Mechanisms and mediation in survival analysis: towards an integrated analytical framework

Depression and Anxiety Mediate the Relationship between Discrimination and Well-Being in a Sample of Latinx Adults with Type 2 Diabetes: Results from a Dual Mediation Analysis

Latinxs experience greater risk for type 2 diabetes, discrimination, and poor mental health. The pathways linking these factors, however, are not well understood. This study tested whether depression and anxiety mediated the relationship between discrimination and well-being. Bootstrapped mediation tests were conducted using a sample of Latinx adults with type 2 diabetes (n = 121) and regression models adjusted for demographic and health covariates. Depression and anxiety fully and jointly mediated the effect of discrimination on well-being; everyday discrimination was linked to elevated symptoms of depression and anxiety which were, in turn, independently linked to reduced emotional well-being. Moreover, the effect size for the anxiety pathway (β=-0.13) was 60% larger than for depression (β=-0.08). Dual mediation suggests depression, and especially anxiety, may be important targets for interventions seeking to mitigate the deleterious effects of discrimination. Findings have important implications for psychotherapeutic treatments and public health policy.

Association of Social Determinants and Tumor Biology With Racial Disparity in Survival From Early-Stage, Hormone-Dependent Breast Cancer

Importance

Black women with hormone receptor-positive breast cancer experience the greatest racial disparity in survival of all breast cancer subtypes. The relative contributions of social determinants of health and tumor biology to this disparity are uncertain.

Objective

To determine the proportion of the Black-White disparity in breast cancer survival from estrogen receptor (ER)-positive, axillary node-negative breast cancer that is associated with adverse social determinants and high-risk tumor biology.

Design, Setting, and Participants

A retrospective mediation analysis of factors associated with the racial disparity in breast cancer death for cases diagnosed between 2004 and 2015 with follow-up through 2016 was carried out using the Surveillance, Epidemiology, and End Results (SEER) Oncotype registry. The study included women in the SEER-18 registry who were aged 18 years or older at diagnosis of a first primary invasive breast cancer tumor that was axillary node-negative and ER-positive, who were Black (Black), non-Hispanic White (White), and for whom the 21-gene breast recurrence score was available. Data analysis took place between March 4, 2021, and November 15, 2022.

Exposures

Census tract socioeconomic disadvantage, insurance status, tumor characteristics including the recurrence score, and treatment variables.

Main Outcomes and Measures

Death due to breast cancer.

Results

The analysis with 60 137 women (mean [IQR] age 58.1 [50-66] years) included 5648 (9.4%) Black women and 54 489 (90.6%) White women. With a median (IQR) follow-up time of 56 (32-86) months, the age-adjusted hazard ratio (HR) for breast cancer death among Black compared with White women was 1.82 (95% CI, 1.51-2.20). Neighborhood disadvantage and insurance status together mediated 19% of the disparity (mediated HR, 1.62; 95% CI, 1.31-2.00; P < .001) and tumor biological characteristics mediated 20% (mediated HR, 1.56; 95% CI, 1.28-1.90; P < .001). A fully adjusted model that included all covariates accounted for 44% of the racial disparity (mediated HR, 1.38; 95% CI, 1.11-1.71; P < .001). Neighborhood disadvantage mediated 8% of the racial difference in the probability of a high-risk recurrence score (P = .02).

Conclusions and Relevance

In this study, racial differences in social determinants of health and indicators of aggressive tumor biology including a genomic biomarker were equally associated with the survival disparity in early-stage, ER-positive breast cancer among US women. Future research should examine more comprehensive measures of socioecological disadvantage, molecular mechanisms underlying aggressive tumor biology among Black women, and the role of ancestry-related genetic variants.

Association of Prediagnosis Obesity and Postdiagnosis Aspirin With Survival From Stage IV Colorectal Cancer

IMPORTANCE

The potential relationship between obesity and colorectal cancer (CRC) outcome is poorly understood in patients with late-stage disease. Increased body mass index may negate aspirin use for cancer prevention, but its role as a factor on the effectiveness of postdiagnosis aspirin use is unclear.

OBJECTIVE

To evaluate how prediagnosis obesity and postdiagnosis aspirin use may be associated with overall survival in patients with late-stage colorectal cancer.

DESIGN, SETTING, AND PARTICIPANTS

This cross-sectional study used self-reported data from patients with metastatic or treatment-refractory disease who consented to a clinical protocol at MD Anderson Cancer Center, a large US cancer treatment center. Patients were enrolled between 2010 and 2018 and followed up for mortality through July 2020. Analyses were conducted through March 2022.

EXPOSURES

Body mass index in the decade prior to initial diagnosis and regular aspirin use at survey completion.

MAIN OUTCOMES AND MEASURES

Overall survival was measured from stage IV diagnosis until death or last follow-up. Cox proportional hazards models were constructed to estimate associations of prediagnosis obesity and postdiagnosis aspirin use with overall survival.

RESULTS

Of 656 patients included in this analysis, 280 (42.7%) were women, 135 (20.6%) were diagnosed with CRC before age 45 years, 414 (63.1%) were diagnosed between ages 45 and 65 years, and 107 (16.3%) were diagnosed at 65 years or older; 105 patients (16.0%) were Black or Hispanic, and 501 (76.4%) were non-Hispanic White. Controlling for age, sex, race, stage at initial diagnosis, and weight change between prediagnosis and survey date, patients with obesity in the decade prior to CRC diagnosis had significantly higher likelihood of death (hazard ratio, 1.45; 95% CI, 1.11-1.91) compared with those with normal prediagnosis body mass index. Furthermore, only patients with normal prediagnosis body mass index experienced significant survival benefit with postdiagnosis aspirin use (hazard ratio, 0.59; 95% CI, 0.39-0.90).

CONCLUSIONS AND RELEVANCE

In this cross-sectional study, our findings suggest potentially differential tumor development in the long-term physiologic host environment of obesity. Confirmation and further evaluation are needed to determine whether prediagnosis body mass index may be used to estimate the benefit from postdiagnosis aspirin use.

Do Early-Life Social, Behavioral, and Health Exposures Increase Later-Life Arthritis Incidence?

OBJECTIVES

This study investigates direct and indirect influences of childhood social, behavioral, and health exposures on later-life osteoarthritis and rheumatoid arthritis development.

METHODS

Drawing from cumulative inequality theory and six waves of the Health and Retirement Study (2004-2014), we estimate structural equation modeling-based discrete-time survival analysis of the association between six childhood exposure domains and both osteoarthritis and rheumatoid arthritis incidence for men (n = 2720) and women (n = 2974). Using the delta method to test for mediation, we examine indirect effects via selected health-related risks and resources.

RESULTS

Risky adolescent behavior is associated with rheumatoid arthritis incidence for women (h.O.R. = 1.883, 95% C.I. [1.016, 3.490]), whereas several types of childhood exposures are associated with later-life osteoarthritis development for both men and women. Experiencing two or more childhood socioeconomic disadvantages is indirectly associated with osteoarthritis (men: coef. = 0.024, 95% C.I. [0.003, 0.045]; women: coef. = 0.111, 95% C.I. [0.071, 0.150]) and rheumatoid arthritis (men: coef. = 0.037, 95% C.I. [0.000, 0.074]; women: coef. = 0.097, 95% C.I. [0.035, 0.159]) development through adult body mass index.

DISCUSSION

Findings highlight the importance of childhood contexts in understanding the development of later-life osteoarthritis and rheumatoid arthritis.

Structural racism is a mediator of disparities in acute myeloid leukemia outcomes

Non-Hispanic Black (NHB) and Hispanic patients with acute myeloid leukemia (AML) have higher mortality rates than non-Hispanic White (NHW) patients despite more favorable genetics and younger age. A discrete survival analysis was performed on 822 adult patients with AML from 6 urban cancer centers and revealed inferior survival among NHB (hazard ratio [HR] = 1.59; 95% confidence interval [CI]: 1.15, 2.22) and Hispanic (HR = 1.25; 95% CI: 0.88, 1.79) patients compared with NHW patients. A multilevel analysis of disparities was then conducted to investigate the contribution of neighborhood measures of structural racism on racial/ethnic differences in survival. Census tract disadvantage and affluence scores were individually calculated. Mediation analysis of hazard of leukemia death between groups was examined across 6 composite variables: structural racism (census tract disadvantage, affluence, and segregation), tumor biology (European Leukemia Network risk and secondary leukemia), health care access (insurance and clinical trial enrollment), comorbidities, treatment patterns (induction intensity and transplant utilization), and intensive care unit (ICU) admission during induction chemotherapy. Strikingly, census tract measures accounted for nearly all of the NHB-NHW and Hispanic-NHW disparity in leukemia death. Treatment patterns, including induction intensity and allogeneic transplant, and treatment complications, as assessed by ICU admission during induction chemotherapy, were additional mediators of survival disparities in AML. This is the first study to formally test mediators for observed disparities in AML survival and highlights the need to investigate the mechanisms by which structural racism interacts with known prognostic and treatment factors to influence leukemia outcomes.

Respective Mediating Effects of Social Position and Work Environment on the Incidence of Common Cardiovascular Risk Factors

Background

Social position and work environment are highly interrelated and their respective contribution to cardiovascular risk is still debated.

Methods and Results

In a cohort of 20 625 French workers followed for 25 years, discrete‐time survival analysis with reciprocal mediating effects, adjusted for sex, age, and parental history of early coronary heart disease, was performed using Bayesian structural equation modeling to simultaneously investigate the extent to which social position mediates the effect of work environment and, inversely, the extent to which work environment mediates the effect of social position on the incidence of common cardiovascular risk factors. Depending on the factor, social position mediates 2% to 53% of the effect of work environment and work environment mediates 9% to 87% of the effect of social position. The mediation by work environment is larger than that by social position for the incidence of obesity, hypertension, dyslipidemia, diabetes, sleep complaints, and depression (mediation ratios 1.32–41.5, 6.67 when modeling the 6 factors together). In contrast, the mediation by social position is larger than that by work environment for the incidence of nonmoderate alcohol consumption, smoking, and leisure‐time physical inactivity (mediation ratios 0.16–0.69, 0.26 when modeling the 3 factors together).

Conclusions

The incidence of behavioral risk factors seems strongly dependent on social position whereas that of clinical risk factors seems closely related to work environment, suggesting that preventive strategies should be based on education and general practice for the former and on work organization and occupational medicine for the latter.

Childhood Disadvantage and Adult Functional Status: Do Early-Life Exposures Jeopardize Healthy Aging?

OBJECTIVES

To examine whether childhood disadvantage is associated with later-life functional status and identify mediating factors.

METHODS

Unique and additive effects of five childhood domains on functional status were assessed at baseline (2006) and over time (2006-2016) in a sample of 13,894 adults from the Health and Retirement Study (>50 years). Adult health behaviors and socioeconomic status (SES) were tested as mediators.

RESULTS

Respondents exposed to multiple childhood disadvantages (OR = .694) as well as low childhood SES (OR = .615), chronic diseases (OR = .694), impairments (OR = .599), and risky adolescent behaviors (OR = .608) were less likely to be free of functional disability by baseline. Over time, these unique and additive effects of childhood disadvantage increased the hazard odds of eventually developing functional disability (e.g., additive effect: hOR = 1.261). Adult health behaviors and SES mediated some of these effects.

DISCUSSION

Given the enduring effects of childhood disadvantage, policies to promote healthy aging should reduce exposure to childhood disadvantage.

Impact of Surgeon Type and Rurality on Treatment and Survival of Ovarian Cancer Patients

BACKGROUND

National Comprehensive Cancer Network guidelines recommend ovarian cancer patients receive cancer-directed surgery from a gynecologic oncologist surgeon. We aimed to determine if rurality impacts type of surgeon and estimate if the interaction between rurality and type of surgeon impacts cytoreductive surgery, chemotherapy initiation, and survival.

METHODS

Our population-based cohort of Iowan (N=675) ovarian cancer patients included women diagnosed with histologically confirmed stages IB-IV cancer in 2010 to 2016 at the ages of 18 to 89 years old and who received cancer-directed surgery in Iowa. Multivariable logistic regression analysis and Cox proportional hazards models were used.

RESULTS

Rural (vs. urban) patients were less likely to receive surgery from a gynecologic oncologist (adjusted odds ratio [OR]: 0.48; 95% confidence interval [CI]: 0.30-0.78). Rural patients with a gynecologic oncologist (vs. nongynecologic oncologist) surgeon were more likely to receive cytoreduction (OR: 2.84; 95% CI: 1.31-6.14) and chemotherapy (OR: 4.22; 95% CI: 1.82-9.78). Gynecologic oncologist-provided surgery conferred a 3-year cause-specific survival advantage among rural patients (adjusted hazard ratio: 0.57; 95% CI: 0.33-0.97) and disadvantage among urban patients (hazard ratio: 1.77; 95% CI: 1.02-3.06) in the model without treatment covariates. Significance dissipated in models with treatment variables.

DISCUSSION

The variation in the gynecologic oncologist survival advantage may be because of treatment, referral, volume, or nongynecologic oncologist surgeons' specialty difference by rurality. This is the first study to investigate the ovarian cancer survival advantage of having a gynecologic oncologist surgeon by rurality.

Longitudinal Moderated Mediation Analysis in Parallel Process Latent Growth Curve Modeling in Intervention Studies

BACKGROUND

Intervention studies are used widely in nursing research to explore the efficacy of intervention programs for changing targeted health outcomes. However, the analyses of such studies have focused predominantly on their main intervention effects; most studies ignore the mechanisms underlying how the intervention programs work partly because of lack of application details of the longitudinal mediation analysis techniques.

OBJECTIVES

The aim of this study was to illustrate an application of parallel process latent growth curve modeling (PP-LGCM) to examine longitudinal moderated mediation effects.

METHODS

Longitudinal data from an online bone health intervention study were used to demonstrate the step-by-step application of PP-LGCM with Mplus statistical software.

RESULTS

With modification indices, we were able to achieve adequate model fit for PP-LGCM in our data. The mediation effects of self-efficacy on the intervention effects on exercise were nonsignificant for the entire sample. However, the conditional indirect effect showed the mediation effects were moderated by age group.

DISCUSSION

PP-LGCM provides an efficient way to analyze and explain the underlying mechanisms for the intervention effects in a trial, especially when the intervention program is guided by a theory.

Comparing alternating pressure mattresses and high-specification foam mattresses to prevent pressure ulcers in high-risk patients: the PRESSURE 2 RCT

BACKGROUND

Pressure ulcers (PUs) are a burden to patients, carers and health-care providers. Specialist mattresses minimise the intensity and duration of pressure on vulnerable skin sites in at-risk patients.

PRIMARY OBJECTIVE

Time to developing a new PU of category ≥ 2 in patients using an alternating pressure mattress (APM) compared with a high-specification foam mattress (HSFM).

DESIGN

A multicentre, Phase III, open, prospective, planned as an adaptive double-triangular group sequential, parallel-group, randomised controlled trial with an a priori sample size of 2954 participants. Randomisation used minimisation (incorporating a random element).

SETTING

The trial was set in 42 secondary and community inpatient facilities in the UK.

PARTICIPANTS

Adult inpatients with evidence of acute illness and at a high risk of PU development.

INTERVENTIONS AND FOLLOW-UP

APM or HSFM - the treatment phase lasted a maximum of 60 days; the final 30 days were post-treatment follow-up.

MAIN OUTCOME MEASURES

Time to event.

RESULTS

From August 2013 to November 2016, 2029 participants were randomised to receive either APM (n = 1016) or HSFM (n = 1013). Primary end point - 30-day final follow-up: of the 2029 participants in the intention-to-treat population, 160 (7.9%) developed a new PU of category ≥ 2. There was insufficient evidence of a difference between groups for time to new PU of category ≥ 2 [Fine and Gray model HR 0.76, 95% confidence interval (CI) 0.56 to 1.04; exact p-value of 0.0890 and 2% absolute difference]. Treatment phase sensitivity analysis: 132 (6.5%) participants developed a new PU of category ≥ 2 between randomisation and end of treatment phase. There was a statistically significant difference in the treatment phase time-to-event sensitivity analysis (Fine and Gray model HR 0.66, 95% CI 0.46 to 0.93; p = 0.0176 and 2.6% absolute difference). Secondary end points - 30-day final follow-up: new PUs of category ≥ 1 developed in 350 (17.2%) participants, with no evidence of a difference between mattress groups in time to PU development, (Fine and Gray model HR 0.83, 95% CI 0.67 to 1.02; p-value = 0.0733 and absolute difference 3.1%). New PUs of category ≥ 3 developed in 32 (1.6%) participants with insufficient evidence of a difference between mattress groups in time to PU development (Fine and Gray model HR 0.81, 95% CI 0.40 to 1.62; p = 0.5530 and absolute difference 0.4%). Of the 145 pre-existing PUs of category 2, 89 (61.4%) healed - there was insufficient evidence of a difference in time to healing (Fine and Gray model HR 1.12, 95% CI 0.74 to 1.68; p = 0.6122 and absolute difference 2.9%). Health economics - the within-trial and long-term analysis showed APM to be cost-effective compared with HSFM; however, the difference in costs models are small and the quality-adjusted life-year gains are very small. There were no safety concerns. Blinded photography substudy - the reliability of central blinded review compared with clinical assessment for PUs of category ≥ 2 was 'very good' (kappa statistic 0.82, prevalence- and bias-adjusted kappa 0.82). Quality-of-life substudy - the Pressure Ulcer Quality of Life - Prevention (PU-QoL-P) instrument meets the established criteria for reliability, construct validity and responsiveness.

LIMITATIONS

A lower than anticipated event rate.

CONCLUSIONS

In acutely ill inpatients who are bedfast/chairfast and/or have a category 1 PU and/or localised skin pain, APMs confer a small treatment phase benefit that is diminished over time. Overall, the APM patient compliance, very low PU incidence rate observed and small differences between mattresses indicate the need for improved indicators for targeting of APMs and individualised decision-making. Decisions should take into account skin status, patient preferences (movement ability and rehabilitation needs) and the presence of factors that may be potentially modifiable through APM allocation, including being completely immobile, having nutritional deficits, lacking capacity and/or having altered skin/category 1 PU.

FUTURE WORK

Explore the relationship between mental capacity, levels of independent movement, repositioning and PU development. Explore 'what works for whom and in what circumstances'.

TRIAL REGISTRATION

Current Controlled Trials ISRCTN01151335.

FUNDING

This project was funded by the National Institute for Health Research Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 23, No. 52. See the NIHR Journals Library website for further project information.

The mediating role of extent of resection in the relationship between the tumor characteristics and survival outcome of glioma

The prognostic value of tumor characteristics for glioma has been controversial, partly because of a lack of knowledge about how these associations develop. Extent of resection may be factors that mediate the relationship between tumor characteristics and the hazard of death from glioma.

Patients and Methods: This consecutive study retrospectively included a group of 393 treatment-naive patients with newly, pathologically confirmed glioma between January 2004 and December 2014. Information on patient age, gender, Karnofsky Performance Status (KPS), tumor grade, tumor size, tumor location, presence or absence of contrast enhancement on MRI and extent of tumor resection have all been collected. The discrete-time survival model integrating survival outcomes within structural equation models was employed to develop and evaluate a comprehensive hypothesis regarding the direct and indirect impact of tumor characteristics on the hazard of death from glioma, mediated by the extent of resection.

Results: Except for tumor location, the indirect effects of tumor grade, contrast enhancement, and tumor size on PFS of glioma through extent of resection were found significant in the model.

Conclusion: This study provides a better understanding of the process through which tumor characteristics is associated with hazard of death from glioma.

Evaluating a Method to Estimate Mediation Effects With Discrete-Time Survival Outcomes

The utility of evaluating mediation effects spans across research domains. The model facilitates investigation of underlying mechanisms of event timing and, as such, has the potential to help strengthen etiological research and inform intervention work that incorporates the evaluation of mediating variables. In order for the analyses to be maximally useful however, it is critical to employ methodology appropriate for the data under investigation. The purpose of this paper is to evaluate a regression-based approach to estimating mediation effects with discrete-time survival outcomes. We empirically evaluate the performance of the discrete-time survival mediation model in a statistical simulation study, and demonstrate that results are functionally equivalent to estimates garnered from a potential-outcomes framework. Simulation results indicate that parameter estimates of mediation in the model were statistically accurate and precise across the range of examined conditions. Type 1 error rates were also tolerable in the conditions studied. Adequate power to detect effects in the model, with binary X and continuous M variables, required effect sizes of the mediation paths to be medium or large. Possible extensions of the model are also considered.

Cohort Study on Laryngeal Cough Reflex, Respiratory Disease, and Death: A Mediation Analysis

OBJECTIVE

To estimate quantitatively whether the presence of cough associated with dysphagia (laryngeal cough reflex) increased mortality through respiratory disease among community-dwelling older Japanese.

DESIGN

A 6-year follow-up prospective cohort study (from 2010 to 2017).

SETTING

Thirteen municipalities in Japan.

PARTICIPANTS

Community-dwelling individuals aged 65 years or older (N = 32,682).

MEASURES

The baseline survey was conducted through self-reported questionnaire. Exposure was experience of laryngeal cough reflex while drinking. The outcome was all-cause mortality. All covariates were selected from demographic, socioeconomic variables, baseline health and functional status, smoking, alcohol drinking, number of remaining teeth, and stroke. The mediator variable was respiratory disease. Cox proportional hazards regression was used to estimate hazard ratios (HRs) for mortality. Parametric mediation analysis was conducted to estimate the effect of laryngeal cough reflex on the mean residual time to death mediated through respiratory disease.

RESULTS

Among the 32,682 participants (mean age = 74.1 years, standard deviation = 5.9 years), 5550 (17.0%) experienced laryngeal cough reflex at baseline. A total of 4037 deaths occurred. Crude mortality rates of the participants with or without laryngeal cough reflex were 16.3% and 11.6%, respectively. After adjusting for covariates, laryngeal cough reflex [HR = 1.10; 95% confidence interval (CI) = 1.02 to 1.19] and respiratory disease (HR = 1.80; 95% CI = 1.62 to 2.00) were associated with mortality. The mediation analysis showed that respiratory disease significantly (P < .001) and partially mediated the association between laryngeal cough reflex, an indicator of 1 or more conditions such as chronic aspiration and mortality.

CONCLUSIONS/IMPLICATIONS

Laryngeal cough reflex was a prevalent condition, and it was associated with all-cause mortality in community-dwelling older Japanese individuals. Clinicians could contribute to reduce mortality risk by addressing swallowing function problems using their interprofessional collaboration team (speech-language pathologist, dentist, rehabilitation doctor, otolaryngologist, respiratory physician, and gerontologist).

Modifiable Risk Factors Explain Socioeconomic Inequalities in Dementia Risk: Evidence from a Population-Based Prospective Cohort Study

BACKGROUND

Differences in dementia risk across the gradient of socioeconomic status (SES) exist, but their determinants are not well understood.

OBJECTIVE

This study investigates whether health conditions and lifestyle-related risk factors explain the SES inequalities in dementia risk.

METHODS

6,346 participants from the English Longitudinal Study of Ageing were followed up from 2008/2009 until 2014/2015. We used Cox regression adjusted for age, gender, wealth/education, and clustering at the household level to examine the association between SES markers (wealth, education) and time to dementia in a structural equation model including potential mediation or effect modification by a weighted compound score of twelve modifiable risk and protective factors for dementia ('LIfestyle for BRAin health' (LIBRA) score).

RESULTS

During a median follow-up of 6 years, 192 individuals (3.0%) developed dementia. LIBRA scores decreased with increasing wealth and higher educational level. A one-point increase in the LIBRA score was associated with a 13% increase in dementia risk (hazard ratio (HR) = 1.13, 95% confidence interval 1.07-1.19). Higher wealth was associated with a decreased dementia risk (HR = 0.58, 0.39-0.85). Mediation analysis showed that 52% of the risk difference between the highest and lowest wealth tertile was mediated by differences in LIBRA (indirect effect: HR = 0.75, 0.66-0.85). Education was not directly associated with dementia (HR = 1.05, 0.69-1.59), but was a distal risk factor for dementia by explaining differences in wealth and LIBRA scores (indirect effect high education: HR = 0.92, 0.88-0.95).

CONCLUSION

Socioeconomic differences in dementia risk can be partly explained by differences in modifiable health conditions and lifestyle factors.

Mediation analysis with a time-to-event outcome: a review of use and reporting in healthcare research

BACKGROUND

Mediation analysis tests whether the relationship between two variables is explained by a third intermediate variable. We sought to describe the usage and reporting of mediation analysis with time-to-event outcomes in published healthcare research.

METHODS

A systematic search of Medline, Embase, and Web of Science was executed in December 2016 to identify applications of mediation analysis to healthcare research involving a clinically relevant time-to-event outcome. We summarized usage over time and reporting of important methodological characteristics.

RESULTS

We included 149 primary studies, published from 1997 to 2016. Most studies were published after 2011 (n = 110, 74%), and the annual number of studies nearly doubled in the last year (from n = 21 to n = 40). A traditional approach (causal steps or change in coefficient) was most commonly taken (n = 87, 58%), and the majority of studies (n = 114, 77%) used a Cox Proportional Hazards regression for the outcome. Few studies (n = 52, 35%) mentioned any of the assumptions or limitations fundamental to a causal interpretation of mediation analysis.

CONCLUSION

There is increasing use of mediation analysis with time-to-event outcomes. Current usage is limited by reliance on traditional methods and the Cox Proportional Hazards model, as well as low rates of reporting of underlying assumptions. There is a need for formal criteria to aid authors, reviewers, and readers reporting or appraising such studies.

Sex Differences in Attaining Cigarette Smoking and Nicotine Dependence Milestones Among Novice Smokers

There may be sex differences in the response to nicotine, according to findings of studies in animals; however, sex differences in the natural course of cigarette smoking and nicotine dependence are documented in few studies. Prevalent (n = 240 girls; n = 184 boys) and incident (n = 231 girls; n = 184 boys) cigarette smokers from the Nicotine Dependence in Teens Study were followed up to 5 years after first puff, from age 12 to 18 years (1999-2005). We used Cox proportional hazards models to compare time to development of 3 cigarette-use (i.e., whole cigarette; 100 cigarettes lifetime; regular smoking), and 3 nicotine-dependence symptom (i.e., "really need a cigarette"; mentally addicted; physically addicted) milestones across sex. Girls were at higher risk than boys of attaining all milestones; hazard ratios (95% confidence intervals) ranged from 1.35 (1.06, 1.72) for 100 cigarettes lifetime to 1.74 (1.44, 2.10) for "really need a cigarette." Among nonregular smokers, 26% (8%; 43%) and 25% (6%; 44%) more girls than boys reported "really need a cigarette" 1 and 2 years, respectively, after first puff. Preventive interventions may need adjustment to incorporate these findings. Additional research should clarify the relative contribution of biological and social underpinnings of these sex differences.

Screening Patterns and Mortality Differences in Patients With Lung Cancer at an Urban Underserved Community

BACKGROUND

The landmark National Lung Screening Trial demonstrated significant reduction in lung cancer-related mortality. However, European lung cancer screening (LCS) trials have not confirmed such benefit. We examined LCS patterns and determined the impact of LCS-led diagnosis on the mortality of newly diagnosed patients with lung cancer in an underserved community.

PATIENTS AND METHODS

Medical records of patients diagnosed with primary lung cancer in 2013 through 2016 (n = 855) were reviewed for primary care provider (PCP) status and LCS eligibility and completion, determined using United States Preventative Services Task Force guidelines. Univariate analyses of patient characteristics were conducted between LCS-eligible patients based on screening completion. Survival analyses were conducted using Kaplan-Meier and multivariate Cox regression.

RESULTS

In 2013 through 2016, 175 patients with primary lung cancer had an established PCP and were eligible for LCS. Among them, 19% (33/175) completed screening prior to diagnosis. LCS completion was associated with younger age (P = .02), active smoking status (P < .01), earlier stage at time of diagnosis (P < .01), follow-up in-network cancer treatment (P = .03), and surgical management (P < .01). LCS-eligible patients who underwent screening had improved all-cause mortality compared with those not screened (P < .01). Multivariate regression showed surgery (hazard ratio, 0.31; P = .04) significantly affected mortality.

CONCLUSION

To our knowledge, this is the first study to assess LCS patterns and mortality differences on patients with screen-detected lung cancer in an urban underserved setting since the inception of United States Preventative Services Task Force guidelines. Patients with a LCS-led diagnosis had improved mortality, likely owing to cancer detection at earlier stages with curative treatment, which echoes the finding of prospective trials.

Incorporating nonlinearity into mediation analyses

Background

Mediation is an important issue considered in the behavioral, medical, and social sciences. It addresses situations where the effect of a predictor variable X on an outcome variable Y is explained to some extent by an intervening, mediator variable M. Methods for addressing mediation have been available for some time. While these methods continue to undergo refinement, the relationships underlying mediation are commonly treated as linear in the outcome Y, the predictor X, and the mediator M. These relationships, however, can be nonlinear. Methods are needed for assessing when mediation relationships can be treated as linear and for estimating them when they are nonlinear.

Methods

Existing adaptive regression methods based on fractional polynomials are extended here to address nonlinearity in mediation relationships, but assuming those relationships are monotonic as would be consistent with theories about directionality of such relationships.

Results

Example monotonic mediation analyses are provided assessing linear and monotonic mediation of the effect of family functioning (X) on a child’s adaptation (Y) to a chronic condition by the difficulty (M) for the family in managing the child's condition. Example moderated monotonic mediation and simulation analyses are also presented.

Conclusions

Adaptive methods provide an effective way to incorporate possibly nonlinear monotonicity into mediation relationships.

Affluence and Private Health Insurance Influence Treatment and Survival in Non-Hodgkin's Lymphoma

BACKGROUND

The aim of this study was to investigate inequalities in survival for non-Hodgkin's lymphoma (NHL), distinguishing between direct and indirect effects of patient, social and process-of-care factors.

METHODS

All cases of NHL diagnosed in Ireland in 2004-2008 were included. Variables describing patient, cancer, stage and process of care were included in a discrete-time model of survival using Structural Equation Modelling software.

RESULTS

Emergency admissions were more common in patients with co-morbid conditions or with more aggressive cancers, and less frequent for patients from more affluent areas. Aggressive morphology, female sex, emergency admission, increasing age, comorbidity, treatment in a high caseload hospital and late stage were associated with increased hazard of mortality. Private patients had a reduced hazard of mortality, mediated by systemic therapy, admission to high caseload hospitals and fewer emergency admissions.

DISCUSSION

The higher rate of emergency presentation, and consequent poorer survival, of uninsured patients, suggests they face barriers to early presentation. Social, educational and cultural factors may also discourage disadvantaged patients from consulting with early symptoms of NHL. Non-insured patients, who present later and have more emergency admissions would benefit from better access to diagnostic services. Older patients remain disadvantaged by sub-optimal treatment, treatment in non-specialist centres and emergency admission.