eTable 1. Data Sources
eTable 2. Cohort Identification Strategy
eTable 3. Covariate Definitions
eTable 4. Surgical Procedure Intensity
eTable 5. Characteristics of Study Cohort by Cancer Type
eTable 6. Cumulative Incidence at 1-year, 3-year, and 5-year Timepoints
eFigure 1. Cohort Selection Flowchart
eFigure 2. Cumulative Incidence of Cancer, Non-Cancer, and All-Cause Death by Cancer Type
eFigure 3. Cumulative Incidence of Cancer, Non-Cancer, and All-Cause Death Stratified by Gastrointestinal Cancer Subtypes
eFigure 4. Cumulative Incidence of Cancer, Non-Cancer, and All-Cause Death Stratified by Genitourinary Cancer Subtypes
Nonauthor Collaborators. Members of the Recovery After Surgical Therapy for Older Adults Research–Cancer (RESTORE-Cancer) Group
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Chesney TR, Coburn N, Mahar AL, et al. All-Cause and Cancer-Specific Death of Older Adults Following Surgery for Cancer. JAMA Surg. 2021;156(7):e211425. doi:10.1001/jamasurg.2021.1425
How long do older adults live after receiving surgical treatment for cancer and do they die of cancer or noncancer causes?
In this population-based cohort study of 82 037 patients 70 years or older, 5 years after resection for cancer, 20% of patients died of cancer and 16% died of other causes. Low-risk cancer types, advancing age, and frailty were associated with death from noncancer causes 3 years after surgery.
For older adults selected for cancer surgery, the relative burden of cancer deaths exceeds death from other causes, even in more vulnerable patients.
Cancer care has inherent complexities in older adults, including balancing risks of cancer and noncancer death. A poor understanding of cause-specific outcomes may lead to overtreatment and undertreatment.
To examine all-cause and cancer-specific death throughout 5 years for older adults after cancer resection.
Design, Setting, and Participants
This population-based cohort study was conducted in Ontario, Canada, using the administrative databases stored at ICES (formerly the Institute for Clinical Evaluative Sciences). All adults 70 years or older who underwent resection for a new diagnosis of cancer between January 1, 2007, and December 31, 2017, were included. Patients were followed up until death or censored at date of last contact of December 31, 2018.
Main Outcome and Measures
Using a competing risks approach, the cumulative incidence of cancer and noncancer death was estimated and stratified by important prognostic factors. Multivariable subdistribution hazard models were fit to explore prognostic factors.
Of 82 037 older adults who underwent surgery (all older than 70 years; 52 119 [63.5%] female), 16 900 of 34 044 deaths (49.6%) were cancer related at a median (interquartile range) follow-up of 46 (23-80) months. At 5 years, estimated cumulative incidence of cancer death (20.7%; 95% CI, 20.4%-21.0%) exceeded noncancer death (16.5%; 95% CI, 16.2%-16.8%) among all patients. However, noncancer deaths exceeded cancer deaths starting at 3 years after surgery in breast, prostate, and melanoma skin cancers, patients older than 85 years, and those with frailty. Cancer type, advancing age, and frailty were independently associated with cause-specific death.
Conclusions and Relevance
At the population level, the relative burden of cancer deaths exceeds noncancer deaths for older adults selected for surgery. No subgroup had a higher burden of noncancer death early after surgery, even in more vulnerable patients. This cause-specific overall prognosis information should be used for patient counseling, to assess patterns of over- or undertreatment in older adults with cancer at the system level, and to guide targets for system-level improvements to refine selection criteria and perioperative care pathways for older adults with cancer.
Within a decade, 70% of new cancers will be diagnosed in older adults.1-4 Heterogenous health statuses and values generate nuances in cancer care with older adults.5-12 While cancer is the predominant risk for death for younger patients with cancer, the risk of noncancer death is an important consideration in older adults with more prevalent comorbidities, frailty, decreased underlying life expectancy, and potentially different tumor biology.5,13-17 It is a concern that older adults with cancer may have a high burden of competing risks, and conventional cancer treatment will result in overtreatment.18,19 As such, the risk of cancer and noncancer death must frame cancer treatment choices, but these risks are not known for older adults undergoing surgery.
Poor understanding of the relative risks of cancer and noncancer deaths may lead to over- and undertreatment.18 Undertreated older adults die earlier of cancer, and overtreated older adults are exposed to nonbeneficial intensive treatments and unnecessary treatment-related toxicity with reduced quality of life.20-28 Recognizing this, a joint statement by the European Organisation for Research and Treatment of Cancer, the Alliance for Clinical Trials in Oncology, and the International Society of Geriatric Oncology recommends reporting cancer and noncancer deaths in conjunction with all-cause death and stratifying results by frailty status.29
For older adults with cancer, cause-specific survival after surgical resection or stratified by frailty status has not been reported to our knowledge, and competing risks are rarely considered.30-33 Applying prognostic data for younger patients is inadequate. Understanding older adults’ cancer and noncancer prognosis within current cancer care strategies is critical to patient counseling, health care system planning, and research to evaluate patterns of care, selection of treatments, improve cancer care strategies, and develop prognostic models.34-38
Given the concern for competing risks of noncancer death and overtreatment in older adults and the limited cause-specific prognostic data currently available, we aimed to examine the prognosis of older adults undergoing cancer surgery, including cancer, noncancer, and all-cause deaths over 5 years after surgery and to explore prognostic factors associated with cause-specific death. This study was conducted in accordance with recommendations for fundamental prognosis research and geriatric oncology research.29,34
We designed a population-based retrospective cohort study of adults 70 years or older with a new diagnosis of solid malignant neoplasm who underwent resection. We assembled the cohort using linked health administrative databases using unique encoded identifiers for each patient housed at ICES (formerly the Institute for Clinical Evaluative Sciences) for the province of Ontario, Canada. Race/ethnicity are not collected in the ICES data set. The Ontario Cancer Registry includes all patients with a cancer diagnosis in Ontario.39,40 The Registered Persons Database contains vital status and demographic data.41 Data on receipt of health care services is captured in several additional link data sets described in eTable 1 in Supplement 1. The Ontario population has universally accessible publicly funded health care.42 The study was approved by the Sunnybrook Health Sciences Centre Research Ethics Board (Toronto, Ontario, Canada). This report followed the PROGnosis RESearch Strategy (PROGRESS) and the REporting of studies Conducted using Observational Routinely collected Data (RECORD) statements.34,43 Two caregivers and family members of older adults who underwent major cancer surgery were involved in developing the research question, defining outcomes measures, and interpreting results. Patient consent was not obtained because administrative anonymized data were used.
We identified individuals 70 years or older with a new diagnosis of oropharyngeal, melanoma, breast, esophageal, gastrointestinal, colorectal, hepatobiliary, pancreatic, genitourinary, gynecologic, and bronchopulmonary cancer between January 1, 2007, and December 31, 2017, using International Classification of Diseases for Oncology, Third Edition codes (eTable 2 in Supplement 1).44 We included patients who underwent resection 90 to 180 days after cancer diagnosis (eTable 2 in Supplement 1).
Individuals were excluded if they had date of death preceding date of diagnosis, had previous cancer diagnosis within 5 years before the index cancer diagnosis, had 2 or more cancer diagnoses recorded on index diagnosis date, were admitted to long-term care prior to surgery, had missing date of death, or follow-up less than 180 days.
Our outcome was death following surgery, and cause-specific death was classified as cancer death or noncancer death. Time to death was calculated from date of surgery to date of death. Cancer death was defined as International Classification of Diseases, Ninth Revision codes 140-239 and noncancer death as other codes in the Ontario Registrar General database (eTable 1 in Supplement 1). The primary cause of death was defined as the antecedent cause of death when available or the immediate cause of death when antecedent causes were not captured. High agreement in cause of death has been reported between the Ontario Registrar General and clinical follow-up in a prospective cohort of cancer patients.45 Participants were followed up until date of death, date of last clinical contact, or December 31, 2018, allowing for an opportunity of a minimum of 12 months of follow-up for each patient.
Covariates are detailed in eTable 3 in Supplement 1. We quantified comorbidity burden using the Adjusted Clinical Groups system with a cutoff of 10 indicating high burden.46-48 We used a consensus classification of high and low surgical procedure intensity (eTable 4 in Supplement 1).49 We identified receipt of perioperative chemotherapy or radiotherapy within 180 days before or after surgery using claims data. We identified preoperative frailty using the Johns Hopkins Adjusted Clinical Groups frailty marker.50 This frailty marker uses 12 clusters of frailty-defining diagnoses within important geriatric domains and has been externally validated against Comprehensive Geriatric Assessment and the Vulnerable Elders Survey.51-53 We determined rural residency using the rurality index of Ontario based on postal codes of primary residences.54 We captured socioeconomic status using the Material Deprivation Index, a composite index of ability to afford goods and activities, categorized into quintiles.55-57
Using time to death, we estimated the cumulative incidence of death throughout 5 years after cancer surgery using the cumulative incidence function for all-cause death and for cancer death treating noncancer death as a competing risk. We reported cumulative incidence as percentages with 95% CIs at 1-, 3-, and 5-year points. Five years was selected as a conventional time horizon in cancer care; however, given that for older adults shorter times may be relevant, we also provide survival curves and estimates reported at shorter intervals. Cumulative incidence was reported for the entire cohort overall and further stratified by cancer type, age groups, and preoperative frailty status.
To explore the association of potential prognostic factors on the relative incidence of cancer and noncancer death, we fit multivariable subdistribution hazard models for cancer and noncancer death.58,59 We selected potential prognostic factors a priori based on clinical relevance and literature review including age, sex, cancer type, material deprivation quintile, rural residence, time of diagnosis, receipt of perioperative therapy, and preoperative frailty.60,61 To assess if the association of frailty cancer and noncancer death with outcomes differed by age, we included an interaction term between age and frailty status. We reported the subdistribution hazard ratios with 95% CI.
Data were missing for rural residency in 632 patients (0.8%) and material deprivation in 514 patients (0.6%). As such, we performed a complete-case analysis for multivariable analysis. Statistical significance was set at P ≤ .05. All analyses were conducted using SAS Enterprise Guide 7.1 (SAS Institute). Data were analyzed from October to December 2019.
We identified 82 037 older adults undergoing surgical resection for cancer (eFigure 1 in Supplement 1). Characteristics of included patients are presented in Table 1 and stratified by cancer type in eTable 5 in Supplement 1. The most common cancer types were breast cancer (22 811 patients [27.8%]) and gastrointestinal cancer (32 036 patients [39.1%]). Of 34 044 deaths, 49.6% (n = 16 900) were cancer related, with a median (interquartile range) follow-up of 46 (23-80) months. Postoperative mortality within 90 days of surgery accounted for 11.2% (n = 3831) of deaths.
The cumulative incidence of death for the whole cohort is displayed in Figure 1. Throughout 5 years, the cumulative incidence of death from cancer was greater than death from noncancer causes. Estimated cumulative incidence of cancer death was 8.2% (95% CI, 8.4%-8.6%) at 1 year, 16.4% (95% CI, 16.2%-16.7%) at 3 years, and 20.7% (95% CI, 20.4%-21.0%) at 5 years after surgery. Estimated cumulative incidence of noncancer death was 5.3% (95% CI, 5.1%-5.5%) at 1 year, 11.9% (95% CI, 11.6%-12.2%) at 3 years, and 18.1% (95% CI, 17.8%-18.5%) at 5 years.
Across cancer types, the cumulative incidence of all-cause death varied from 22.0% (95% CI, 21.4%-22.6%) to 50.8% (95% CI, 47.2%-54.4%) at 5 years (eFigure 2 and eTable 6 in Supplement 1). Differences in the incidence of death between cancer types were driven mostly by cancer deaths. The lowest 5-year cumulative incidence of cancer death occurred in breast cancer at 9.1% (95% CI, 8.6%-9.5%) and the highest in oropharyngeal cancers at 29.1% (95% CI, 25.9%-32.3%). The incidence of death from cancer was greater than death from noncancer causes across most cancer types. The only exceptions were breast cancer and melanoma for which noncancer deaths were higher than cancer deaths by 5 years. Across gastrointestinal cancers, the cumulative incidence of noncancer death was similar; however, cancer deaths varied, with highest cumulative incidence in in hepato-pancreatico-biliary and esophageal cancers (eFigure 3 in Supplement 1). Across genitourinary cancers, bladder cancer had the greatest cumulative incidence of cancer death (eFigure 4 in Supplement 1).
When stratified by age group, the cumulative incidence of all-cause death increased with advancing age, from 26.4% (95% CI, 25.8%-26.9%) to 57.0% (95% CI, 56.0%-58.1%) at 5 years (Figure 2; eTable 6 in Supplement 1), mostly owing to increases in noncancer deaths as patients got older. Only in patients 85 years and older did the incidence of death from noncancer causes became greater death from cancer throughout the 5 years after surgery.
Preoperative frailty also was associated with the risk of death and patterns in cause of death. The cumulative incidence of all-cause death at 5 years was 34.9% (95% CI, 34.9%-35.3%) for those without frailty and 56.6% (95% CI, 55.2%-58.0%) for those with frailty (Figure 3; eTable 6 in Supplement 1). In patients with preoperative frailty, the incidence of death from noncancer causes became greater than death from cancer starting 3 years after surgery.
Preoperative patient and disease factors associated with the cumulative incidence of cancer deaths and of noncancer deaths were examined in separate multivariable subdistribution hazard models (Table 2). Advancing age was associated with cumulative incidence of cancer death for individuals aged 75 to 79 years (subdistribution hazard ratio, 1.33 [95% CI, 1.28-1.38]) to individuals 85 years and older (subdistribution hazard ratio, 2.06 [95% CI, 1.96-2.16]), compared with those aged 70 to 74 years). A similar association was observed in noncancer deaths but with greater magnitude of the effect estimates than for cancer death. Preoperative frailty was associated with increased cumulative incidence of both cancer and noncancer death. The magnitude of the association between frailty and noncancer death increased with age (interaction P < .001), and there was no frailty-age interaction observed for cancer death (interaction P = .09). Perioperative therapy was associated with increased cumulative incidence of cancer death and decreased incidence of noncancer death. Compared with breast cancer, all cancer types were strongly associated with an increased cumulative incidence of cancer death with subdistribution hazard ratios ranging from 2.34 (95% CI, 2.14-2.57) for melanoma to 4.69 (95% CI, 4.40-5.01) for bronchopulmonary cancers; this was not consistently observed for noncancer deaths.
This is the first study, to our knowledge, to describe the population-level overall prognosis of older adults after cancer surgery including an analysis of cancer and noncancer deaths. At 5 years, older adults selected for surgery had a 37% cumulative incidence of all-cause death, 21% of cancer deaths, and 16% of noncancer deaths. The relative burden of cancer compared with noncancer deaths varied by cancer type, age group, and preoperative frailty status. Noncancer deaths became greater than cancer deaths only in patients who underwent a surgical procedure for breast cancer or melanoma, who were 85 years and older, and with preoperative frailty.
This study contributes important new insight into cancer- and noncancer-specific death.31,32,62 A systematic review identified only 1 study reporting cause-specific end points using a competing risks approach in geriatric oncology.18 Another study reanalyzed data from a clinical trial of adjuvant endocrine therapy in breast cancer to evaluate cause-specific death.63,64 Most population-based studies have not reported outcomes for older adults, not reported cause-specific death, or not used a competing risks approach.31-33,60,61,65-71 The few studies using a competing risks approach were focused only on breast cancer, did not focus on cancer surgery, and did not address preoperative frailty status, which is a key concept in older adults care.30,72,73 Our study addresses these gaps in the literature by reporting on long-term survival outcomes for older adults undergoing cancer surgery stratified by cancer type, age, and preoperative frailty status and accounting for competing risks to avoid overestimation of the risk of cancer death in older adults.13,18,58,59 Previously, overall prognostic estimates for patients, clinicians, and health care systems planning were only available by applying data from younger patients.
Our results indicate that the incidence of death from cancer was greater than death from noncancer causes overall. Thus, even with cancer surgery, cancer remains the major driver of death, and there is no evidence of overtreatment. In certain circumstances, cancer death did not exceed noncancer death such as with breast, melanoma, and prostate cancers. If treated with surgery, patients with those cancers are less likely to die of cancer than from other causes. This difference can be related to lower-intensity surgery, highly effective cancer treatment, favorable cancer biology, low noncancer life expectancy, or a combination of these factors. Similarly, the risk of noncancer death was greater in patients 85 years and older and in those with preoperative frailty but only starting at 3 years after surgery. This suggests that cancer surgery is not overtreatment in older adults selected for surgery, even in vulnerable subgroups because cancer death still represents an immediate greater threat of mortality. This is important information considering observed patterns of undertreatment in older adults.23-26,28 It is important to personalize estimates of noncancer death to avoid foregoing surgery in those who would benefit; lack of cancer treatment would further increase the risk of early cancer-specific death.74-76
Our data can be directly used to provide overall prognosis estimates specific to older adult populations with cancer, and this can be balanced against underlying life expectancy estimates without cancer using tools such as ePrognosis.77 These data can be combined with geriatric-specific care processes, such as integrated geriatric oncology pathways, comprehensive geriatric assessment, prehabilitation, and geriatric comanagement, which are still rare within surgical cancer care, to further support decision-making, mitigate risks, and support cancer treatment, particularly those at higher risk.74,78-84
There are study limitations. Because of the retrospective design and administrative data sources, the data used were not specifically collected for the purposes of the research question. In particular, our results must be interpreted in the context of patients selected for surgical treatment. For example, only 8% of our cohort had preoperative frailty compared with 10% to 20% in the general older adult population, pointing toward the expected selection bias for cancer surgery.85-87 We could not decipher the rationale for not undergoing a surgical procedure using administrative data, nor could we determine if surgery was offered to some patients but declined. Because comparisons with older adults without cancer, older adults with cancer not having surgery, or younger surgery patients with cancer bear selection bias and are not informative for decision-making in the individual older adults selected for cancer surgery, such comparative analyses were not undertaken. We focused on describing overall prognosis of patients treated with surgery rather than an assessment of the causal effect of surgery compared with no surgery. Although high agreement is documented between our registries and primary sources in cause of death ascertainment, prevalent comorbidities could affect the accuracy of the listed cause of death on death certificates and lead to misclassification, even more so if misclassification is related to age.45 This would lead to overestimating noncancer deaths; therefore the observations of patterns of cancer vs noncancer deaths would remain and the conclusions would stand. Finally, cancer stage information is not reliably available in our databases, so we cannot comment on cause-specific mortality by stage. Despite these limitations, our results provide important knowledge regarding the burden of cancer- and noncancer-related deaths, which demonstrates that cancer surgery is not overtreatment for older adults currently selected for surgery and highlight the importance of considering multiple patient-centered factors for individualized decision-making and need to avoid de-escalation of care and risk of undertreatment in older adults.
The population-based design of this study is a strength that allowed for a real-world assessment of the prognosis of older adults undergoing cancer surgery. These data represent outcomes within a universal health care system avoiding bias by insurance status–limited access. We used high-quality linked administrative databases and definitions with known accuracy and reliability including comprehensive availability of cause of death, limited missing data, and measurement error. This allowed for an accurate assessment of system-level cancer care performance regarding postoperative cancer and noncancer deaths.
In this study, we reported novel cause-specific death throughout 5 years after cancer surgery in older adults accounting for competing risks. Overall, the incidence of death from cancer exceeded that from noncancer causes. In a few subgroups, the incidence of noncancer death was greater: breast, prostate, and melanoma skin cancer; patients 85 years and older; or preoperative frailty after 3 years. For most patients, cancer surgery should not be avoided due to concern about overtreatment related to competing risks resulting in noncancer death. Clinicians and patients can directly use this knowledge about the association of surgery for cancer with outcomes in shared decision-making, including data stratified by cancer type, age, and frailty. For patients with a higher risk of noncancer death (breast, melanoma skin, and prostate cancers; age >85 years; and frailty) further interventions should be considered to mitigate risks of noncancer death, such as integrated geriatric oncology pathways, comprehensive geriatric assessment, prehabilitation, and geriatric comanagement. Future work should focus on integrating quality of life and functional outcomes with cause-specific death information to support comprehensive patient counseling and using the current findings to assist in building prognostic models for more individualized risk prediction.
Corresponding Author: Tyler R. Chesney, MD, MSc, St Michael’s Hospital, 30 Bond St, Cardinal Carter North, Room 16-064, Toronto, ON M5B 1W8, Canada (firstname.lastname@example.org).
Accepted for Publication: February 14, 2021.
Published Online: May 12, 2021. doi:10.1001/jamasurg.2021.1425
Author Contributions: Dr Hallet and Mr Zhao had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Chesney, Coburn, Davis, Zuk, Hallet.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Chesney, Hallet.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Chesney, Mahar, Zhao, Hallet.
Obtained funding: Chesney, Mahar, Hsu, Hallet.
Administrative, technical, or material support: Chesney, Coburn, Zuk, Hallet.
Supervision: Hsu, Haas, Hallet.
Conflict of Interest Disclosures: Dr Coburn receives salary support from Cancer Care Ontario as lead for patient-reported outcomes. Dr Hallet reports personal fees (speaker honoraria) from Ipsen Biopharmaceuticals Canada, Advanced Accelerator Applications, and Novartis Oncology outside the submitted work. No other disclosures were reported.
Funding/Support: This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC).
Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Group Information: The Members of the Recovery After Surgical Therapy for Older Adults Research–Cancer (RESTORE-Cancer) Group are listed in Supplement 2.
Disclaimer: No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. However, the analyses, conclusions, opinions, and statements expressed herein are those of the author and not necessarily those of the Canadian Institute for Health Information. The opinions, results, view, and conclusions reported in this article are those of the authors and do not necessarily reflect those of Cancer Care Ontario (CCO). No endorsement by CCO is intended or should be inferred.
Additional Contributions: We thank the patient and family advisors, Grace Paladino and Pietro Galluzzo, for their contributions to this work. No compensation was received