Increase in the Life Expectancy of Cancer Patients in the United States

Abstract

Background:

Cancer is becoming more of a chronic disease due to improvements in treatment and early detection for multiple cancer sites. To gain insight on increased life expectancy due to these improvements, we quantified trends in the loss in expectation of life (LEL) due to a cancer diagnosis for six cancer sites from 1975 through 2018.

Methods:

We focused on patients diagnosed with female breast cancer, chronic myeloid leukemia (CML), colon and rectum cancer (CRC), diffuse large B-cell lymphoma (DLBCL), lung cancer, or melanoma between 1975 and 2018 from nine Surveillance, Epidemiology, and End Results (SEER) cancer registries. Life expectancies for cancer patients aged 50+ were modeled using flexible parametric survival models. LEL was calculated as the difference between general population life expectancy and life expectancy for cancer patients.

Results:

Over 2 million patients were diagnosed with one of the six cancers between 1975 and 2018. Large increases in life expectancy were observed between 1990 and 2010 for female breast, DLBCL, and CML. Patients with CRC and melanoma had more gradual improvements in life expectancy. Lung cancer LEL only began decreasing after 2005. Increases in life expectancy corresponded with decreases in LEL for cancer patients.

Conclusions:

The reported gains in life expectancy largely correspond to progress in the screening, management, and treatment of these six cancers since 1975.

Impact:

LEL provides an important public health perspective on how improvements in treatment and early detection and their impacts on survival translate into changes in cancer patients’ life expectancy.

Introduction

In the last several decades, there have been important improvements in treatment and early detection for many cancer sites. Cancer patients are being cured and surviving longer (1-3), and cancer is becoming more of a chronic disease. While cancer survival is the most widely used measure to describe cancer patients’ prognosis (4), it is important to consider alternative measures that can offer different perspectives of the impact of cancer and improvements in cancer care on cancer patients’ life expectancy. From a public health perspective, it is informative to examine how advancements in treatment and screening and their corresponding effects on survival translate into changes in the life expectancy of cancer patients and how their life expectancy compares to the life expectancy of the general population.

However, life expectancy has been underutilized in the context of cancer survival research, except for a handful of studies conducted in Sweden (5-8) and Australia (9-12) and a single study using US cancer data (13). The main challenge is the requirement of an extensive follow-up period to accurately assess survival rates over an individual’s entire lifetime, which proves particularly difficult for patients diagnosed with cancer in recent years. Nonetheless, this obstacle can potentially be overcome by employing advanced models capable of extrapolating survival outcomes, for example the flexible parametric survival model which has been implemented in standard statistical packages (14-19).

The aim of this study was to evaluate the changes in life expectancy and the loss in expectation of life (LEL) due to a cancer diagnosis among cancer patients diagnosed between 1975 and 2018. The loss in expectation of life (LEL) due to a cancer diagnosis is defined as the difference between the life expectancy in the general population and the life expectancy among the cancer patients (13, 20). To illustrate this, we specifically focused on female breast cancer, chronic myeloid leukemia (CML), colon and rectum cancer (CRC), diffuse large B-cell lymphoma (DLBCL), lung cancer, and melanoma of the skin. These cancer types were chosen due to significant reported survival improvements (3). By utilizing LEL measures, we aimed to estimate the impact of improvements in breast, CML, CRC, DLBCL, lung, and melanoma survival over time on patients’ life expectancy and number of years lost due to cancer from 1975 to 2018. Importantly, our analysis specifically aimed to assess how the diagnosis of cancer affects life expectancy for recently diagnosed patients.

Materials and Methods

Study Population

We used survival data from 1975 through 2018 in nine Surveillance, Epidemiology, and End Results (SEER) Program registries representing about 9% of the US population. We identified females diagnosed with invasive breast cancer and males and females diagnosed with CML, CRC, lung cancer, and melanoma using the SEER site recode variable (21). Males and females diagnosed with DLBCL were identified using the 2020 revision of the SEER lymphoid neoplasm recode variable (22). In constructing the study cohort, we excluded cases diagnosed by autopsy or death certificate and cases alive with no survival time. We further restricted analysis to cases that were a man or woman’s first or only cancer since a prior cancer diagnosis may affect patient prognosis. Finally, to maintain consistency and avoid extensive survival extrapolations based on a small number of cases, we restricted the cohort to patients diagnosed at ages 50 and above. The final analytic set consisted of 472,496 females diagnosed with invasive breast cancer and males and females diagnosed with CML (11,862), CRC (404,344), DLBCL (44,866), lung (980,220), or melanoma (103,040).

Statistical Analysis

The LEL due to cancer is defined as the difference between the life expectancy of the cancer patients and the life expectancy of comparable individuals from the general population matched by sex, age, and calendar year. To calculate life expectancy of the cancer cohort, we utilized the approach outlined in Andersson et al (14). Specifically, we employed a flexible parametric survival model, which regresses the cumulative excess hazard on the log scale and enables extrapolation of relative survival data by single age at diagnosis, single calendar year, sex, and other factors using spline models as described in greater detail below.

Regression analysis of relative survival for cancer patients diagnosed in 1975-2018 with follow-up to December 31, 2018 was carried out using flexible parametric survival models for each cancer site separately, as described by Lambert et al and others (16-18). We included age at diagnosis and year of diagnosis as covariates in the models, and sex was included as a covariate for all sites except female breast cancer. To capture the nonlinear relationships and potential complexities associated with age and year at diagnosis, we modeled these variables using restricted cubic splines. We included all two-way interaction terms between age, sex, and year in the models for melanoma, lung cancer, DLBCL, CRC, and CML. For female breast cancer, an interaction between age and year was included in the model. The models included time-dependent effects with two degrees of freedom for age and year at diagnosis to allow for nonproportional excess hazards. Candidate models were assessed using the Akaike information criterion (AIC) and Bayesian information criterion (BIC). For each cancer site, the LEL was predicted from a flexible parametric relative survival model with six knots for the baseline cumulative hazard. We did a sensitivity analysis for knot positions and the number of knots (2 to 7 knots) for the baseline cumulative hazard and assessed the impact of these choices by plotting the excess hazards and LELs over time. The LELs were found to be quite similar across different choices of knot positions and the number of knots used (Supplementary Tables S1, S2 for 55-year-olds). We used proc nlmixed in the %stpm2 macro in SAS 9.4 (SAS Institute, Cary, NC) to fit the flexible parametric models and proc iml for post-fit estimation. The 95% confidence intervals were calculated using variances that were derived via the delta method. The Supplementary Materials and Methods provides directions for obtaining access to the required SAS macros and an example SAS program for fitting a model and computing the LEL.

We used US population life tables by year, sex, and single year of age available from SEER*Stat for the years 1975 - 2018. For estimates after 2018, we appended the predicted life tables published by the Social Security Administration (23). These predicted life tables presented mortality rates using 10-year future intervals from 2020 to 2100. We applied linear interpolation to obtain the annual rates required for the calculation of expected life expectancy. We also applied a correction factor so that the projected rates were consistent with the observed data in 2018. Race-specific projected life tables were unavailable, so we combined all races for LEL analysis.

To calculate life expectancy among the cancer cohort, we extrapolated cancer patients’ all-cause survival by multiplying the predicted relative survival from the flexible parametric survival model with expected survival estimates. Once these extrapolated estimates are obtained, one can calculate survival curves for the cancer cohort and the general population for a given age, sex, and calendar year. The difference in area under these two curves is the estimated LEL due to cancer. In Supplementary Figure S1, we show the survival curve of a 55-year-old female diagnosed in 2015 with DLBCL (red curve) and the survival curve of a cancer-free 55-year-old female (blue curve). The area between the two survival curves is the LEL due to cancer.

Data Availability

The data used in this study were obtained from nine SEER registries. Information on how to access SEER data is available at https://seer.cancer.gov/data/access.html. For availability of specific files used in this analysis, please contact the corresponding author.

Results

In Tables 1 and 2 and Figures 1-4, we present life expectancy among the general population and cancer cases and LEL due to cancer for specific ages (55, 65, 75, and 85 years old at diagnosis) in males and females over 1975-2018. For breast cancer, life expectancy improved gradually, stabilizing around 2000 for younger age groups. For example, a 55-year-old female breast cancer patient diagnosed in 1980 had an expected remaining life of 17.3 years (95% CI: 17.2-17.5) compared to 27.4 life-years remaining for a female of the same age without breast cancer; whereas a corresponding patient diagnosed in 2000 had 24.0 life-years (95% CI: 23.9-24.2) remaining (Table 1 and Figure 1A). Improvements were observed for older age groups, though on a smaller scale. For 75- and 85-year-old females, life expectancy among women diagnosed with breast cancer was similar to that of the general population and improved slightly over time.

Table 1.

Life expectancy of cancer cohort and general population and life expectancy lost (LEL) by cancer type (female breast, CML, CRC, DLBCL, lung, melanoma), age at diagnosis (55, 65, 75, 85), and year of diagnosis (1980, 1990, 2000, 2010) for females.

Cancer	Year	Life expectancy	Females (55)	Females (65)	Females (75)	Females (85)
Female Breast	1980	General population	27.4	18.9	11.9	6.5
		Cancer	17.3 [17.2-17.5]	12.5 [12.4-12.6]	8.6 [8.6-8.7]	4.7 [4.6-4.8]
		LEL	10.0 [9.9-10.2]	6.3 [6.3-6.4]	3.2 [3.2-3.3]	1.8 [1.7-1.8]
	1990	General population	28.2	19.4	12.0	6.6
		Cancer	20.9 [20.7-21.0]	15.2 [15.1-15.2]	9.9 [9.8-9.9]	5.3 [5.3-5.4]
		LEL	7.3 [7.2-7.4]	4.3 [4.2-4.4]	2.1 [2.1-2.2]	1.3 [1.3-1.4]
	2000	General population	28.9	20.2	12.5	6.6
		Cancer	24.0 [23.9-24.2]	17.2 [17.2-17.3]	11.0 [10.9-11.0]	5.7 [5.6-5.7]
		LEL	4.9 [4.7-5.0]	2.9 [2.9-3.0]	1.5 [1.5-1.6]	0.9 [0.9-1.0]
	2010	General population	29.1	20.6	13.0	7.0
		Cancer	25.3 [25.1-25.5]	18.3 [18.2-18.4]	11.7 [11.6-11.8]	6.1 [6.0-6.1]
		LEL	3.8 [3.6-4.0]	2.4 [2.3-2.5]	1.3 [1.2-1.4]	0.9 [0.9-0.9]
CML	1980	General population	27.4	18.9	11.9	6.5
		Cancer	4.7 [4.2-5.2]	3.7 [3.4-4.0]	2.5 [2.3-2.7]	1.4 [1.2-1.5]
		LEL	22.7 [22.2-23.1]	15.2 [14.8-15.5]	9.4 [9.2-9.6]	5.1 [5.0-5.2]
	1990	General population	28.2	19.4	12.0	6.6
		Cancer	6.2 [5.5-6.9]	4.3 [3.9-4.7]	2.8 [2.6-3.0]	1.7 [1.5-1.8]
		LEL	22.0 [21.3-22.7]	15.1 [14.7-15.5]	9.2 [9.0-9.4]	5.0 [4.8-5.1]
	2000	General population	28.9	20.2	12.5	6.6
		Cancer	15.1 [14.0-16.2]	8.4 [7.8-9.1]	4.2 [3.9-4.5]	1.9 [1.8-2.1]
		LEL	13.8 [12.7-14.9]	11.7 [11.1-12.4]	8.3 [8.0-8.6]	4.7 [4.6-4.9]
	2010	General population	29.1	20.6	13.0	7.0
		Cancer	22.0 [21.0-23.0]	13.2 [12.4-13.9]	6.3 [5.9-6.7]	2.7 [2.5-2.9]
		LEL	7.0 [6.1-8.0]	7.5 [6.7-8.2]	6.7 [6.3-7.1]	4.2 [4.0-4.4]
CRC	1980	General population	27.4	18.9	11.9	6.5
		Cancer	14.7 [14.5-15.0]	10.3 [10.1-10.4]	6.6 [6.5-6.7]	3.6 [3.5-3.6]
		LEL	12.7 [12.4-12.9]	8.6 [8.5-8.7]	5.3 [5.2-5.4]	2.9 [2.8-2.9]
	1990	General population	28.3	19.5	12.0	6.6
		Cancer	16.8 [16.5-17.0]	11.8 [11.7-12.0]	7.5 [7.4-7.5]	4.1 [4.1-4.1]
		LEL	11.5 [11.3-11.8]	7.7 [7.5-7.8]	4.6 [4.5-4.6]	2.5 [2.5-2.6]
	2000	General population	29.0	20.3	12.6	6.6
		Cancer	19.0 [18.8-19.2]	13.1 [12.9-13.2]	8.3 [8.2-8.3]	4.3 [4.2-4.3]
		LEL	10.0 [9.8-10.3]	7.2 [7.1-7.4]	4.3 [4.2-4.4]	2.3 [2.3-2.4]
	2010	General population	29.2	20.8	13.1	7.1
		Cancer	20.3 [20.1-20.6]	13.5 [13.3-13.6]	8.5 [8.5-8.6]	4.4 [4.4-4.5]
		LEL	8.9 [8.7-9.1]	7.3 [7.1-7.5]	4.6 [4.5-4.7]	2.6 [2.6-2.7]
DLBCL	1980	General population	27.4	18.9	11.9	6.5
		Cancer	10.7 [9.9-11.4]	6.4 [5.9-6.8]	3.7 [3.4-3.9]	1.9 [1.7-2.0]
		LEL	16.7 [16.0-17.5]	12.5 [12.1-12.9]	8.2 [8.0-8.5]	4.6 [4.4-4.7]
	1990	General population	28.3	19.5	12.0	6.6
		Cancer	12.4 [11.8-13.1]	7.7 [7.3-8.1]	4.3 [4.1-4.5]	2.0 [1.9-2.2]
		LEL	15.9 [15.2-16.5]	11.8 [11.4-12.2]	7.8 [7.6-8.0]	4.6 [4.5-4.7]
	2000	General population	29.0	20.3	12.6	6.6
		Cancer	16.3 [15.7-16.8]	10.5 [10.1-10.8]	5.7 [5.5-5.9]	2.6 [2.5-2.7]
		LEL	12.8 [12.2-13.3]	9.8 [9.5-10.2]	6.8 [6.7-7.0]	4.0 [3.9-4.1]
	2010	General population	29.2	20.8	13.1	7.1
		Cancer	19.6 [19.0-20.3]	12.8 [12.4-13.2]	7.2 [7.0-7.5]	3.4 [3.3-3.5]
		LEL	9.6 [9.0-10.2]	8.0 [7.6-8.4]	5.9 [5.7-6.2]	3.7 [3.5-3.8]
Lung	1980	General population	27.4	18.9	11.9	6.5
		Cancer	4.2 [4.1-4.3]	2.8 [2.7-2.8]	1.7 [1.6-1.7]	1.0 [0.9-1.0]
		LEL	23.2 [23.1-23.3]	16.1 [16.0-16.2]	10.2 [10.2-10.3]	5.5 [5.5-5.5]
	1990	General population	28.3	19.5	12.0	6.6
		Cancer	4.4 [4.3-4.5]	3.0 [2.9-3.0]	1.9 [1.8-1.9]	1.0 [1.0-1.0]
		LEL	23.9 [23.8-24.0]	16.5 [16.5-16.6]	10.2 [10.1-10.2]	5.6 [5.6-5.6]
	2000	General population	29.0	20.3	12.6	6.6
		Cancer	4.8 [4.8-4.9]	3.4 [3.4-3.5]	2.2 [2.2-2.2]	1.1 [1.1-1.1]
		LEL	24.2 [24.1-24.2]	16.9 [16.8-16.9]	10.4 [10.4-10.4]	5.5 [5.5-5.5]
	2010	General population	29.2	20.8	13.1	7.1
		Cancer	6.0 [5.9-6.1]	4.4 [4.4-4.5]	2.9 [2.9-3.0]	1.5 [1.5-1.5]
		LEL	23.2 [23.1-23.3]	16.3 [16.3-16.4]	10.2 [10.2-10.2]	5.6 [5.5-5.6]
Melanoma	1980	General population	27.4	18.9	11.9	6.5
		Cancer	24.0 [23.6-24.4]	16.1 [15.8-16.4]	9.6 [9.4-9.9]	4.8 [4.6-5.0]
		LEL	3.4 [3.0-3.8]	2.8 [2.5-3.0]	2.3 [2.0-2.5]	1.7 [1.5-1.9]
	1990	General population	28.3	19.5	12.0	6.6
		Cancer	26.2 [25.9-26.5]	17.7 [17.5-17.9]	10.5 [10.4-10.7]	5.5 [5.3-5.6]
		LEL	2.1 [1.8-2.3]	1.8 [1.6-2.0]	1.5 [1.3-1.7]	1.2 [1.0-1.3]
	2000	General population	29.0	20.3	12.6	6.6
		Cancer	27.8 [27.6-27.9]	19.2 [19.1-19.3]	11.6 [11.5-11.7]	5.8 [5.7-5.9]
		LEL	1.2 [1.1-1.4]	1.1 [1.0-1.2]	1.0 [0.9-1.1]	0.8 [0.7-0.9]
	2010	General population	29.2	20.8	13.1	7.1
		Cancer	28.6 [28.5-28.7]	20.2 [20.1-20.2]	12.6 [12.5-12.6]	6.5 [6.5-6.6]
		LEL	0.6 [0.5-0.8]	0.6 [0.5-0.7]	0.6 [0.5-0.7]	0.5 [0.4-0.6]

Table 2.

Life expectancy of cancer cohort and general population and life expectancy lost (LEL) by cancer type (CML, CRC, DLBCL, lung, melanoma), age at diagnosis (55, 65, 75, 85), and year of diagnosis (1980, 1990, 2000, 2010) for males.

Cancer	Year	Life expectancy	Males (55)	Males (65)	Males (75)	Males (85)
CML	1980	General population	22.9	14.9	9.1	5.2
		Cancer	3.9 [3.5-4.2]	2.7 [2.5-2.9]	1.9 [1.8-2.1]	1.2 [1.1-1.3]
		LEL	19.1 [18.7-19.4]	12.2 [12.0-12.5]	7.2 [7.1-7.4]	4.0 [3.9-4.1]
	1990	General population	24.6	16.2	9.6	5.3
		Cancer	5.0 [4.5-5.5]	3.1 [2.8-3.3]	2.1 [2.0-2.2]	1.4 [1.3-1.5]
		LEL	19.6 [19.1-20.1]	13.1 [12.8-13.3]	7.5 [7.4-7.7]	3.9 [3.8-4.1]
	2000	General population	25.7	17.5	10.5	5.5
		Cancer	12.5 [11.6-13.5]	6.2 [5.7-6.7]	3.1 [2.9-3.4]	1.6 [1.5-1.7]
		LEL	13.1 [12.2-14.1]	11.3 [10.9-11.8]	7.3 [7.1-7.5]	3.9 [3.7-4.0]
	2010	General population	25.7	18.1	11.2	5.9
		Cancer	18.8 [17.9-19.7]	10.4 [9.7-11.0]	4.9 [4.6-5.2]	2.3 [2.1-2.4]
		LEL	6.9 [6.0-7.8]	7.8 [7.1-8.4]	6.4 [6.0-6.7]	3.6 [3.4-3.8]
CRC	1980	General population	22.9	14.9	9.1	5.2
		Cancer	11.8 [11.6-12.0]	8.1 [8.0-8.2]	5.1 [5.0-5.2]	2.8 [2.8-2.9]
		LEL	11.2 [11.0-11.4]	6.8 [6.7-6.9]	4.0 [4.0-4.1]	2.3 [2.3-2.4]
	1990	General population	24.7	16.2	9.6	5.3
		Cancer	14.3 [14.1-14.5]	9.9 [9.8-10.0]	6.1 [6.0-6.1]	3.3 [3.3-3.4]
		LEL	10.4 [10.2-10.6]	6.3 [6.2-6.4]	3.5 [3.5-3.6]	2.0 [2.0-2.1]
	2000	General population	25.7	17.6	10.5	5.5
		Cancer	16.4 [16.2-16.6]	11.3 [11.2-11.4]	6.9 [6.9-7.0]	3.5 [3.5-3.6]
		LEL	9.4 [9.2-9.5]	6.3 [6.1-6.4]	3.6 [3.5-3.6]	1.9 [1.9-2.0]
	2010	General population	25.9	18.2	11.3	5.9
		Cancer	17.3 [17.1-17.4]	11.5 [11.4-11.7]	7.2 [7.1-7.3]	3.6 [3.6-3.7]
		LEL	8.6 [8.5-8.8]	6.7 [6.5-6.8]	4.1 [4.0-4.2]	2.3 [2.3-2.4]
DLBCL	1980	General population	22.9	14.9	9.1	5.2
		Cancer	7.9 [7.4-8.5]	4.9 [4.5-5.2]	2.9 [2.7-3.1]	1.5 [1.4-1.7]
		LEL	15.0 [14.5-15.6]	10.0 [9.7-10.4]	6.2 [6.0-6.4]	3.6 [3.5-3.8]
	1990	General population	24.7	16.2	9.6	5.3
		Cancer	9.7 [9.2-10.2]	6.2 [5.9-6.5]	3.5 [3.3-3.6]	1.7 [1.5-1.8]
		LEL	15.0 [14.5-15.5]	10.0 [9.7-10.3]	6.1 [6.0-6.3]	3.7 [3.6-3.8]
	2000	General population	25.7	17.6	10.5	5.5
		Cancer	13.4 [13.0-13.9]	8.9 [8.6-9.2]	4.9 [4.7-5.0]	2.2 [2.1-2.3]
		LEL	12.3 [11.9-12.8]	8.7 [8.4-9.0]	5.6 [5.5-5.8]	3.3 [3.2-3.4]
	2010	General population	25.9	18.2	11.3	5.9
		Cancer	16.7 [16.2-17.2]	11.1 [10.8-11.5]	6.4 [6.2-6.6]	2.9 [2.8-3.0]
		LEL	9.2 [8.7-9.7]	7.1 [6.7-7.4]	4.9 [4.7-5.1]	3.0 [2.9-3.1]
Lung	1980	General population	22.9	14.9	9.1	5.2
		Cancer	2.7 [2.7-2.8]	1.9 [1.9-2.0]	1.3 [1.2-1.3]	0.8 [0.8-0.8]
		LEL	20.2 [20.1-20.3]	13.0 [12.9-13.0]	7.9 [7.8-7.9]	4.4 [4.4-4.4]
	1990	General population	24.7	16.2	9.6	5.3
		Cancer	3.0 [2.9-3.1]	2.2 [2.1-2.2]	1.4 [1.4-1.5]	0.8 [0.8-0.9]
		LEL	21.7 [21.6-21.7]	14.0 [14.0-14.0]	8.2 [8.1-8.2]	4.5 [4.5-4.5]
	2000	General population	25.7	17.6	10.5	5.5
		Cancer	3.2 [3.2-3.3]	2.5 [2.4-2.5]	1.7 [1.6-1.7]	0.9 [0.9-0.9]
		LEL	22.5 [22.5-22.6]	15.1 [15.1-15.2]	8.8 [8.8-8.9]	4.5 [4.5-4.6]
	2010	General population	25.9	18.2	11.3	5.9
		Cancer	3.7 [3.6-3.8]	3.0 [2.9-3.0]	2.1 [2.1-2.1]	1.5 [1.5-1.5]
		LEL	22.2 [22.1-22.3]	15.2 [15.2-15.3]	9.2 [9.2-9.3]	5.6 [5.5-5.6]
Melanoma	1980	General population	22.9	14.9	9.1	5.2
		Cancer	18.1 [17.8-18.5]	12.0 [11.7-12.2]	7.2 [7.0-7.4]	3.8 [3.6-3.9]
		LEL	4.8 [4.4-5.2]	3.0 [2.7-3.2]	2.0 [1.8-2.1]	1.4 [1.2-1.5]
	1990	General population	24.7	16.2	9.6	5.3
		Cancer	21.7 [21.5-22.0]	14.3 [14.1-14.4]	8.3 [8.2-8.4]	4.4 [4.3-4.5]
		LEL	3.0 [2.7-3.2]	1.9 [1.8-2.1]	1.3 [1.2-1.4]	0.9 [0.8-1.1]
	2000	General population	25.7	17.6	10.5	5.5
		Cancer	23.8 [23.6-23.9]	16.2 [16.1-16.3]	9.5 [9.4-9.6]	4.8 [4.7-4.8]
		LEL	2.0 [1.8-2.2]	1.4 [1.3-1.5]	1.0 [0.9-1.1]	0.7 [0.6-0.8]
	2010	General population	25.9	18.2	11.3	5.9
		Cancer	24.7 [24.6-24.9]	17.3 [17.2-17.4]	10.6 [10.5-10.7]	5.4 [5.3-5.5]
		LEL	1.2 [1.0-1.3]	0.9 [0.8-1.0]	0.7 [0.6-0.8]	0.5 [0.5-0.6]

Figure 1.

Life expectancies for female breast cancer and CML. Life expectancy of the general population and of patients with A) breast cancer, B) CML diagnosed at age 55 or 65, C) CML diagnosed at age 75 or 85 in the US over year of diagnosis and by age at diagnosis. Dashed line = general population, solid line = cancer cohort. Shaded area around the red solid line represents 95% CI. For breast cancer, general population is for female only.

Figure 4.

Loss in expectation of life (LEL) by cancer site. LEL presented for patients with A) breast, B) CML, C) CRC, D) DLBCL, E) lung, and F) melanoma in US, over year of diagnosis, by age at diagnosis and sex. Shaded area represents 95% CI.

There was a dramatic improvement in life expectancy among CML patients, especially in the 55- and 65-year age groups, but less so in the 75- and 85-year age groups. The increase in life expectancy was notable for patients diagnosed after 1990 in the younger age groups and came at a later point around 2000 in the 75- and 85-year age groups. For example, a 55-year-old male CML patient diagnosed in 2010 had 18.8 life-years (95% CI: 17.9-19.7) remaining as compared to 3.9 life-years (95% CI: 3.5-4.2) remaining for an analogous CML patient diagnosed in 1980 (Table 2 and Figures 1B-1C).

For CRC, life expectancy has improved steadily from 1975 to 2000 and stabilized after 2000 in all age groups. A 55-year-old male patient diagnosed with CRC in 1980 had 11.8 life-years (95% CI: 11.6, 12.0) remaining as compared to 22.9 life-years remaining without CRC. A similar patient diagnosed in 2010 had 17.3 life-years (95% CI: 17.1, 17.4) remaining (Table 2 and Figures 2A-2B). Improvements were observed for individuals aged 75 or 85 at diagnosis at a smaller scale.

Figure 2.

Life expectancies for CRC and DLBCL. Life expectancy of the general population and of A) CRC patients diagnosed at age 55 or 65, B) CRC patients diagnosed at age 75 or 85, C) DLBCL patients diagnosed at age 55 or 65, and D) DLBCL patients diagnosed at age 75 or 85 cancer in US, over year of diagnosis and by age at diagnosis. Dashed line = general population, solid line = cancer cohort. Shaded area around the red solid line represents 95% CI.

For DLBCL, life expectancy has improved steadily, stabilizing around 2010 in all age groups. In 1980, a 55-year-old male DLBCL patient had 7.9 life-years (95% CI: 7.4, 8.5) remaining as compared to 22.9 life-years remaining without DLBCL. A corresponding patient diagnosed in 2010 had 16.7 life-years (95% CI: 16.2-17.2) remaining (Table 2 and Figures 2C-2D). Improvements were observed for older age groups at a smaller scale.

For lung cancer, life expectancy for cancer patients remained stagnant until after 2005 and was generally much lower than that of the general population. For example, a 65-year-old male diagnosed with lung cancer in 1980 had 1.9 remaining life-years (95% CI: 1.9-2.0), whereas a 65-year-old male without lung cancer had 14.9 life-years remaining (Table 2 and Figures 3A-3B). By 2010, a 65-year-old male lung cancer patient had 3.0 life-years (95% CI: 2.9-3.0) remaining compared to 18.2 life-years remaining for a 65-year-old without lung cancer.

Figure 3.

Life expectancies for lung and melanoma. Life expectancy of the general population and of A) lung patients diagnosed at age 55 or 65, B) lung patients diagnosed at age 75 or 85, C) melanoma patients diagnosed at age 55 or 65, and D) melanoma patients diagnosed at age 75 or 85 cancer in US, over year of diagnosis and by age at diagnosis. Dashed line = general population, solid line = cancer cohort. Shaded area around the red solid line represents 95% CI.

For melanoma patients, on the other hand, their life expectancy was closer to the life expectancy of the general population and has increasingly approached that of the general population, being very similar in recent years. For example, a 65-year-old female melanoma cancer patient diagnosed in 1980, had 16.1 life-years (95% CI: 15.8-16.4) remaining as compared to 18.9 life-years remaining for the same age female without melanoma (Table 1 and Figures 3C-3D). By 2010, a 65-year-old female melanoma cancer patient had 20.2 life-years (95% CI: 20.1-20.2) remaining as compared to 20.8 life-years remaining for a 65-year-old female without melanoma.

The notable improvements in life expectancy among cancer cases compared to the general population resulted in a substantially decreased LEL due to cancer over the study period (Tables 1-2 and Figure 4). For example, a 55-year-old female cancer patient diagnosed in 2010 with breast cancer, CML, CRC, DLBCL, lung cancer, or melanoma is estimated to live an additional 8.0, 17.3, 5.6, 8.9, 1.8, or 4.6 years, respectively, compared to a 55-year-old female diagnosed in 1980. In 2010, a 55-year-old woman is estimated to lose 3.8 life-years (95% CI: 3.6-4.0) due to breast cancer, 7.0 life-years (95% CI: 6.1-8.0) due to CML, 8.9 life-years (95% CI: 8.7-9.1) due to CRC, 9.6 life-years (95% CI: 9.0-10.2) due to DLBCL, 23.2 life-years (95% CI: 23.1-23.3) due to lung cancer, and 0.6 life-years (95% CI: 0.5-0.8) due to melanoma. In general, life expectancy was slightly higher for females than males in both the general population and in the cancer cohort. As expected, life expectancy was also highest for younger ages and lowest for older ages for both sexes and increased steadily over time. Estimates of the proportion of loss in expectation of life (PLEL), which illustrates improvements in life expectancy for each cancer site relative to the life expectancy of the general population, are presented in the supplementary appendix (Supplementary Figure S2).

Discussion

The study findings indicate notable reductions in the loss in expectation of life for patients diagnosed with female breast cancer, CML, CRC, DLBCL, lung cancer, or melanoma between 1975 and 2018, reflecting advancements in cancer control. Younger patients aged 55 or 65 demonstrated the largest improvements in life expectancy. Except for melanoma, females demonstrated larger decreases in the loss in expectation of life compared to males. The most substantial enhancements in life expectancy for breast cancer, CML, and DLBCL were observed between 1990 and 2010. For CRC, improvements in life expectancy were mainly observed between 1975 and 2000. For lung cancer, little improvement in life expectancy was observed, though it was found to increase after 2005. Despite these improvements, life expectancy for cancer patients remains lower than that of individuals without cancer, except for melanoma where life expectancy increases occurred gradually and eventually reached the level of general population life expectancy. Results for CML are in line with those of Bower et al (6), who examined life expectancy among patients diagnosed with CML in Sweden between 1973 and 2013. Since 2010, life expectancy for cancer patients has mostly plateaued in each of the age groups examined.

Improvements in cancer treatments have had a significant impact on the survival of cancer patients, directly influencing their life expectancy. Additionally, screening can play a role in shaping life expectancy for screen-detected cancers such as breast cancer, colorectal cancer, lung cancer, and melanoma by altering the distribution of cancer stages among patients. This is achieved by diagnosing more individuals with early-stage or favorable prognosis cancers. However, quantifying the specific contributions of treatment versus screening in driving these improvements remains challenging. Recent papers have sought to quantify the value of cancer screening and potential gains in life expectancy due to screening (24, 25). For breast cancer, increases in life expectancy from 1975 to 2005 align with the FDA approval of numerous drugs for breast cancer over this period (Supplementary Figure S3) as well as the use of mammography for early detection especially among 50 to 74-year-old women (26, 27).

For CML, interferon-alpha was introduced in the 1980s, and few drugs were approved in the 1990s (28-30). Imatinib, which is a tyrosine kinase inhibitor (TKI) intended to treat patients in the chronic phase of disease, received FDA approval in 2001 (Supplementary Figure S4). As there are no screening programs for CML and advancements in treatment have been well-documented for this cancer site, the dramatic increases in life expectancy observed for patients diagnosed with CML after 2000 may correspond to the uptake of imatinib.

For CRC, increases in life expectancy largely occurred before 2005 for those aged 55 or 65, and for older individuals, there were slight improvements between 1975 and 2018. Numerous drugs received FDA approval between 1998 and 2006, which may, in part, account for the increase in life expectancy among CRC patients observed during this period (Supplementary Figure S5; 29, 30). Since 2015, multiple therapies have been approved for the treatment of CRC, however their impact has yet to be seen in our data. Colonoscopy was introduced to the US population in the late 1960s but gained popularity in the 1980s. One of the key benefits of colonoscopy is the ability to remove polyps during the procedure, which has the potential to alter the stage distribution of colorectal cancer cases and have a positive impact on improving the overall stage distribution and subsequent prognosis of colorectal cancer.

For DLBCL, the largest increase in life expectancy occurred between 1990 and 2005. No drugs were approved specifically for DLBCL in the 1990s, but during that period, CHOP (cyclophosphamide, doxorubicin, vincristine, plus prednisone) was the standard of care (Supplementary Figure S6; 29-31). Rituximab was FDA approved for relapsed/refractory low-grade or follicular non-Hodgkin’s lymphoma in 1997 and later received FDA approval for DLBCL specifically in 2006. Since 2006, the combined therapy R-CHOP (rituximab + CHOP) has become the standard of care. While rituximab was only approved for DLBCL in 2006, it is possible that doctors began treating DLBCL patients with R-CHOP in the late 1990s, which may have led to improvements in average life expectancy for patients diagnosed with DLBCL prior to 2006. Gains in life expectancy after 2006 have been modest. Since 2017, additional therapies have gained FDA approval, which may result in further increases in life expectancy. No screening programs are available for DLBCL, so the improvements in LEL that were observed since the mid-to-late 1990’s are most likely attributable to the treatment advances for this cancer site.

For lung cancer, the decline in LEL after 2005 is in line with decreases in non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC) mortality after 2006 (32). NSCLC and SCLC account for approximately 76% and 13%, respectively, of all lung cancer cases in the US (33). In recent years, the uptake of immune-based programmed cell death protein 1-programmed death ligand 1 (PD-1-PD-L1) inhibitors has significantly improved outcomes for NSCLC patients (Supplementary Figure S7; 32). The approval of a number of immune checkpoint inhibitors for NSCLC since 2010 (27) and the slow adoption of lung cancer screening (34) may both be contributing to the observed decrease in lung cancer LEL after 2005.

For melanoma, life expectancy for cancer patients closely matched that of the general population by 2000. Prior to 2000, dacarbazine was approved for metastatic melanoma in 1975 and interleukin-2 was approved in 1998 (29, 30, 35). Since 2011, numerous drugs have received FDA approval for melanoma treatment, which may have helped to close the gap in life expectancy between cancer patients and the non-cancer population (Supplementary Figure S8).

Our study has its limitations. Estimated life expectancies are based on extrapolations from a parametric model, which are pronounced for patients more recently diagnosed. This is reflected through the width of the confidence intervals. Our data includes cases with follow-up through 2018. Thus, life expectancy for patients more recently diagnosed is based on shorter follow-up and extrapolations from the regression model. These extrapolations may not reflect the impact of more recently approved therapies. However, the use of spline models for the baseline cumulative hazard reduces the impact of parametric assumptions on the resulting estimated life expectancies. Additionally, our sensitivity analysis demonstrated that the estimates were robust to changing the number or location of knots in the spline function. We also limited the analysis to patients aged 50 and over for whom data were available. Detailed information on confounders was largely unavailable and thus excluded from our models.

For breast cancer, healthy screener bias may be present, which indicates that women diagnosed with breast cancer have higher non-cancer survival than what is expected in general population life tables, likely due to healthier behavior or greater access to care. A prior study by Cho et al (36) reported that for women diagnosed with breast cancer at ages 50 and above, the non-cancer survival probability was higher than that of the general population, which is indicative of a healthy screening effect. For colorectal and lung cancer, which are other cancer sites with screening, no such bias was present. The healthy screener effect can lead to underestimated LEL for breast cancer because we used a lower life expectancy based on the US population life tables. Currently, the method has no way of accounting for this bias, but that is a potential area for future research.

One additional limitation is the use of all-races-combined life tables after 2018 as opposed to using race-stratified constant mortality rates after 2018. Currently, our estimates are based on all races combined due to the limitation of the projected life tables being only available for all races. However, there may be potential disparities in how LEL has changed over time that are important to explore. While we did not quantify the potential effect of using the projected life tables as opposed to using a constant mortality rate, this is a topic worth exploring in future research.

There are alternative approaches for calculating life expectancies and alternative metrics to evaluate the impact of cancer control on cancer survivorship. An alternative nonparametric approach for the calculation of LEL (37) does not rely on extrapolating survival beyond follow-up time. However, this method does not provide trend estimates. A model-based approach uses reference-adjusted expected mortality rates to remove the impact of other-cause mortality when making comparisons between cancer patients and the general population across calendar time (38). As in our study, Andersson et al (38) utilize flexible parametric survival models for relative survival. However, unlike our LEL estimates, reference-adjusted LEL may have less utility as a public health measure and may be more suited for comparison of groups over time. The joinpoint survival model (JPSurv) is another approach that involves modeling survival trends using joinpoint regression models (39, 40). The JPSurv method estimates when changes in survival occurred and the rate at which survival is either increasing or decreasing. Other potential metrics include the restricted mean survival time (RMST) and median survival time (41). RMST is similar to the LEL, but it is defined only up to a certain time point post-diagnosis (e.g., 5 or 10 years). Median survival is useful for assessing the average survival time, but it is unable to be calculated for cancer sites with very good prognosis. Both RMST and median survival do not require modeling or extrapolating beyond the range of the observed data.

In contrast, LEL compares the life expectancy of cancer patients with the life expectancy of the general population and can quantify the number of years of life that individuals with cancer may lose due to their disease. The number of life years lost due to cancer is an interpretable and intuitive measure and covers the entire future life course of cancer patients. It is gaining traction as a cancer control measure, but it comes at the cost of requiring modeling and extrapolation assumptions.

The major strength of this study is the use of population-based registry data. We included all patients with a breast cancer, CML, CRC, DLBCL, lung cancer, or melanoma diagnosis between 1975 and 2018 in a collection of nine cancer registries representing approximately 9% of the US. This resulted in over 2 million patients being included in our analysis, with the smallest group of patients considered in a single model being over 11,000 for CML. By using registry data, we can understand the impacts of treatment and screening on life expectancy as they disseminate into the population. Although life expectancy and LEL are not new measures, this is the first time that these measures are reported in the US covering a long period of time. While trends in cancer survival measures provide more relevant information to cancer patients and clinicians and can be better tailored to an individual, life expectancies and LEL measures provide a related public health perspective that better describes the impact of cancer control on a cancer patient’s life expectancy.

In conclusion, our study demonstrated that patients with breast cancer, CML, CRC, DLBCL, lung cancer, and melanoma are living longer since 1975 and that their life expectancy is getting closer to the life expectancy of non-cancer patients. In particular, large increases in life expectancy were observed between 1990 and 2010 for breast, CML, and DLBCL patients. For CRC, improvements in life expectancy largely occurred before 2000. For lung cancer, life expectancy is still substantially lower for cancer patients, though there have been improvements since 2005. For melanoma, life expectancy in the most recent years has nearly equaled that of the general population without cancer. The improvements in life expectancy observed among cancer patients over the past three decades may largely be the result of improved general population life expectancy, treatment advancements, public awareness, and improved screening. Life expectancy and the loss in expectation of life are useful statistics that inform patients, clinicians, and the general population about the impact of cancer control efforts and should therefore be included in national cancer progress reports and publications.