Background

In 2022, there were more than 18 million cancer survivors in the United States,¹ and this number is expected to increase to 22 million by 2030 due to population aging and improvements in early cancer detection and effective treatments.² Cancer diagnosis and treatment can have long-term adverse effects, including physical problems, such as cognitive issues and hearing loss,^3,4 and emotional difficulties, such as anxiety and depression.^5–7 Collectively, these adverse effects can negatively affect survivors’ social relationships and contribute to loneliness,^8,9 a subjective feeling of social isolation.

Loneliness is associated with higher morbidity and mortality risk in the general population^10–14; its importance as a medical issue was underscored by the widespread social isolation imposed by the COVID-19 pandemic.¹⁵ A study of healthy individuals in Sweden found that those who self-reported having a high level of loneliness had a 27% higher mortality risk compared with individuals who did not.¹² Another study of an elderly population in the United States found that loneliness was associated with higher mortality risks.¹⁰ Although informative, most of these studies reported on the associations of loneliness and adverse health outcomes, rather than causality, due to the observational nature of the study designs.

Cancer survivors can experience unique sources of loneliness associated with their diagnosis and treatment history. Families and friends might not understand their disease-related concerns, they may hesitate to share their cancer-related fears and worries, and they may not have received the support they expected and/or needed during their cancer experience, all of which could potentially heighten feelings of social isolation.^16,17 Understanding the prevalence and health implications of loneliness among cancer survivors can have important clinical implications for both cancer care and primary care, given that screening and interventions may require coordination between physicians and specialties, and primary care providers can help provide continuity of care, routine screenings, and emotional support to patients.¹⁸ Previous studies found that loneliness was associated with adverse health outcomes, such as cognitive problems, mental health problems, and pain,^19–21 but few studies examined the associations of loneliness and mortality risk among cancer survivors. One study found that socially isolated people (measured by living arrangements) were at a higher risk of cancer-related mortality.²² Although social isolation and loneliness are closely related, they are distinct phenomena—loneliness reflects an individual’s subjective feeling, and people with high levels of social engagement might still feel lonely. Therefore, in this study, we examined whether loneliness was associated with mortality risk in a large, population-based longitudinal cohort of cancer survivors.

Methods

Data were obtained from the 2008–2018 Health and Retirement Study (HRS), a nationally representative longitudinal panel survey conducted biennially with questions about health, employment, and psychosocial well-being among individuals aged ≥50 years.²³ Each respondent was surveyed multiple times, and loneliness and other time-varying variables were measured repeatedly throughout the years. Follow-up for vital status was obtained through December 31, 2020. Cause of death was not available in this dataset, and thus all-cause mortality, which reflects mortality related to overall health status and access to all aspects of care, was used as the outcome in this analysis. All-cause mortality is an especially relevant outcome for long-term cancer survivors²⁴ because cancer and its treatment can increase the risk of other chronic conditions (eg, cardiomyopathy, congestive heart failure) that can also affect their mortality risk.²⁵

Loneliness measures were collected by the HRS since 2008,²⁶ and subsequently measured in surveys every 4 years. Cancer survivors were identified by an affirmative response to a question about prior cancer or malignant tumor diagnosis, excluding minor skin cancer (person-years, 6,306). We excluded cancer survivors with missing information on any loneliness measure and with missing data on race/ethnicity, educational attainment, marital status, or depression status (9.4% of the sample). A total of 3,371 cancer survivors with 5,711 person-years of observation were included in the final analytic sample (see Table S1 in the supplementary materials, available online with this article).

Loneliness was measured using an 11-item abbreviated version of the UCLA Loneliness Scale (Version 3), a validated and widely used loneliness measure²⁷ (see Supplementary Table S2 for the exact wording of questions). Example items include lacking companionship and feeling isolated from others. Response options were “often,” “some of the time,” and “hardly ever or never.” The Cronbach’s alpha for loneliness measure was 0.88 in our sample, indicating relatively high reliability of these measures. For each question, scores were assigned based on loneliness level: 1 for least lonely, 2 for moderately lonely, and 3 for the loneliest option. Items were summed to create total loneliness scores for each individual, which were then categorized into 4 levels: 11–12 (low/no loneliness), 13–15 (mild loneliness), 16–19 (moderate loneliness), and 20–33 (severe loneliness) based on sample distribution and results from spline analyses (Supplementary Table S3 and Figure S1). Respondent demographic characteristics included sex, race/ethnicity, marital status, educational attainment, number of health conditions other than cancer, years since cancer diagnosis, and depression status measured by whether the person felt depressed in the past year.

Distributions of loneliness scores were compared by sociodemographic characteristics. Kaplan-Meier curves were used to visualize survival rates by loneliness level. To examine the association of loneliness and survival among cancer survivors while accounting for any changes in loneliness during the follow-up period, we used time-varying univariate and multivariable Cox proportional hazard models with age as the time scale²⁸ after confirming proportional hazards with log-log Kaplan-Meier curves. Sex, marital status, educational attainment, number of health conditions other than cancer, years since cancer diagnosis, and depression status were included as covariates in the multivariable models because earlier studies showed that these variables were linked with both loneliness level and mortality risk.^29–38 Because there is little consensus on the best approach for measuring loneliness, we conducted a sensitivity analysis with the loneliness score as a continuous variable rather than categorized as quartiles.^39,40 To address concerns about biases introduced by missing data on our exposure measure, we used the average loneliness score to approximate total score for survivors who answered some, but not all, loneliness questions in another sensitivity analysis. This approach reduced the percentage of respondents excluded due to missing data from 9.4% to 4.8%. We explored effect modification of the association of loneliness and survival by key factors, including sex, race/ethnicity, marital status, educational attainment, number of health conditions other than cancer, years since cancer diagnosis, and depression status, with an interaction term of loneliness score in separate multivariable models. We also performed a sensitivity analysis where we censored follow-up time on February 29, 2020, to exclude the potential effects of the COVID-19 pandemic on the study findings. Statistical analysis was performed in SAS 9.4 (SAS Institute Inc.). All analyses used sample weights to account for the complex HRS design. Statistical comparisons were 2-sided, and significance was defined as P<.05.

Results

A total of 3,371 cancer survivors with 5,711 person-years of observation were included in this study. Most of them were long-term survivors, diagnosed >2 years prior to the survey. Of them, 1,382 (weighted percentage: 24.5%), 1,428 (24.6%), 1,391 (23.5%), and 1,510 (27.4%) reported a loneliness score of 11–12 (low/no loneliness), 13–15 (mild loneliness), 16–19 (moderate loneliness), and 20–33 (severe loneliness), respectively. The variables of male sex, non-White race/ethnicity, unmarried status, less education, more health conditions (other than cancer), and feeling depressed in the past year were linked to a higher probability of being lonelier (Table 1).

Table 1.

Characteristics of Cancer Survivors by Loneliness Score

View Table

There were 686 deaths among the sample during the study period. Higher loneliness was associated with worse survival among cancer survivors (Figure 1). In adjusted analysis (weighted), compared with the low/no loneliness group, survivors with higher loneliness scores had higher mortality risk, with the highest adjusted hazard ratio (aHR) observed among the group with severe loneliness (aHR, 1.19 [95% CI, 0.86–1.63] for mild loneliness; 1.41 [95% CI, 1.01–1.96] for moderate loneliness; and 1.67 [95% CI, 1.25–2.23] for severe loneliness), following a dose–response association (P=.004) (Table 2). The aHR was 1.03 (95% CI, 1.01–1.06) when loneliness score was included in multivariable models as a continuous variable.

View Full Size

Kaplan-Meier survival curve by loneliness score among cancer survivors.

Citation: Journal of the National Comprehensive Cancer Network 22, 4; 10.6004/jnccn.2023.7114

Table 2.

Association of Loneliness and Survival Among Cancer Survivors^a

View Table

Missing data were nonrandom (see Supplementary Table S4), but findings were largely unchanged when we used the average loneliness score to approximate the total score for survivors responding to some, but not all, of the loneliness questions (Supplementary Table S5). We observed similar patterns of loneliness and mortality risk when stratifying the analysis by key factors, such as race/ethnicity, years since cancer diagnosis, and depression status (Supplementary Table S6). The findings were unchanged after censoring follow-up time before the start of the COVID-19 pandemic in the United States (Supplementary Table S7).

Discussion

In this nationally representative cohort of cancer survivors in the United States, higher levels of loneliness were associated with higher mortality risk, with the highest mortality risk observed among the loneliest group, following a dose–response pattern. Overall, these findings extend the growing body of research linking loneliness to important health outcomes, including morbidity and mortality, in the context of cancer. Our findings suggest that screening patients for loneliness should be a component of survivorship care, and of concern to primary care and cancer clinicians in light of the potential adverse associations of loneliness and mortality risk. This study is especially timely given the recent Advisory on the Healing Effects of Social Connection and Community by the US Surgeon General, which highlighted the adverse effects of loneliness for individuals and for public health,⁴¹ and the increased loneliness among people with cancer during the COVID-19 pandemic.⁴² Ongoing evaluations of any lasting effects of the widespread social isolation that occurred during the pandemic are warranted.

This study expanded on previous studies reporting negative associations of loneliness and health outcomes, such as cognitive problems, mental health problems, and pain,^19–21 and found that loneliness was also associated with higher mortality risks among cancer survivors. As described in the Surgeon General’s Advisory, prior research has suggested that loneliness may be linked to worse survival through multiple mechanisms or pathways, such as increased risk of experiencing negative emotions (eg, hostility, stress, anxiety), increased engagement in unhealthy behaviors (eg, smoking, alcohol abuse, less physical activity), and/or through physiologic pathways (eg, immune system disorders).⁴³ For instance, people with cancer might feel their families and friends did not support or share their cancer-related concerns, and/or may have unrealistic expectations of emotional and instrumental support from other people,¹⁷ which can add to negative emotions. Additionally, lonelier cancer survivors might not receive the practical and emotional support they need for their symptoms, which could then impose greater health burdens, with earlier studies finding that loneliness was associated with symptoms such as fatigue, pain, and low sleep quality among cancer survivors, even many years after diagnosis.^21,44 Some cancer survivors may experience socially constraining behaviors, such as criticism, when sharing their feelings, which has been associated with poor health.¹⁶ Unfortunately, we were not able to examine these specific potential mechanisms in the present analysis, but this avenue of research will be important for future studies.

Our study findings have clear clinical implications. For instance, our findings underscore the importance of incorporating loneliness screening as a routine part of cancer survivorship care. Health care providers should consider adding loneliness assessment tools to their standard practice, and these tools should be employed at multiple time points, from initial diagnosis throughout survivorship care, because loneliness may change throughout the cancer journey. Early detection of loneliness is crucial because it may prevent the escalation and complications of loneliness among cancer survivors. In addition, identification of survivors experiencing loneliness can inform needs for intervention, such as mental health counseling, community support, social network involvement, and integrating loneliness intervention into cancer treatment and cancer survivorship care. Future studies should focus on the optimal screening tools for loneliness and examine the effectiveness of loneliness screening and interventions.

Our findings also highlight the importance of care coordination by cancer care clinicians and other specialty and primary care providers to ensure that cancer survivors receive timely screening and effective intervention for loneliness.¹⁸ More specifically, primary care providers can provide comprehensive routine assessment of loneliness among cancer survivors and help them manage some of the physical health consequences of loneliness, such as sleep problems and chronic stress, by prescribing appropriate medications or lifestyle modifications. Oncologists can also provide loneliness screening during cancer treatment and provide education about the emotional challenges that often accompany a cancer diagnosis and treatment, and some of the unique needs and challenges of cancer survivors. Referrals to psychologists can provide more specialized assessment and tailored counseling. In addition, a health care coordinator or nurse navigator can help facilitate communication and coordinate care among these specialists to ensure timely surveillance and intervention with regard to the loneliness status among patients with cancer and survivors. In addition to clinical interventions, the Surgeon General’s Advisory highlighted the importance of relevant training for providers and the need for research on loneliness to help increase public awareness.⁴¹ Identification of the optimal approaches to coordinate and collaborate among different specialties to address patient loneliness and any adverse health outcomes is needed.

This study has limitations, including the use of an abbreviated measure to assess loneliness and the use of self-reported data. Cancer-related information, such as cancer type and number of diagnoses, stage at diagnosis, and whether respondents were receiving cancer treatment at the time of the survey, was not systematically collected in the HRS, nor did it include cause of death. Furthermore, the sample of cancer survivors was not sufficient to explore the modification effects of all key factors. Meanwhile, approximately 5% of respondents were excluded due to missing key variables, and we did not use multiple imputation to impute the missing values because they were not missing at random. Despite these limitations, this study contributes to the growing literature demonstrating the association of loneliness and compromised health outcomes among cancer survivors, who face unique experiences of loneliness due to the nature of their disease and treatment. Future research should determine the optimal methods for the assessment and management of loneliness in cancer survivors to inform interventions to reduce associations with adverse health outcomes.

Conclusions

Using national longitudinal data, we found that loneliness was associated with higher mortality risks among cancer survivors. Programs to enhance coordination and collaboration among providers across different specialties and integrate loneliness screening into primary care and oncology care are warranted.