exactly the same due to discreteness of weighted estimates of totals.

The risk-based modeled fixed effectiveness strategy maintains the same estimated NNS=194 as modeled for the USPSTF recommendations. This strategy would select for CT screening an extra 3.1 million individuals (12.1 million total; 28% of ever-smokers ages 50-80) at highest 5-year lung-cancer risk (≥1.7%), and was modeled to prevent an estimated 62,382 (95%CI=59,567-65,196) lung-cancer deaths (76% of CT-preventable deaths) over 5 years. Compared to the USPSTF guidelines, this was a 34% relative increase in modeled CT-preventable deaths (19% absolute increase; p<0.0001), while maintaining the same estimated screening efficiency (false-positive screens per prevented death: 134 (95%CI=131-138) vs. 133 (95%CI=128-137), p=0.5) and the same estimated ratio of extra lung-cancers diagnosed per prevented death (0.92 (95%CI=0.92-0.93) vs. 0.93 (95%CI=0.92-0.93), p=0.006) (Table 4).

Figure 1 shows modeled 5-year outcomes for risk-based CT screening strategies over a range of lung-cancer risk thresholds. For example, preventing 90% of CT-preventable lung-cancer deaths was estimated to require screening the 49% of ever-smokers at highest lung-cancer risk (≥0.7%; 21.2 million people), yielding an estimated NNS of 287 (95%CI=279-295) per prevented death and an estimated 185 (95%CI=181-190) false-positives per prevented death. Strategies below the curve, such as USPSTF and CMS recommendations, were estimated as having less screening effectiveness than risk-based strategies.

Figure 1. 5-year modeled outcomes from different risk-based CT lung-cancer screening strategies in US ever-smokers ages 50-80.

For example, a lung cancer risk threshold of 0.7% is estimated to screen 49% (21M) of ever-smokers ages 50-80, prevent 90% (74,021) of preventable deaths over 5 years, screen 287 people to prevent 1 death, result in 185 false-positive CT screens per prevented death, and diagnose 0.94 extra lung cancers per prevented death. The asterisks on the figure denote the data markers for current USPSTF and CMS recommendations but the only axes that apply to these two points are the estimated number and % preventable, and the estimated # and % screened. USPSTF recommendations are estimated to screen 9.0 million (21%) of ever-smokers age 50-80, might prevent 46,488 lung-cancer deaths over 5 years (57% of the preventable deaths), screen 194 people to prevent one death, result in 133 false-positive CT screens per prevented death, and diagnose 0.93 extra lung cancers per prevented death. CMS recommendations are estimated to screen 8.7 million (20%) of ever-smokers age 50-80, might prevent 41,559 lung-cancer deaths over 5 years (51% of the preventable deaths), screen 208 people to prevent one death, result in 142 false-positive CT screens per prevented death and diagnose 0.94 extra lung cancers per prevented death. Strategies below the curve, such as USPSTF and CMS recommendations, are estimating as having less screening effectiveness than risk-based strategies. USPSTF recommendations are estimated as having more screening effectiveness than CMS recommendations because CMS recommendations exclude older smokers (ages 78-80), who can have higher risks of lung cancer.

Comparison of USPSTF-eligible population to risk-based populations

Risk-based strategies retain the highest-risk USPSTF-eligible smokers but replace lower-risk USPTF-eligible smokers with higher-risk USPSTF-ineligible smokers. Compared to USPSTF-eligibility, risk-based screening strategies preferentially include more current smokers overall, more low-intensity long-term current-smokers, and more high-intensity former-smokers who have quit for more than 15 years (Table 5).

Table 5. Characteristics of US ever-smokers ages 50-80 years under different selecting criteria for CT lung-cancer screening.

	All U.S. smokers ages 50-80 years a N=43,413,257 %	USPSTF- eligible smokers b N=9,018,130 e %	Risk-based eligibility: fixed USPSTF population size c N=9,018,693 e %	Risk-based eligibility: modeled fixed USPSTF effectiveness d N=12,101,749 %
Age, Years
50-54	23.8	0.0	1.4	3.5
55-59	20.3	29.5	9.5	12.5
60-64	19.9	28.2	21.0	21.9
65-69	15.1	20.5	23.7	22.4
70-74	11.1	12.6	21.7	19.2
75-80	9.8	9.3	22.7	20.6
Gender
Male	55.0	61.4	60.8	59.8
Female	45.0	38.6	39.2	40.2
Race
White, Non-Hispanic	80.2	85.0	82.4	82.0
Black, Non-Hispanic	9.8	7.7	12.8	12.9
Hispanic	7.1	4.7	3.2	3.6
Asian or Other	2.8	2.6	1.6	1.6
Education
Less than high-school	15.9	21.3	27.1	25.6
High-school graduate	27.2	29.6	32.7	32.2
Post-high-school, no college	3.6	4.9	4.6	4.7
Some college	19.0	19.6	16.4	17.1
Bachelor’s degree	25.1	19.5	15.2	16.0
Graduate school	9.2	5.1	4.1	4.5
Body Mass Index, Kg/m2
<18	1.0	1.9	3.4	2.8
18-30	67.6	67.6	75.8	73.9
> 30	31.4	30.5	20.8	23.3
Emphysema
No	93.3	85.0	78.8	81.6
Yes	6.7	15.0	21.2	18.4
First-degree relatives with lung-cancer
None	91.7	90.4	88.0	88.3
1	8.1	9.3	11.4	11.1
2	0.3	0.3	0.6	0.6
Smoking status
Former smoker	66.5	48.2	42.6	44.9
Current smoker	33.5	51.8	57.4	55.1
Duration of smoking, Years
<15	22.1	0.0	0.1	0.3
15-29	24.1	3.0	4.0	6.2
30-44	38.3	52.9	37.1	43.8
45+	15.6	44.0	58.8	49.7
Time since quit, Years
< 5	41.9	68.8	70.2	67.4
5-9	6.0	13.0	7.0	6.9
10-15	10.1	18.3	9.2	9.3
>15	42.0	0.0	13.6	16.3
Pack-years of smoking, Years
<15	42.2	0.0	10.3	12.7
15-29	23.9	0.0	11.2	13.4
30-45	17.7	44.1	24.7	27.4
>45	16.2	55.9	53.8	46.5
Cigarettes per day f
<10	25.0	0.0	13.1	15.2
10-19	24.9	11.6	17.1	18.0
20-29	32.9	57.4	39.5	38.1
30-39	6.6	13.7	12.8	11.8
40-49	7.2	12.5	12.0	11.4
50+	3.5	4.9	5.6	5.5

Note: Percentages are column percentages for each factor.

Abbreviations: NLST= National Lung Screening Trial; CT= computed tomography; US= United States; USPSTF= United States Preventive Services Task Force.

^aEstimates for US ever-smokers age 50-80 years indicate weighted estimates from the National Health Interview Survey (NHIS) 2010-2012

^bWeighted estimated number of smokers age 55-80 years in the NHIS 2010-2012 who meet the USPSTF criteria for lung-cancer screening: at least 30 pack-years of smoking and no more than 15 years since quitting smoking.

^cThe fixed USPSTF population-size was selected by choosing the lung-cancer risk threshold such that the number screened matches the number of USPSTF-eligible smokers in the US.

^dThe modeled fixed USPSTF effectiveness population-size was selected by choosing the lung-cancer risk threshold such that the NNS matches the NNS based on screening all USPSTF-eligible smokers in the US.

^eTotals are not exactly the same due to discreteness of weighted estimates of totals.

^f20 cigarettes = 1 pack of cigarettes.

In the risk-based fixed population-size strategy, 36% of the USPSTF-eligible smokers are replaced by an equal number of USPSTF-ineligible smokers at much higher lung-cancer risk (average 5-year lung-cancer risk=1.3% vs. 3.2%) and lower NNS (647 vs. 226) (Table S5a). The replacements are preferentially current smokers, ages 65-80 years, African-Americans, less educated and lower BMI individuals, those with emphysema, and those with a family history of lung-cancer (Table S5a). The subgroup of replacements who smoked less than 30 pack-years tend to be current long-term (45+ year) smokers, but 99% of whom smoke less than 1 pack per day, and 61% of whom smoke less than half a pack per day (Table S5a). This subgroup is also majority female and disproportionately African-American. The subgroup of replacements who quit more than 15 years ago were high-intensity smokers, almost all of whom smoked at least 30 pack-years, and 53% smoked at least 45 pack-years (Table S5a).

Similar conclusions hold for the risk-based modeled fixed effectiveness strategy, which replaces 24% of the USPSTF-eligible with more USPSTF-ineligible smokers at higher lung-cancer risk (average 5-year lung-cancer risk=1.1% vs. 2.6%) and lower NNS (813 vs. 281) (Table S5b).

DISCUSSION

Empirical individual risk models were developed, validated, and applied to US health survey data to estimate the 5-year effect of NLST-like CT lung-cancer screening (3 annual screens) in the US. The risk models validate well in US research cohorts (PLCO and NLST) as well as in the US general population (NHIS), suggesting transportability of these models. The key observation from the models is that compared to selecting risk-factor-based subgroups for screening (such as current USPSTF recommendations), individual-risk-based selection of smokers was estimated to prevent more deaths, improve screening effectiveness (defined as the number needed to screen to prevent 1 lung-cancer death) and improve screening efficiency (defined as the ratio of false-positive CT screens to prevented deaths).

The superior performance of risk-based screening is highlighted by the estimate that 90% of CT-preventable lung-cancer deaths are possibly preventable by a risk-based strategy that screens only 49% of US ever-smokers ages 50-80. This strategy is modeled to screen 287 individuals per prevented death, which may be comparably effective as other cancer screening programs.²⁰

Risk-based screening strategies appear superior to USPSTF recommendations because they preferentially replace the 36% of the USPSTF-eligible who are low-risk, low-benefit, ever-smokers (5-year lung-cancer risk=1.3% and NNS=647) with USPSTF-ineligible high-risk, high-benefit, ever-smokers (5-year lung-cancer risk=3.2% and NNS=226). These USPSTF-ineligible high-risk individuals cannot be identified by subgroups and require a risk calculation to identify, such as the 22% who smoke less than 30 pack-years²¹, the 13% who smoke less than a half-pack per-day, or the 14% who quit smoking more than 15 years ago. Conversely, 36% of USPSTF-eligible individuals are actually at low risk and thus may benefit less from screening than would be recognized without a risk calculation. Risk-based selection would also increase the number of African-Americans and women selected for CT lung screening. Post-hoc NLST analyses suggest possibly higher efficacy of CT lung screening for women.²²

Substantially higher effectiveness of an NLST-like CT-screening program among USPSTF-eligible smokers in the US was estimated than that observed in the NLST² (NNS=194 vs. 320, respectively). Although including individuals ages 75-80 (per USPSTF recommendations) increases effectiveness, the bigger contributor is the substantially lower lung-cancer mortality in the NLST than expected based on lung-cancer mortality rates in the PLCO and the US (NHIS). Consequently, the benefits of CT-screening in the US could be higher than that observed in the NLST², or estimated using NLST rates²³, or estimated by microsimulation models calibrated to the NLST²⁴. The deficit in lung-cancer mortality in the NLST was not fully explained by study-specific differences in risk factors, lung-cancer incidence, treatments, or histology/stage at lung-cancer diagnosis. A small mortality reduction from chest radiography in the NLST, but not PLCO, cannot be ruled out. These observations demonstrate the importance of validating risk models to population-representative data, such as the NHIS, rather than to research studies, which can have healthy-volunteer effects.^25,26

Our empirical methodology has limitations. The estimates are model-based rather than directly observed outcomes. The estimates presume the implementation of screening programs in the US with short-term performance similar to the NLST. In particular, the key assumption was that that the 20.4% reduction in lung-cancer mortality and the 12.4% increase in lung-cancer detection from CT screening observed in the NLST would be the same in NLST-ineligible smokers in the US population. Notably, in the NLST, the mortality reduction and increased lung cancer detection were unrelated to modeled lung-cancer risk, lending support to the validity of the assumption. The NLST-based model for false-positive CT screens was assumed to apply to lung-screening programs in the US. Screening performance may change with innovation, for example, if CT findings are classified by Lung-RADS rather than NLST protocols.²⁷ Our work contrasts with microsimulation-based estimates^5,6 in that, to avoid extrapolation beyond the observed NLST follow-up, only the short-term effect of an NLST-like CT screening program in the contemporary US population was considered. Since the NHIS does not collect cancer incidence, the lung-cancer incidence model was validated only in research cohorts. In addition, there is no external data for validating the model for false-positive risk.

Risk-based selection for screening is justified only if the benefits and harms of screening primarily depend on individual cancer risk, in that, two individuals with different combinations of risk factors but equal modeled risk would have similar outcomes. If true, this implies the principle of “equal management of people at equal risk,” which provides a intellectual framework for the development of simplified and consistent recommendations for risk-based precision medicine.¹³ This principle was adopted as the basis of current risk-based cervical-cancer screening recommendations ^28-30 and underlies the official risk-based decision aid.³¹ For lung cancer, however, this principle may not hold for certain high-risk subgroups that may be at increased risk of procedure-related complications, such as individuals with chronic obstructive pulmonary disease. Indeed, NLST individuals with multiple pulmonary comorbidities did not benefit from CT screening, despite having a high modeled lung-cancer risk.¹¹

However, implementing risk-based screening in clinical practice poses many challenges. The models provide estimates that could be useful for justifying a cost-effective risk threshold to define screening eligibility. Risk thresholds could be based on either lung-cancer incidence^32-36 or mortality, which are highly correlated. Although our incidence model and the PLCO_m2012 incidence model³⁶ were fit to data from the PLCO, the models use different predictors and are of different forms (our model is a Cox hazard-ratio model accounting for competing mortality; PLCO_m2012 is a logistic-regression model). In the clinic, accurate and user-friendly risk-based decision aids^37-39 are required to lay the foundation for shared decision-making.⁴ Although risk communication is challenging, it is an evolving field of research.⁴⁰ Much research remains to be done to develop and evaluate shared decision-making processes to make precision prevention a reality.

Although CT screening can reduce lung-cancer mortality by approximately 20%, the majority of lung-cancer deaths are not screen-preventable at this time. The best way for smokers to avoid lung-cancer, and all smoking-related illness, remains to quit smoking as early as possible.

Conclusions

Among a cohort of US ever-smokers age 50-80 years, application of a risk-based model for CT screening for lung cancer compared with a model based on USPSTF recommendations was estimated to be associated with a greater number of lung cancer deaths prevented over 5 years along with a lower NNS to prevent 1 lung cancer death.

Key Points.

Question

What might be estimated outcomes from individual risk-based selection strategies, compared to USPSTF recommendations, for selecting ever-smokers for CT lung-cancer screening?

Findings

In this empirical modeling study within a cohort of US ever-smokers age 50-80 years, application of a risk-based model for CT screening for lung cancer compared with a model based on USPSTF recommendations was estimated to be associated with a greater number of lung-cancer deaths prevented over 5 years along with a lower NNS to prevent 1 lung-cancer death.

Meaning

Risk-based selection strategies for CT lung-cancer screening might improve screening effectiveness.