Exceptional Survival Among Kentucky Stage IV Non-small Cell Lung Cancer Patients: Appalachian Versus Non-Appalachian Populations

Abstract

Purpose:

To determine differences in exceptional survival (ES)—survival of 5 years or more past diagnosis—between stage IV non-small cell lung cancer (NSCLC) patients residing in the Appalachian versus non-Appalachian regions of Kentucky.

Methods:

This was a population-based, retrospective case-control study of Kentucky patients, diagnosed with stage IV NSCLC between January 1, 2000, and December 31, 2011. The data were drawn from the Kentucky Cancer Registry.

Findings:

Findings from the multivariable logistic regression revealed no significant differences in the odds of ES between patients who resided in Appalachian versus non-Appalachian Kentucky. Being female and undergoing surgery only as the first course of treatment were associated with higher odds of ES. Increasing age, unspecified histology, having poorly differentiated or undifferentiated carcinomas, and receiving radiation therapy only as the first course of treatment were associated with decreased odds of ES.

Conclusion:

Differences in the odds of ES among stage IV NSCLC patients were not related to residence in Appalachian versus non-Appalachian Kentucky. ES was associated with other nongenetic and treatment factors that warrant further investigations.

Lung cancer continues to be the leading cause of cancer deaths in the United States.¹ According to the US National Cancer Institute, the majority of lung cancer patients are diagnosed at a late stage and have poor prognosis.² In the most common type of lung cancer, non-small cell lung cancer (NSCLC), the median survival time for patients with stage IV disease is less than 1 year³ and the 5-year survival rate is very low, less than 5%.⁴ Previous research has demonstrated that a patient’s age,^5,6 gender,⁷ stage of diagnosis,^5,8 tumor morphology^9,10 and other biological markers,^11,12 overall pulmonary function,⁶ performance status¹³ (an estimate of patients’ ability to perform activities of daily living without help¹⁴), treatment,^6,13,15-19 and presence of other tumors¹⁰ are all associated with lung cancer survival. While many of these studies have focused on assessing factors associated with overall survival, few studies have examined exceptionally long-term survival in lung cancer patients,^18,20,21 especially with stage IV NSCLC.

Appalachian Kentucky and Lung Cancer

In 2017, lung cancer incidence and mortality rates per 100,000 people were notably higher in Kentucky than in the US overall (87.2 vs 55.0 and 55.5 vs 37.0, respectively).²² Within the state, rates differ with respect to the region of residence: Appalachian versus non-Appalachian Kentucky. According to authorizing legislation of the Appalachian Regional Commission (ARC), 54 out of 120 counties in Kentucky belong to Central Appalachia.²³ This area is 1 of the 5 subregions of the entire Appalachian region that encompasses a large territory along the Appalachian Mountains.²³ Most of Appalachian Kentucky counties are nonmetropolitan (50 counties or 93%), half of which are extremely rural.²⁴

In 2017, age-adjusted all-cancer incidence and mortality rates for lung cancer per 100,000 people were significantly higher in 54 Appalachian than in 66 non-Appalachian Kentucky counties: 105.2 versus 80.2 and 74.6 versus 48.0, respectively²⁵ (Figure 1). Researchers have linked increased rates for lung cancer in the Appalachian region to environmental contaminants associated with coal mining^26,27; higher than national levels of exposure to arsenic, chromium, and nickel²⁸; and to highly prevalent tobacco use.²⁹ Genetic predisposition has also been discussed in the literature.³⁰ The region’s poverty levels,³¹ poor access to health care,³² significant educational disparities,³³ as well as health inequalities,³³ including higher rates of morbidity and mortality for chronic diseases and injury,³³ make the burden of lung cancer in Appalachian Kentucky even greater.

Figure 1.

Age-Adjusted Cancer Incidence (A) and Mortality (B) in Kentucky. All Rates per 100,000. Based on Data Released November 2019 (A) and May 2020 (B).

The Present Study

The focus of this study is exceptional survival (ES) in stage IV NSCLC patients who resided in Kentucky. Considering the rarity of a 5-year survival in stage IV NSCLC, we defined ES as having lived at least 5 years past diagnosis (ie, 60 months or more). To our knowledge, there is no generally agreed definition of ES in lung cancer research. It is more common to see the term “prolonged” or “long-term” survival when describing patients with advanced lung cancer who survived at least 18 months.²⁰ Shahidi and Kvale¹⁸ as well as Julien and associates²⁰ used the term “exceptional” to describe survival in NSCLC patients who lived 10 years or more after diagnosis; whereas Frenkel and colleagues labeled cancer patients as “exceptional” if their survival, including cancer of lung, ranged from 4 to 23 years.³⁴

Given how disproportionately Appalachian Kentucky is affected by lung cancer and other chronic diseases, our primary aim was to determine differences in ES between stage IV NSCLC patients residing in the Appalachian versus non-Appalachian regions of the state. We hypothesized that Appalachian Kentucky would have a significantly lower proportion of exceptional survivors among stage IV NSCLC patients. Because ES in stage IV NSCLC patients is underinvestigated, we also aimed to explore its associations with demographic characteristics, biological and treatment factors.

Methods

Data Sources and Study Design

This was a population-based, retrospective case-control study, and the data were drawn from the Kentucky Cancer Registry (KCR). The KCR, which is both a National Cancer Institute (NCI) Surveillance Epidemiology and End Results (SEER) cancer registry and a Centers for Disease Control and Prevention (CDC) National Program of Cancer Registries (NPCR) member, tracks data on all cancer cases in Kentucky.^35,36 KCR data are linked with the National Death Index (NDI³⁷) and state death certificate data.³⁶ The Institutional Review Board (IRB) at the University of Kentucky approved the study protocol (# 44081). We requested the data in December 2018; therefore, given the availability of the data submitted to KCR, the cutoff study date was the end of 2016. Since we defined ES as having lived at least 5 years past diagnosis, the last year of diagnosis for eligible lung cancer patients (with the minimum follow-up time of 5 years) would have been 2011. The inclusion criteria for the study cohort were the following: (1) patients living in Kentucky, (2) aged 20 years and older, and (3) diagnosed with stage IV NSCLC between January 1, 2000, and December 31, 2011.

Initially, we included 14,673 primary invasive stage IV NSCLC records (not limited to first primary only) that were not identified through autopsy/death certification. After exclusion of 45 duplicate records with a later sequence number of lung cancers, there were 361 records of ES (ES cases); the remaining 14,267 records were classified as non-ES cases. Among the non-ES cases, we excluded 9 patients who were lost to follow-up, as their survival time was unknown. Assessing survival time of these patients was important for determining the control group. The controls should have had a notably shorter survival to minimize bias in comparison with exceptional survivors. We excluded patients from the non-ES group who demonstrated rather extended survival, though not long enough to qualify for ES. Therefore, the control group included patients who survived up to 18 months. The cutoff point of 18 months was selected because (1) patients should have had enough time to demonstrate a response to treatment and (2) our data showed that almost 90% of the non-ES cases were deceased after 18 months from their diagnosis, hence the control group would include all true short-term survivors.

Furthermore, we excluded 32 patients with a zero survival time, who were likely diagnosed with stage IV NSCLC at the time of a surgery, and 31 patients who had received an unknown therapy to treat their lung cancer. The final sample included 13,099 Kentucky patients, diagnosed with stage IV NSCLC between January 1, 2000, and December 31, 2011 (Figure 2).

Figure 2.

Flow Diagram of Selection of the Final Sample.

Examined Characteristics

The KCR database³⁶ provided data on age (in years), sex at birth, race, ethnicity, county of residence where initial diagnosis was made, tobacco use, type of insurance, histology of cancer, grade of tumor, lung cancer sequence number, first course of treatment received, and information on active patient follow-up for vital status. KCR assigns rural/urban status to counties based on the Rural-Urban Continuum Codes,³⁸ with codes of 8-9 assigned to counties located in extremely rural areas, as defined by the US Department of Agriculture’s Economic Research Service.²⁴ The main exposure of interest was Appalachian versus non-Appalachian Kentucky operationalized by the county of residence at diagnosis. KCR assigns Appalachian status to 54 Kentucky counties according to the geographic classification provided by the ARC.²³ Histology was categorized in the following way: adenocarcinoma: 8,140, 8,550, 8,255, 8,251, 8,250, 8,252, 8,253, 8,254, 8,310, 8,470, 8,480, 8,481, 8,260, 8,490, 8,230, 8,333; large cell: 8,012, 8,013, 8,014, 8,082, 8,123, 8,310; sarcomatoid cell: 8,033, 8,031, 8,022, 8,972, 8,032; squamous cell: 8,052,8,070, 8,071, 8,072, 8,073, 8,083, 8,084; non-small-cell carcinoma not otherwise specified (NSC NOS): 8,046, 8,000, 8,001, 8,010, 8,020, 8,021. Furthermore, we identified 13 different types of first course of treatment received. To assess associations of ES with individual therapies and combination therapy, we collapsed treatments in the 4 main categories: (1) chemotherapy only, (2) radiation therapy only, (3) surgery at primary site only, and (4) polytherapy (any combination of chemotherapy and/or radiotherapy and/or surgery at primary site and/or other nonsystemic therapy). For descriptive purposes, we also included the group of patients who had no definite therapy or underwent surgery at regional and/or distant sites only.

Charlson Comorbidity Index: Subset Analysis

Previous studies demonstrated associations between the Charlson Comorbidity Index (CCI)³⁹ and survival in NSCLC patients who underwent surgery.^40,41 In CCI, the higher the score the higher the risk of health care use and/or mortality the person would have; 0 means no comorbidities.³⁹ We treated CCI as a proxy for general health status. To control for a strong potential confounding factor—the overall patient’s health—we conducted an additional analysis for a subpopulation of patients diagnosed between 2007 and 2011, for whom CCI data were available. The KCR obtained information regarding the CCI by linking data to insurance claims databases, including private insurance, KY government employee insurance, Medicare claims, and Medicaid claims data. The comorbidity information was captured from claims 1 year prior to a patient’s cancer diagnosis. In the regression modeling, the CCI variable was treated as a 3-level categorical covariate (1: CCI = 0; 2: CCI of 1 or 2; and 3: CCI of 3 or higher).

Statistical Analysis

All analyses were conducted using SAS software, version 9.4 (SAS Institute, Cary, NC). To determine statistically significant differences in proportions of demographic characteristics and treatment factors between the case and control groups as well as between the Appalachian versus non-Appalachian groups, Pearson’s χ² or Fisher’s Exact Test (if the expected cell was less than or equal to 5) was used for categorical variables, and the Wilcoxon-Mann-Whitney test was used for nonnormally distributed, continuous variables. In unadjusted and adjusted analyses (single variable and multivariable logistic regression models, respectively), to assess associations between ES and the first course of treatment, we excluded patients without a definite therapy and those who underwent surgery at regional and/or distant sites only. Residence in the Appalachian region, histology, grade of tumor, and treatment were the variables of interest. Backward elimination with a significance level of 5%, while keeping the Appalachian status variable at each step, showed that among other potential confounding factors such as age, sex at birth, race, ethnicity, tobacco use, and insurance type, the first 2 were significant factors and were further included in the final model. We then carried out a second model selection with backward elimination including the same main effects together with 4 interactions of interest determined based on the previous literature: residence in the Appalachian region and smoking (1), residence in the Appalachian region and histology (2), age at diagnosis and treatment (3), as well as smoking and treatment (4). None of the interactions were significant; we excluded them from further analysis.

Sensitivity Analyses

The sample was not limited to primary lung cancers only as we aimed to examine the overall cohort of stage IV NSCLC patients. However, to explore the impact of having other cancers, we conducted 2 sensitivity analyses limiting the sample to patients whose lung cancer was the first primary (1) and patients who had lung cancer only (2).

Results

Descriptive Statistics

Table 1 presents summary statistics of descriptive analysis. There were 361 (2.8%) patients who demonstrated ES. No significant differences were found between the proportions of patients who demonstrated ES among residents of the Appalachian versus non-Appalachian region (P = .2938; 107 [2.5%] vs 254 [2.9%]). Overall, the survival time for the total study sample (n = 13,099) ranged from 0.03 to 220.6 months. Survival time of those who were residents of the Appalachian region ranged from 0.03 to 208.4 months, whereas for the residents of the non-Appalachian region the survival time ranged from 0.03 to 220.6 months. Patients who resided in the Appalachian region (4,216, 32.2%) were predominantly of male sex (63.4%), White and non-Hispanic (98.4% and 98.0%, respectively), and they were diagnosed at a slightly younger age than their counterparts from the non-Appalachian region (median age of 67 vs 68, P < .0001). Our sample included 1,458 (11.1%) patients who resided in completely rural areas. Among them, 1,065 (73.0%) were from Appalachian Kentucky (results not shown in the table). Compared to residents from the non-Appalachian region, a greater proportion of those from Appalachian Kentucky had Medicare, Medicaid, or were medically uninsured (P < .0001: 63.4% vs 62.2%, 11.7% vs 5.8%, and 4.5% vs 4.4%, respectively); and fewer had private insurance (7.0% vs 14.2%, P < .0001).

Table 1.

Characteristics of the Study Sample

Characteristic	All (n = 13,099)	Appalachia (n = 4,216)	Non-Appalachia (n = 8,883)	P	Controls (n = 12,738)	Cases (n = 361)	P
Age at diagnosis (in years)	68 (60, 76)	67 (59, 74)	68 (60, 76)	<.0001	68 (60, 76)	63 (56, 70)	<.0001
Sex at birth (female)	5,080 (38.8)	1,545 (36.6)	3,535 (39.8)	.0005	4,891 (38.4)	189 (52.4)	<.0001
Race				<.0001			.3781
Black	866 (6.6)	63 (1.5)	803 (9.0)		841 (6.6)	25 (6.9)
White	12,201 (93.1)	4,147 (98.4)	8,054 (90.7)		11,868 (93.2)	333 (92.2)
Other	29 (0.2)	6 (0.1)	23 (0.3)		27 (0.2)	2 (0.6)
Missing	3 (0.0)	0 (0.0)	3 (0.03)		2 (0.0)	1 (0.3)
Ethnicity				.0002			.0863
Non-Hispanic	12,877 (98.3)	4,131 (98.0)	8,746 (98.5)		12,523 (98.3)	354 (98.1)
Hispanic	52 (0.4)	9 (0.2)	43 (0.5)		48 (0.4)	4 (1.1)
Spanish surname only or unknown	170 (1.3)	76 (1.8)	94 (1.1)		167 (1.3)	3 (0.8)
Tobacco use				.0017			.4884
Never used	819 (6.3)	260 (6.2)	559 (6.3)		791 (6.2)	28 (7.8)
Users of different tobacco products	10,096 (77.1)	3,182 (75.5)	6,914 (77.8)		9,822 (77.1)	274 (75.9)
Not recorded/unknown	2,184 (16.7)	774 (18.4)	1,410 (15.9)		2,125 (16.7)	59 (16.3)
Insurance type				<.0001			<.0001
Not insured or self-pay	581 (4.4)	191 (4.5)	390 (4.4)		568 (4.5)	13 (3.6)
Private insurance	1,558 (11.9)	295 (7.0)	1,263 (14.2)		1,485 (11.7)	73 (20.2)
Medicaid	1,013 (7.7)	494 (11.7)	519 (5.8)		982 (7.7)	31 (8.6)
Medicare	8,196 (62.6)	2,673 (63.4)	5,523 (62.2)		8,018 (62.9)	178 (49.3)
TRI-CARE/military/veterans affairs	458 (3.5)	110 (2.6)	348 (3.9)		450 (3.5)	8 (2.2)
Insurance NOS	1,013 (7.7)	327 (7.8)	686 (7.7)		962 (7.6)	51 (14.1)
Insurance status unknown	280 (2.1)	126 (3.0)	154 (1.7)		273 (2.1)	7 (1.9)
Histologic type				<.0001			.0002
Adenocarcinoma	4,444 (33.9)	1,233 (29.2)	3,211 (36.1)		4,285 (33.6)	159 (44.0)
Large cell carcinoma	558 (4.3)	150 (3.6)	408 (4.6)		540 (4.2)	18 (5.0)
Sarcomatoid cell carcinoma	61 (0.5)	18 (0.4)	43 (0.5)		60 (0.5)	1 (0.3)
Squamous cell carcinoma	2,853 (21.8)	915 (21.7)	1,938 (21.8)		2,775 (21.8)	78 (21.6)
NSC NOS	5,183 (39.6)	1,900 (45.1)	3,283 (37.0)		5,078 (39.9)	105 (29.1)
Grade of tumor				0.489			<.0001
I and II	1,216 (9.3)	377 (8.9)	839 (9.4)		1,151 (9.0)	65 (18.0)
III and IV	3,655 (27.9)	1,163 (27.6)	2,492 (28.1)		3,551 (27.9)	104 (28.8)
Unknown	8,228 (62.8)	2,676 (63.5)	5,552 (62.5)		8,036 (63.1)	192 (53.2)
Lung cancer sequence number				0.202			<.0001
First primary only	10,670 (81.5)	3,471 (82.3)	7,199 (81.0)		10,438 (81.9)	232 (64.3)
First primary	258 (2.0)	81 (1.9)	177 (2.0)		190 (1.5)	68 (18.8)
Second	2,171 (16.6)	664 (15.7)	1,507 (17.0)		2,110 (16.6)	61 (16.9)
First course of treatment				<.0001			<.0001
No definite therapy or surgery at regional/distant site	4,969 (37.9)	1,794 (42.6)	3,175 (35.7)		4,907 (38.5)	62 (17.2)
Chemotherapy only	2,048 (15.6)	541 (12.8)	1,507 (17.0)		1,967 (15.4)	81 (22.4)
Radiation therapy only	2,602 (19.9)	833 (19.8)	1,769 (19.9)		2,575 (20.2)	27 (7.5)
Surgical procedure at primary site only	214 (1.6)	72 (1.7)	142 (1.6)		166 (1.3)	48 (13.3)
Polytherapy	3,266 (24.9)	976 (23.1)	2,290 (25.8)		3,123 (24.5)	143 (39.6)
ES (surviving 5 years or more)	361 (2.8)	107 (2.5)	254 (2.9)	.2938	–	–	–
Survival months	3.4 (1.4, 7.5)	3.3 (1.3, 7.2)	3.4 (1.4, 7.7)	–	3.2 (1.3, 7.0)	93.1 (75.4, 120.0)	–
Residence in Appalachia	4,216 (32.2)	–	–	–	4,109 (32.3)	107 (29.6)	.2938
CCI (n = 2,573)				0.01			0.136
0	1,063 (41.3)	300 (37.3)	763 (43.1)		1,007 (41.0)	56 (48.7)
1-2	1,184 (46.0)	404 (50.3)	780 (44.1)		1,134 (46.1)	50 (43.5)
≥3	326 (12.7)	99 (12.3)	227 (12.8)		317 (12.9)	9 (7.8)

Notes: Numbers are presented as counts and percentages for categorical variables, whereas median (25th, 75th percentiles) for nonnormally distributed, continuous variables. Percentages may not add up to 100% because of rounding.

Private insurance Includes fee-for-service/managed care, Health Maintenance Organization (HMO), and Preferred Provider Organization (PPO).

Polytherapy included patients who received any combination of the following therapies: chemotherapy/radiation therapy/surgery at primary site/other nonsystemic therapy.

For comparison analyses, listwise deletion of 3 missing observations was performed for the race variable.

Charlson comorbidity index was available for a subsample of 2,573 patients diagnosed between 2007 and 2011.

Abbreviations: NOS, not otherwise specified; NSC NOS, non-small-cell carcinoma not otherwise specified; ES, exceptional survival; CCI, Charlson comorbidity index.

We observed a larger percentage of adenocarcinomas (36.1% vs 29.2%, P < .0001) and significantly lower proportions of NSC NOS carcinomas (37.0% vs 45.1%, P < .0001) among residents of the non-Appalachian region. Furthermore, as compared to residents of non-Appalachian Kentucky, a notably higher proportion of patients residing in the Appalachian region did not receive any definite therapy or underwent surgery at a regional or distant site (42.6% vs 35.7%, P < .0001), slightly more underwent surgery only (1.7% vs 1.6%), but fewer had received other systemic treatments such as chemotherapy, radiotherapy, or polytherapy (12.8% vs 17.0%, 19.8% vs 19.9%, 23.1% vs 25.8%, respectively).

Results From Unadjusted and Adjusted Analyses

As seen in Table 2, results from the bivariate logistic regression revealed no significant difference in ES between patients from Appalachian versus non-Appalachian Kentucky (odds ratio [OR] = 0.93; 95% CI: 0.72-1.21). Being female, having private insurance versus Medicare, and undergoing surgery versus chemotherapy were significantly associated with increased odds of ES (OR = 1.84 [1.46-2.32]; OR = 1.63 [1.21-.20]; OR = 7.02 [4.75-10.38], respectively). Increase in age at diagnosis, having a NSC NOS carcinoma, grade III/IV or unknown grade of tumor, and undergoing radiation therapy were associated with decreased odds of ES (OR = 0.98 [0.97-0.99]; OR = 0.48 [0.35-0.64]; OR = 0.54 [0.39-0.75]; OR = 0.42 [0.31-0.57]; OR = 0.26 [0.16-0.40], respectively).

Table 2.

Results From the Single Variable Logistic Regression Models, Using ES (Survival of 5 Years and More) as the Outcome of Interest (N = 8,130; 4,969 Patients Who Did Not Receive Any Definite Treatment Were Excluded; ES Cases: n = 299)

Characteristic	OR (95% CI)	P
Appalachia region	0.93 (0.72-1.21)	.5996
Age at diagnosis		.0002
1-year increase	0.98 (0.97-0.99)
5-year increase	0.90 (0.86-0.95)
Age at diagnosis (cat.)
Less than 50 years old	Reference
Those aged 50-64	0.65 (0.46-0.93)	.0179
Those aged 65-74	0.55 (0.38-0.79)	.0012
75 years or older	0.39 (0.25-0.61)	<.0001
Sex at birth (female)	1.84 (1.46-2.32)	<.0001
Race
White	Reference
Black	0.95 (0.58-1.54)	.8251
Other	2.92 (0.68-12.66)	.1516
Ethnicity
Non-Hispanic	Reference
Hispanic	2.47 (0.75-8.10)	.1370
Spanish surname only or unknown	0.84 (0.26-2.67)	.7663
Tobacco use
Never used	Reference
Users of tobacco products	0.80 (0.50-1.28)	.3542
Not recorded/unknown	0.79 (0.46-1.36)	.3982
Insurance type
Medicare	Reference
Medicaid	1.15 (0.74-1.78)	.5454
TRICARE/military/veterans affairs	0.70 (0.32-1.50)	.3596
Private insurance	1.63 (1.21-2.20)	.0015
Not insured or self-pay	0.47 (0.21-1.07)	.0714
Insurance, NOS	1.80 (1.29-2.54)	.0007
Insurance status unknown	1.76 (0.70-4.40)	.2274
Histologic type ICD-O-3
Adenocarcinoma	Reference
Large cell carcinoma	0.88 (0.53-1.47)	.6325
NSC NOS	0.48 (0.35-0.64)	<.0001
Sarcomatoid cell carcinoma	0.54 (0.07-3.98)	.5481
Squamous cell carcinoma	0.77 (0.58-1.03)	.0740
Grade
Grade I and II	Reference
Grade III and IV	0.54 (0.39-0.75)	.0002
Unknown	0.42 (0.31-0.57)	<.0001
Lung cancer sequence # (cont.)	1.32 (1.17-1.48)	<.0001
Lung cancer sequence # (cat.)
First primary only	Reference
First	16.04 (11.43-22.51)	<.0001
Second or higher	1.55 (1.14-2.11)	.0056
First course of treatment
Chemotherapy	Reference
Radiation therapy	0.26 (0.16-0.40)	<.0001
Surgical procedure at primary site only	7.02 (4.75-10.38)	<.0001
Polytherapy	1.11 (0.84-1.47)	.4549
CCI (n = 2,573)
0	Reference
1-2	0.79 (0.54-1.17)	.2443
≥3	0.51 (0.25-1.04)	.0654

Notes: To compare different types of therapies used to treat lung cancer, 4,969 patients who did not receive any definite treatment were not included in the regression modeling. ES cases: n = 299.

Complete case analysis with listwise deletion of 3 missing observations was performed for the race variable.

Private insurance Includes fee-for-service/managed care, Health Maintenance Organization (HMO), and Preferred Provider Organization (PPO).

Polytherapy included patients who received any combination of the following therapies: chemotherapy/radiation therapy/surgery at primary site/other nonsystemic therapy.

Charlson comorbidity index was available for a subsample of 2,573 patients diagnosed between 2007 and 2011.

Abbreviations: NOS, not otherwise specified; NSC NOS, non-small-cell carcinoma not otherwise specified; ES, exceptional survival; CCI, Charlson comorbidity index; OR, odds ratio; CI, confidence interval.

As seen in Table 3, results from the final adjusted logistic regression revealed no significant differences between patients residing in Appalachian versus non-Appalachian Kentucky concerning the odds of ES (adjusted odds ratio [AOR]: 0.96 [0.74-1.25]). Being female and undergoing surgery only were associated with higher odds of ES (AOR: 1.82 [1.43-2.31]; AOR: 5.99 [3.96-9.06], respectively). Increasing age, having an unspecified histologic type of cell carcinoma, grade III/IV or unknown grade of tumor, and undergoing radiotherapy were associated with decreased odds of ES (AOR: 0.98 [0.97-0.99]; AOR: 0.65 [0.48-0.90]; AOR: 0.67 [0.47-0.95]; AOR: 0.61 [0.43-0.86]; 0.27 [0.17-0.42], respectively).

Table 3.

Results From the 2 Multivariable Logistic Regression Models, Using ES (Survival of 5 Years and More) as the Outcome of Interest

Characteristic	All Observations: n = 8,130		Subset Analysis: n = 2,573
	AOR (95% CI)	P	AOR (95% CI)	P
Appalachia region	0.96 (0.74-1.25)	.7581	1.04 (0.69-1.60)	.8287
Age at diagnosis	0.98 (0.97-0.99)	.0019	0.98 (0.96-1.00)	.0946
Sex at birth (female)	1.82 (1.43-2.31)	<.0001	1.55 (1.05-2.29)	.0277
Histologic type ICD-O-3
Adenocarcinoma	Reference		Reference
Large cell carcinoma	1.05 (0.62-1.79)	.8502	0.42 (0.10-1.78)	.2379
NSC NOS	0.65 (0.48-0.90)	.0089	0.58 (0.34-1.01)	.0541
Squamous cell carcinoma	0.90 (0.67-1.21)	.4719	0.92 (0.59-1.45)	.7232
Sarcomatoid cell carcinoma	0.53 (0.07-4.15)	.5447	<.001 (0, ∞)	.9850
Grade
Grade I and II	Reference		Reference
Grade III and IV	0.67 (0.47-0.95)	.0263	0.82 (0.47-1.43)	.4820
Unknown	0.61 (0.43-0.86)	.0048	0.53 (0.30-0.92)	.0251
First course treatment
Chemotherapy only	Reference		Reference
Radiation therapy only	0.27 (0.17-0.42)	<.0001	0.25 (0.12-0.53)	.0003
Surgical procedure at primary site only	5.99 (3.96-9.06)	<.0001	5.69 (2.99–10.80)	<.0001
Polytherapy	1.03 (0.77-1.36)	.8641	1.03 (0.65-1.63)	.8010
CCI
0	–	–	Reference
1-2			0.81 (0.54-1.22)	.3229
≥3			0.48 (0.22-1.02)	.0558

Note: Model #1: To compare different types of therapies used to treat lung cancer, 4,969 patients who did not receive any definite treatment were not included in the regression modeling; ES cases: n = 299. Model #2: The subset analysis included 2,573 patients diagnosed between 2007 and 2011 with available information on the Charlson Comorbidity Index; ES cases: n = 115.

Polytherapy included patients who received any combination of the following therapies: chemotherapy/radiation therapy/surgery at primary site/other nonsystemic therapy.

Abbreviations: NOS, not otherwise specified; NSC NOS, non-small-cell carcinoma not otherwise specified; CCI, Charlson comorbidity index; AOR, adjusted odds ratio; CI, confidence interval.

Results From Sensitivity Analyses

Among 8,130 patients with definite therapies, 6,664 had NSCLC only, 191 had first primary NSCLC, and 1,275 had second primary NSCLC. When limiting to 6,855 patients with lung cancer as the first or only primary (1) and then to 6,664 patients with NSCLC only (2), we observed similar results regarding the significance and direction of all effects of interest. The only difference from the findings for the total sample (n = 8,130) was nonsignificant P values for the III/IV and unknown grade of tumor (1: P = .2414 and P = .0619; 2: P = .2436 and P = .1032, respectively), and unspecified histology (1: P = .0761; 2: P = .1005, respectively).

Results for the Subset Analysis

For 2,573 patients with available information on Charlson Comorbidity Index (CCI), the CCI ranged from 0 to 6, with a mean value of 1.084 (±SD: 1.288, median = 1). The CCI value ranged from 0 to 4 for the cases and from 0 to 6 for the controls. The cases had a lower average CCI compared to the controls (0.817 [±SD: 0.996] vs 1.096 [±SD: 1.299]). For residents of both regions, the CCI ranged from 0 to 6. The mean value for patients from the Appalachian region was slightly lower: 1.082 (±SD: 1.185) versus 1.085 (±SD: 1.333); the ES cases from Appalachian Kentucky had a slightly higher average CCI value as compared to cases from the non-Appalachian region (1.000 [±SD: 0.956] vs 0.734 [±SD: 1.009], respectively).

Table 3 presents the results of the subset analysis assessing associations between ES and various characteristics while controlling for the CCI. Our findings revealed that patients from Appalachian Kentucky had 4% higher odds of ES as compared to those from the non-Appalachian region; patients whose CCI was 1 or 2 had 19% and patients whose CCI was 3 or higher had 52% lower odds of ES as compared to patients with no comorbid conditions (ie, CCI = 0). However, none of these results were statistically significant (AOR: 1.04 [0.69-1.60], 0.81 [0.54-1.22], and 0.48 [0.22-1.02], respectively). Consistent with the findings for the total sample (n = 8,130), being female and undergoing surgery only were associated with higher odds of ES, whereas having an unknown grade of tumor and receiving radiotherapy only were associated with decreased odds of ES. Unlike the results for the total sample, factors such as age, having an unspecified histology and a III/IV grade of tumor were no longer significant (Table 3, subset analysis).

Discussion

This retrospective population-based case-control study aimed to examine ES and compare patients with stage IV NSCLC residing in Appalachian versus non-Appalachian Kentucky. Our findings revealed that differences in ES were not related to geography alone, but rather to age at diagnosis, sex at birth, histology, and treatment. Although the proportion of ES was lower among patients who resided in Appalachian versus non-Appalachian Kentucky, the difference was negligible and not statistically significant. A small proportion of exceptional survivors in the total population of stage IV NSCLC patients in Kentucky may account for the lack of statistical significance of this finding, especially for such a small magnitude of the difference observed. Younger age and slightly higher percentage of patients who underwent surgery could be some of the reasons why we observed a higher than expected proportion of ES among residents of Appalachian Kentucky; both age and undergoing surgery were independently associated with increased odds of ES. Perhaps strong social support values that are normative⁴² among rural communities like Appalachian Kentucky could have also exerted a positive effect on ES outcomes among patients residing in this region. Evidence shows that social support can decrease the burden of depression,^43,44 especially in rural older adults,⁴⁵ and improve the quality of life (QoL),^46-48 which can positively affect survival in lung cancer patients.⁴⁹ Furthermore, the lack of information on where patients had obtained treatment and on patients’ mobility following stage IV lung cancer diagnosis could have biased our results toward the null. Quality of medical services and access to specialized cancer centers play significant roles in lung cancer survival.⁵⁰ Early palliative care can also improve survival in advanced lung cancer.⁵¹ In this study, the place of residence was captured only at the time of diagnosis; we assumed that patients were longtime residents of their counties. However, patients moving from Appalachia to the non-Appalachian region after diagnosis to seek access to better health care, specialized providers and facilities due to existing health care shortages⁵² in rural Appalachian Kentucky could account for the lack of observed differences.

In addition, several other factors that are presumed to be related to Appalachian Kentucky were not associated with ES. Inconsistent with previous research indicating that less or no smoking is associated with better survival in lung cancer patients,⁵³ our analysis showed no differences in ES by tobacco use. It is highly likely that because stage IV NSCLC is a terminal disease characterized by short mean and median survival times for all patients regardless of their smoking status, it was hard to observe significant differences in ES between smokers and nonsmokers in our study. Besides, the number of ES cases was not very large and there were few nonsmokers among them. No associations between tobacco use and ES, however, should not lead to a wrong conclusion and underestimation of the harmful effect of smoking,¹ especially regarding the risk of lung cancer associated with it in Appalachian Kentucky.^26,28,54 Quitting smoking after a lung cancer diagnosis should be considered on any stage of the disease as it can prepare patients for treatment⁵⁵ and reduce the risk of new lung cancer.^55,56

In contrast to previous studies,⁵³ our results showed that patients with adenocarcinomas were not different from those with large or squamous cell carcinomas regarding their odds of ES; patients with adenocarcinomas had better odds of ES only compared to those with unspecified histology. Perhaps a positive association of adenocarcinoma histology and survival outcomes documented by previous research could be owed to interactions between treatment and histology, as the latter plays a significant role in determining lung cancer therapies.^57,58 Adjusting for histology was important as large clusters of squamous and small cell, as well as other types of lung cancers but adenocarcinomas are observed in Appalachian Kentucky.⁵⁹ According to Liu and associates, there were differences in genetic mutations in patients from Appalachian Kentucky who were diagnosed with squamous cell lung cancer, indicating a greater proportion of mutations of IDH1 and PCMTD1 in patients from this region.³⁰

Such a strong factor as general health status can affect the choice of treatment in lung cancer⁶⁰ or can be independently associated with a patient’s ability to cope with the disease and thus, a patient’s prognosis.⁶¹ The data included information regarding patients’ comorbidity score (ie, CCI) for those who were diagnosed between 2007 and 2011. We used CCI as a proxy to overall patient health; the lower the CCI value the better the overall health. The analysis did not reveal clear associations between the CCI and ES, but we observed a trend toward lower odds of ES with higher CCI. Interestingly, in contrast to the results from the total sample, after adjusting for CCI, the odds of ES for patients from the Appalachian region were slightly, although not significantly, higher than for their counterparts from non-Appalachian Kentucky. Perhaps a bigger sample size, including patients diagnosed earlier than 2007, would have given us more power to detect significant differences.

Independent Factors Associated With ES

Consistent with previous research showing associations of survival outcomes with age⁶² and gender,^3,63-66 our findings revealed that the younger the patient’s age at diagnosis, the better the odds of ES; women had better odds of ES compared to men. Evidence shows that higher grade of tumors in patients with NSCLC is associated with worse prognosis.⁶⁷ In line with previous research,⁶⁷ our findings revealed that poorly differentiated and undifferentiated carcinomas were associated with lower odds of ES.

Similar to previous studies,^13,15,17 patients with chemotherapy alone had higher odds of ES as compared to those who received radiation therapy only. Indeed, evidence shows the survival benefit of chemotherapy in patients with advanced NSCLC, both local¹⁵ and extensive,⁶⁸ for whom neither surgery nor radiotherapy would be recommended.^15,68

Strikingly, surgery as the only treatment was associated with much higher odds of ES as compared to other treatments including chemotherapy, which is the number one recommended therapy for the most advanced forms of NSCLC either progressive or newly diagnosed.⁶⁹ In our study, the surgery group was the smallest (214, 1.6%) as compared to other treatment groups; however, it included more patients than we had expected to see in a study on patients with stage IV NSCLC.⁷⁰ The available information was limited to fully understand why these 214 patients had undergone surgery as their only treatment for an advanced lung cancer. Perhaps the burden of stage IV NSCLC was less severe in these patients; they likely had operable tumors and good enough health indicators to qualify for a surgery.

Although radiotherapy is believed to improve survival for all stages of NSCLC,⁷¹ it is mostly effective in cases of early stage cancers.⁷² In patients with inoperable and latestage NSCLC, radiotherapy may reduce the risk for death if combined with chemotherapy,⁷³ but evidence shows that a thoracic radiotherapy alone has no effect on prolonged survival for advanced NSCLC.⁷⁴ Our findings revealed that patients who received radiotherapy alone had 73% decreased odds of ES as compared to chemotherapy only. More research needs to be done in this area, but if radiotherapy alone as the first course of treatment indeed has a negative effect on prolonged survival for patients with stage IV NSCLC, clinical practice might need to be changed. Radiotherapy has many side effects including but not limited to cough and shortness of breath, fatigue, hair loss, and skin and throat changes.⁷⁵ Research shows that postoperative radiotherapy can negatively affect patients’ survival.^76,77

Perhaps severity of stage IV NSCLC and burden of symptoms could better explain differential outcomes in ES and its associations with treatment. Besides other factors (age, comorbidity, histology, and performance status), the decision about treatment for stage IV NSCLC depends on location and extent of cancer spread.⁶⁹ When cancer is limited to the lung only, patients can benefit from resection of the tumor and/or affected lymph nodes.⁶⁹ To qualify for a surgery, the functioning of the lungs and good overall health of patients are taken into account.⁷⁰ For most newly diagnosed patients with nonoperable tumors or progressive and metastatic stage IV NSCLC, chemotherapy alone or in combination with other therapies remains the main choice of treatment.⁶⁹ Only in selective patients, for example those who have lung cancer metastases in the brain, surgery and/or radiotherapy may be more appropriate.⁷⁸ For stage IV lung cancer patients who are not candidates for chemotherapy or surgery,⁶⁹ radiotherapy alone is recommended only as an effective method for symptom palliation and QoL improvement.^79,80 Additional information on symptom burden, and types⁸¹ and extent of metastases that are reflected in the latest staging system for lung cancer⁸² can assist in measuring confounding by indication and better understanding of the relationship between treatment and ES.

Limitations

This study had several limitations. First, we lacked data on potential confounding factors such as occupational exposures, genetic profiling, and patient-related factors such as performance status,⁵³ QoL,^49,53 and social support.^43,44,46-48 Second, due to limited data, we did not account for possible subsequent treatment(s) nor for the actual cause of death, as the data were limited only to patients’ vital status. Third, advances in surveillance and lung cancer management that happened in different years between 2000 and 2011 could have influenced some of our results, but they would likely have a nondifferential effect on associations of ES with the place of residence. There might have been an increase in the number of patients receiving (1) adjuvant chemotherapy following surgery, (2) maintenance therapy (ie, chemotherapy drug or targeted therapy), (3) palliative care following initial chemotherapy, (4) annual computed tomography scanning among heavy smokers, as well as (5) aggressive chemotherapy in older patients.⁸³ The updated 2004 World Health Organization histological classification of lung cancer could have also had a significant impact on histology-related treatment selection.⁸⁴

Additionally, we could not determine exactly how many patients were without treatment. The original KCR variable combined the 2 groups into 1 category—those who received no definitive treatment for their lung cancer or underwent a surgery at regional and/or distant sites only. We included 4,969 (37.9%) of patients who belonged to this group in descriptive analysis only and found that 62 (1.2%) of them demonstrated ES. Further research is needed to determine characteristics related to such an extraordinary survival among patients without definite treatment(s).

Despite its limitations, this study is the first to examine ES in Kentucky patients with stage IV NSCLC. We did not find enough evidence to support the hypothesis that a significantly lower proportion of exceptional survivors exist among patients who resided in the Appalachian region. In summary, differences in ES in patients diagnosed with stage IV NSCLC were not related to residence in Appalachian Kentucky. Being female and undergoing surgery only as the first course of treatment were associated with higher odds of ES. Increasing age, unspecified histology, having poorly differentiated or undifferentiated carcinomas, and receiving radiation therapy alone as the first course of treatment were associated with decreased odds of ES.