Physician Training Stress and Accelerated Cellular Aging

Abstract

Background

Stress is a key precipitant for many common diseases but established biological markers to track stress and guide investigations into mechanism linking stress and disease are lacking. Cross-sectional studies have identified correlations between stress and telomere attrition, but no large, longitudinal studies examining the impacts of chronic stress on telomere length exist. Residency training for physicians is a well-established stressful experience and can be used as a prospective stress model.

Methods

In a longitudinal cohort study of 250 interns (first-year residents) at 55 U.S hospital systems serving during the 2015–16 academic year, we examined associations between measures of the residency experience and saliva-measured telomere attrition.

Results

Telomere length shortened significantly over the course of internship year, from 6465.1 ± 876.8 base pairs before internship to 6321.5 ± 630.6 base pairs at the end of internship (t(246) = 2.69; P=0.008). Stressful early family environments and neuroticism were significantly associated with shorter pre-internship telomere length. Longer work hours were associated with greater telomere intern telomere loss over the year (p = 0.002). Of note, the mean telomere attrition during internship year was six times greater than the typical annual attrition rate identified in a recent meta-analysis.

Conclusions

This work implicates telomere attrition as a biologically measurable consequence of physician training, with the magnitude of attrition associated with workload. Identification of an objective, biological sequelae of residency stress may help to facilitate the development of effective interventions. Further, the findings implicate telomere attrition as an objective biomarker to follow the pathologic effects of stress, in general.

INTRODUCTION

Medical residency, the first phase of professional medical training in the United States, is a stressful period in the career of physicians. Residents are faced with long work hours, sleep deprivation, and a new degree of responsibility for patients (1–4). Despite efforts from the accrediting agencies and training institutions to limit work hours, ensure adequate supervision, and implement programs to monitor and promote wellness, residents continue to report high stress and burnout (1, 3, 5–7). In addition to precipitating mental health problems (8, 9), stress exposure is an important risk factor for systemic diseases such as obesity, type 2 diabetes and cardiovascular disease (10). To date, studies of physician training stress have largely focused on self-report measures of perceived stress, mental health, and well-being (2, 4, 6, 11). Direct assessment of the objective, biological effects of physician training hold the promise to broaden our understanding of the effects of the training on health and more precisely identify targets for intervention.

Telomeres are complexes consisting of tandem DNA nucleotide TTAGGG repeats complexed with proteins found at the ends of chromosomes that help ensure stability and maintenance of the nuclear genome (10, 12). Telomere length decreases with each cell replication cycle (12), and once telomeres reach a critically shortened length, cells enter a state of replicative senescence (12), A recent systematic review of 129 primary studies found that telomeres shorten at a rate of 24.7 base pairs per year in somatic tissues on average (13). In cross-sectional studies, shorter telomere length has been correlated with multiple forms of stress exposure, including psychological stress, perceived stress, social stress, caregiver stress, post-traumatic stress, and childhood maltreatment (10, 14–18). Telomere attrition is also a predictor of somatic disease risk, including heart disease and diabetes, as well as of general senescence and mortality (10, 12). However, no large, longitudinal studies examining the impacts of chronic stress on telomere length have been reported.

The objective of this study was to examine the biological impacts of residency training by longitudinally assessing telomere length attrition across internship year, and how this marker of cellular aging relates to factors of the training experience.

METHODS

Participants

The Intern Health Study is a multisite prospective cohort study that aims to assess stress and mood during internship. Training physicians (N = 4,904) beginning internship at 55 institutions (Box) in 2015 were sent an e-mail 8–10 weeks prior to commencing internship and invited to participate in the study (Figure 1) (9, 19). E-mail invitations were returned as undeliverable for 44 subjects and a valid e-mail address was unable to be obtained. Of the remaining invited subjects, 64.3% (3,127/4,860) agreed to participate. The first 250 subjects to enroll in the parent study were included in the current analysis of telomere length. Subjects were given $50 Amazon gift cards for participation. The institutional review board at the University of Michigan approved the study. Power analysis estimated that 76 subjects would allow detection of a moderate effect size change in telomere length over intern year (d ≥ 0.5) with a = 0.05 and power of 0.99.

Figure 1.

Flow of Participants Through the Study

Baseline Assessment

All surveys were conducted through a secure, web-based platform designed to maintain confidentiality. Subjects completed a baseline survey upon study enrollment prior to commencing internship. The survey assessed general demographic factors (age, sex, and marital status), specialty, self-reported history of depression, and various psychological measures. Depressive symptoms were measured with the nine-item Patient Health Questionnaire (PHQ-9), which scores each of the nine criteria for depression from “0” (not at all) to “3” (nearly every day), with PHQ-9 scores ≥10 having a sensitivity of 81.3% (95% CI, 71.6%−89.3%) and specificity of 85.3% (95% CI, 81.0%−89.1%) for major depressive disorder (20). Early life stress has been strongly implicated with telomere attrition in prior studies (18). Neuroticism has been associated with telomere attrition in some but not all prior studies (21, 22). Neuroticism (defined as tendency for quick arousal, slow relaxation from arousal and tendency to respond with negativity (23)) was assessed using the NEO-Five Factor Inventory, which is comprised of twelve questions scored on a scale from “1” (disagree strongly) to “5” (agree strongly) (24). Early family environment was assessed using the Risky Families Questionnaire, which scores the extent to which respondents felt loved, were shown affection, were abused, lived with a substance abuser, lived in an organized and well managed household, and had adults who “knew what they were up to” from “1” (rarely or none of the time) to “4” (most or all of the time) (25).

Within-Internship Assessments

Participants were contacted via e-mail at months 3, 6, 9, and 12 of their internship year and asked to complete a web-based survey that included the PHQ-9, week average work hours and sleep, and outside residency stressful life events (defined as a serious illness, death or serious illness in a close family member or friend, financial problems, end of a serious relationship, or becoming a victim of crime or domestic violence) during the past 3 months (9). These measures have previously been shown to be associated with increased subjective and objective markers of residency stress (9).

DNA Collection and Telomere Length Analysis

Subjects submitted Oragene OG-500 salivary DNA self-collection kit saliva samples via postal mail for telomere assessment 2–6 weeks before internship and at the completion of internship (DNA Genotek Inc., Ottawa, ON, Canada) (26). For each individual subject, baseline and follow-up samples were extracted and assayed for telomere length on the same plate. Telomeres were measured relative to a single copy gene (human beta-globin) using an assay adapted from previously published methods (27). Telomere primer sequences were [5′-CGGTTT(GTTTGG)5GTT-3′] and [5′-GGCTTG(CCTTAC)5CCT-3′] at final concentrations of 100 and 900 nM, respectively. Beta-globin primer sequences were [5′ GCTTCTGACACAACTGTGTTCACTAGC-3′] and [5′-CACCAACTTCATCCACGTTCACC-3′] at final concentrations of 300 and 700 nM, respectively. The final reaction mix contained 20 mM Tris-HCl, pH 8.4; 50 mM KCl; 200 μM of each dNTP; 1% DMSO; 0.4 × concentration SYBR Green I (Invitrogen, Carlsbad, CA); 22 ng E. coli DNA (MP Biomedicals, Solon, OH); 0.4 units of Platinum Taq DNA polymerase (Invitrogen); and 0.5–10 ng of genomic DNA per 11-pl reaction. Reference DNA from HeLa cancer cells was used to generate a standard curve for all PCR runs using a LightCycler 480 real-time PCR machine (Roche Diagnostics, Indianapolis, IN). The inter-assay coefficient of variation for telomere length measurement was 4%. Samples were run in duplicate. If the results for the same sample varied by >7%, the sample was run a third time and the average of the two closest values was used. T/S ratio was converted to base pairs using the formula T/S*2413+3274, which was based on comparison of the ratio to Southern blot analysis (28, 29).

Statistical Analysis

Pre- to post-internship change in telomere length was assessed using a paired samples t-test. Relationships between demographic characteristics and telomere length were assessed using bivariate correlation analysis. Associations between other independent variables and telomere length were examined using Pearson correlations. Univariate generalized linear models (GLMs) were used to examine the effects of the following within-internship variables on telomere attrition: 1) average intern year work hours (calculated by averaging each quarter reported hours), 2) stressful life events (calculated by averaging each quarter reported events), 3) PHQ-9 depressive symptoms scores (calculated by averaging each quarter reported PHQ-9), 4) sleep length (calculated by averaging each quarter reported average sleep), and 5) specialty. We created quartiles for work hours by derived work shift length, calculated by dividing the reported average number of work hours each quarter by 5 (assuming a 5-day work week). As prior literature has identified age and sex as significant predictors of telomere length (10, 13), we adjusted for these variables in all Pearson correlations and GLMs. Missing telomere data from three subjects were dealt with by listwise deletion. Statistical significance was defined as a two-tailed P value of <0.05. All p values in this study were subjected to the Benjamini-Hochberg procedure; reported significant p values were significant at the Benjamini-Hochberg calculated critical value (calculated using the formula (i/m)Q, where i=the individual p-value’s rank, m=total number of tests, and Q=the false discovery rate) with the false discovery rate set to 0.15. Data conformed to the assumptions of normality of distribution. Analyses were performed using IBM SPSS Statistics Software Version 24.0 (IBM Corp., Armonk, NY). Summary data are presented as mean (standard deviation, SD) unless otherwise indicated.

College Student Control Sample

In order to gain additional insight into the expected telomere attrition rate in young adults, we assessed a sample of 84 incoming University of Michigan freshmen in 2016 and 2017 (average age - 18.3 (0.8) years at baseline) who provided DNA samples for longitudinal analysis in the August before starting college and then again in August after their freshman year. Telomere length in these samples proceeded as described above.

RESULTS

Participant Characteristics

250 interns agreed to participate in this one-year study of telomere attrition during the intern year, of which 98.8% (247/250) provided salivary DNA samples for both pre- and post-internship analysis of telomere length. The mean age of the interns at baseline was 27.4 (2.9) years, and 52.6% (130/247) were women (Table 1). The subjects taking part in the present telomere study (N=247) did not differ significantly from the overall 2015 Intern Health Study Cohort (N=3,127) on gender, age, history of depression, neuroticism, work hours, baseline depressive symptom score or internship depressive symptom score (all P values > 0.05). Compared to all individuals in the American Association for Medical Colleges (AAMC) database beginning internship, this sample was slightly younger (27.4 vs. 28.8 years old) and included a slightly higher percentage of women (52.6% vs. 48.6%).

Table 1.

Resident Characteristics (N = 247)

Characteristic	Value*
Age — years	27.4 (2.9)
Male — no. (%)	117 (47.4)
Hours worked per week	64.5 (9.2)
PHQ-9 score at baseline	2.6 (2.6)
PHQ-9 score at follow-up	5.2 (3.6)
Positive history of depression — no. (%)	138 (55.9)
Positive family history of depression — no. (%)	121 (49.0)
Risky families questionnaire score	11.8 (8.4)
Neuroticism score	21.2 (8.6)
Number of stressful life events during intern year	0.8 (1.0)
Telomere length at baseline	6465.0 (876.8)
Telomere length at follow up	6321.5 (630.5)

^*Summary data are presented as mean (standard deviation) unless otherwise indicated.

Abbreviations: no., number; PHQ-9, nine-item Patient Health Questionnaire

Telomere Attrition with Intern Year

The mean gap between collection of the baseline and follow-up samples was 385.2 (11.6) days. There was significant attrition of telomere length base pairs from the start of internship (6465.1 (876.8); T/S ratio 1.32(0.36)) to follow up (6321.5 (630.6); T/S ratio 1.26(0.26)), amounting to a loss of 143.5 (839.1) base pairs (−.060(.35) T/S ratio) over the course of intern year (t(246) = 2.69, 95% CI 38.4 to 248.7, P=0.008).

Correlates of Baseline Telomere Length

Neither age (r=−0.04, P=0.58) nor sex (r=−0.12, P=0.07) were significantly correlated with pre-internship telomere length (Table 2). Controlling for age and sex, a more stressful early family environment (r=0.13, P=0.04) and higher levels of neuroticism (r=0.18, P=0.006) were significantly associated with pre-internship telomere length. Personal or family history of depression, depressive symptom scores prior to intern year, and sleep were not associated with pre-internship telomere length on age- and sex-adjusted analysis (P>0.05 for all comparisons).

Table 2.

Correlation Coefficients Between Baseline Pre-Internship Characteristics and Telomere Length (n = 247)

Baseline characteristic	Pre-internship telomere length	P value	Change in telomere length	P value
Age	0·04	0·58	−0·15	0·02
Sex	−0·12	0·07	0·03	0·62
Baseline depressive symptom (PHQ-9) score*	0·07	0·31	−0·07	0·28
Personal history of depression*	−0·01	0·84	−0·03	0·58
Family history of depression*	0·04	0·63	−0·04	0·59
Early family environment (RFQ) score*	0·13	0·04	−0·04	0·53
Neuroticism score*	0·18	0·006	−0·09	0·15
Sleep hours	−0·09	0·15	0·10	0·14

^*Analysis performed while adjusting for age and sex.

Abbreviations: PHQ-9, 9-item Patient Health Questionnaire; RFQ, Risky Families Questionnaire.

Predictors of Telomere Attrition

There was a significant correlation between baseline and follow-up telomere length (r = .418; p < .0001). Among demographic variables, older age (r=−0.15, P=0.02), but not sex (r=0.03, P=0.62), was significantly associated with telomere length attrition over the internship year. Early family environment and neuroticism were not significantly associated with telomere attrition over the year (P>0.05 for both comparisons). Interns reported working a mean of 64.5 (9.1) hours per week. Currently, the American Council for Graduate Medical Education (ACGME) requires medical interns work shifts that are no longer than 16 hours a day. We categorized average work hours by projected shift length to gain insight regarding a potential dose-response to working shifts below this threshold (15 hour or less shifts) compared to working at or above this threshold (16 hours or more). When controlling for age, sex, there was a main effect of work hours on telomere attrition (F(3, 236)=5.20; P=0.002; Cohen’s d=−0.51; Figure 2), such that working more hours, and potentially more shifts at or above the ACGME threshold, was associated with greater telomere attrition (Supplemental Table 1). These results remained significant when including pre-internship telomere length, early family environment and neuroticism in the models (F=2.93, P=.034, F=5.18, P=0.002 and F=5.03, P=0.002, respectively). Other within-internship variables were not significantly correlated with telomere attrition (P>0.05, data not shown).

Figure 2. Univariate Generalized Linear Model of Telomere Length Change With Internship Stress Stratified by Work Hours.

Categorization of average work hours was structured to examine the impact of projected shift hours below (9-, 12, and 15) or above (16 or above) the ACGME-required threshold for interns. There was a main effect of work hours on telomere attrition (F(3, 236)=5.20; P=0.002), such that working more shifts at or above the ACGME threshold was associated with increased telomere attrition. Data are mean (SE). Reference telomere loss is expected average telomere loss per year from article¹⁴

Comparison to Normative and Control Sample Attrition Rates

In the meta-analysis¹³ utilized to establish the typical annual telomere attrition of 25 bp/year, the average age of subjects was substantially higher than the average age of subjects in the current study. Thus, we obtained access a control sample of young adults to gain additional insight into the expected telomere attrition in our sample. In contrast to the intern sample, telomere length did not decrease over the year. In fact, there was a non-significant increase in telomere length across freshman year (t(83)=−2.026, p=0.05).

DISCUSSION

This prospective cohort study of 250 interns demonstrated that physician training is associated with a significant decrease in telomere length, with an average telomere attrition of 143.5 (839.1) base pairs over the course of intern year. Telomere shortening remained significant after accounting for previously reported predictors of telomere attrition, such as personality traits and number of stressful life events outside of residency training. The rate of attrition related to work hours, with higher work hours associated with greater telomere attrition (Figure 2). Further, the rate of attrition during internship was substantially larger than the typical attrition of 25 bp/year in a recent systematic review and meta-analysis of over 100 studies and in a control sample of college students (13).

The longitudinal model of this study provides a powerful within subject design that suggests that, in this relatively homogenous group in regard to age and education-level, work load predicts accelerated attrition of telomeres, which can have long lasting consequences on individual’s health. These are the first data showing a link between the subjectively-stressful experience of physician training and an objective marker of cellular stress exposure. By identifying a relation between work hours and telomere attrition, these results demonstrate that a quantifiable marker, such as telomere length, can facilitate the identification of the critical components of physician stress that may have long-lasting consequences of physical health and wellbeing, which are critical to target for intervention. Potentially, telomere length may help evaluate the effectiveness of such interventions. While our results indicate that internship is associated with accelerated telomere attrition, the mechanisms underlying the association between stress and telomere length are unknown. Telomeres shorten after repeated cellular divisions and cellular stress exposures (10). It is possible that the stress associated with the experience of intern year directly activates or is associated with mechanisms activating increased cellular stress and replication, resulting in accelerated telomere shortening (30). Immune cell replication could also have been increased through infectious agent exposure in health care settings. Decreased telomerase activity, a key regulator of telomere length, has been associated with stress exposure, and may play a role in linking internship and telomere attrition (10).

A large and growing body of work has identified that physician training and practice adversely affects a range mental health outcomes, including burnout, depression and suicide risk (1, 31, 32). The present findings compliments and extends this body of work. First, the identification of an objective, biological marker of physician training stress further substantiates the harmful consequences of this stress on well-being and mental health and adds the urgency to the case that reform in graduate medical education is needed. Further, the identified telomere attrition raises concern that the consequences of physician training stress extend beyond well-being and include increased risk for systemic, chronic diseases previously associated with shortened telomeres (10, 18, 30, 33). Future studies are needed to assess whether the telomere attrition associated with internship persist through training and practice and if so, their relationship with disease.

Beyond physician training stress specifically, this study suggests that telomere length may be useful as a biomarker for stress in general. Chronic activation of the stress response system is a leading cause of death and disability, through increasing the risk of psychiatric disorders such as depression and anxiety as well as systemic diseases such as obesity, type 2 diabetes and cardiovascular disease. Shorter telomere length has been correlated with multiple forms of stress exposure in cross-sectional studies, including psychological stress, perceived stress, social stress, caregiver stress, post-traumatic stress, and childhood maltreatment (10, 14–18). Demonstrating accelerated telomere attrition in a longitudinal study and identifying workload as a critical component of the stress impact substantially adds to the evidence supporting telomere length as a stress marker.

Our study has a number of limitations. First, the sample was restricted to interns and, as such, results may not be generalizable to all training physicians or fully representative of all residents. Related, training physicians differ from the general population in important ways. The longitudinal association between stress and telomere attrition should be replicated in other population samples. Second, while our study measured many factors previously reported to impact resident perceived stress and mental health, we were unable to measure other potential confounders such as smoking, weight, diet and exercise (12, 34). Of note, effect sizes reported for telomere length and smoking (d=−.011, (35)), diet (no significant effect size found (36, 37)), and exercise (no significant effect size found (34, 38)) have typically been in the small to nonsignificant range while the effect found here was in the medium range (d=−.051). However, such variables may be helpful in understanding behavioral mechanisms linking the internship experience and telomere attrition. Third, the data that we used to establish normative rates of telomere rates are imperfect. Most notably, most prior studies include samples that are, on average, substantially older than the intern sample. However, evidence indicates that typical telomere attrition is low to absent in young adults (13), a conclusion consistent with the findings of no telomere attrition in our control sample of college students. Together, these findings raise confidence that the annual telomere attrition identified in interns is substantially accelerated compared to the expected attrition in young adults. Finally, this study measured telomere length from saliva, which is considered a relatively new measure compared to blood. However, there is good correlation between saliva and blood telomere length measurements (39) and previous literature suggests telomeres isolated from saliva samples show consistent results with stress exposure compared to samples isolated from blood (18).

In summary, the present study identified telomere attrition with residency training and found that extended work hours were associated with accelerated rates of attrition. These results underscore the need to improve resident health and well-being during training. Further, the findings identify a quantifiable, objective biological marker that holds the promise to track physician stress and, potentially, stress more generally.

Supplementary Material

2

Supplemental Table1 S1. Telomere length change by work hour quartile