Occupational characteristics and epigenetic aging among older adults in the United States

Figures & data

Figure 1. Sample selection diagram. Data are from the health and retirement study. VBS = Venous blood study.

Table 1. Summary statistics of analytic sample: percent or mean (SD).

	Mean (SD) or %	Percent Missing
Number of respondents	1,251
Chronological age in 2016	62.4 (3.8)	0
Epigenetic clocks in 2016		0
Horvath	59.7 (5.8)
Hannum	61.2 (6.3)
PhenoAge	59.5 (6.8)
GrimAge	72.1 (4.8)
Dunedin PACE	0.99 (0.14)
Current occupational category in 2010		<1
Professional/managerial	43.8
Sales/clerical	24.8
Service	14.1
Manual	17.3
Self-reported occupational characteristics in 2010
High stress	21.7	<1
High physical effort	17.1	<1
Long hours (55+ hours per week)	10.3	1.3
Covariates
Sex		0
Male	49.0
Female	51.0
Race/ethnicity		<1
Non-Hispanic White	79.3
Non-Hispanic Black	9.1
Hispanic	8.1
Non-Hispanic Other	3.5
Educational attainment		<1
Less than high school	36.8
High school or equivalent	10.7
Some college	46.7
College or more	5.8
Smoking status in 2016		<1
Never smoker	45.2
Former smoker	43.3
Current smoker	11.5
Alcohol consumption in 2016		<1
Non-drinker	29.0
Moderate drinker	49.3
More than moderate drinker	21.7
Physical activity in 2016		<1
Active	91.6
Inactive	8.4
Body Mass Index in 2016		<1
Underweight/Healthy weight ($<$25)	23.6
Overweight (25 to $<$30)	39.1
Obese class I (30 to $<$35)	24.2
Obese class II/III ($\ge$35)	13.1

Note: Data are from the Health and Retirement Study. The sample is restricted to participants aged 51–64 in 2010 who reported working for pay in a non-military occupation in the 2010 HRS core wave and who subsequently participated in the 2016 Venous Blood Study (VBS) and were part of the DNA methylation assay subsample. Summary statistics are weighted with VBS DNA methylation sample weights.

Figure 2. Associations between current occupation category in 2010 and epigenetic age measures in 2016. The displayed effect sizes are the coefficients from linear regressions along with the 95% confidence intervals around these coefficients. The reference occupational category is managerial/professional and the coefficients for the other categories indicate the estimated difference in the epigenetic age measure between that category and managerial/professional after accounting for the other included covariates. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Figure 3. Associations between self-reported working conditions in 2010 and epigenetic age measures in 2016. The displayed effect sizes are the coefficients from linear regressions along with the 95% confidence intervals around these coefficients. These self-reported working conditions are all binary measures and the coefficients refer to the estimated difference in a measure between individuals with and without this occupational characteristic after accounting for the other included covariates. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Figure 4. Joint association between current occupation category and self-reported working conditions in 2010 and epigenetic age measures in 2016. The displayed effect sizes are the coefficients from linear regressions along with the 95% confidence intervals around these coefficients. The reference occupational category is managerial/professional and the coefficients for the other categories indicate the estimated difference in the epigenetic age measure between that category and managerial/professional after accounting for the other included covariates. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Table A1. Associations between occupational category and epigenetic age measures.

	PCHorvath			PCHannum			PCPhenoAge			PCGrimAge			DunedinPACE
	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3
Managerial/
Professional (reference)
Sales/Clerical	1.025**	0.835*	0.831*	1.327***	1.120**	1.141**	1.551***	0.681	0.716^+	0.856***	0.233	0.256	0.033**	0.006	0.006
p-value	0.005	0.029	0.028	<0.001	0.007	0.005	<0.001	0.112	0.08	<0.001	0.384	0.261	0.001	0.584	0.511
Service	0.44	0.717	0.518	0.883^+	0.977^+	0.759	2.442***	1.101*	0.761	1.546***	0.879**	0.616*	0.05***	−0.001	−0.007
p-value	0.319	0.129	0.27	0.062	0.057	0.135	<0.001	0.038	0.131	<0.001	0.008	0.028	<0.001	0.921	0.57
Manual	0.342	0.038	0.009	0.441	0.089	0.158	1.913***	0.564	0.616	1.993***	1.049**	0.384	0.052***	0.005	−0.003
p-value	0.419	0.935	0.985	0.334	0.861	0.756	<0.001	0.282	0.222	<0.001	0.001	0.171	<0.001	0.681	0.799
Controls for age & sex?	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Controls for race & education?		X	X		X	X		X	X		X	X		X	X
Controls for health			X			X			X			X			X
behaviors in 2016?
R-squared	0.242	0.277	0.3	0.248	0.266	0.297	0.3	0.334	0.407	0.469	0.494	0.645	0.036	0.119	0.259
AIC	8,036	7,988	7,964	8,219	8,201	8,164	8,324	8,276	8,145	7,131	7,084	6,656	−947	−1,048	−1,249
N	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251

⁺p < 0.10, * p < 0.05, ** p < 0.01, *** p <0.001. Results are pooled across 10 imputations. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Table A2. Associations between self-reported working conditions and epigenetic age measures.

	PCHorvath			PCHannum			PCPhenoAge			PCGrimAge			DunedinPACE
	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3
High stress	−0.175	−0.294	−0.413	−0.276	−0.337	−0.448	−0.248	0.029	−0.227	0.473^+	0.582*	0.178	0.012	0.022*	0.008
p-value	0.623	0.401	0.236	0.472	0.377	0.238	0.542	0.941	0.545	0.061	0.017	0.388	0.226	0.018	0.339
High physical effort	0.097	0.133	0.193	0.559	0.515	0.58	0.477	−0.384	−0.319	0.564*	0.031	−0.025	0.027*	0	−0.002
p-value	0.802	0.733	0.616	0.182	0.226	0.165	0.277	0.378	0.44	0.04	0.908	0.912	0.01	0.968	0.81
Long working hours	0.26	0.109	0.092	0.255	0.151	0.112	−0.087	0.177	0.169	0.518	0.615^+	0.26	0.001	0.009	0.001
p-value	0.592	0.82	0.847	0.628	0.774	0.83	0.873	0.739	0.741	0.131	0.063	0.363	0.969	0.496	0.916
Controls for age & sex?	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Controls for race & education?		X	X		X	X		X	X		X	X		X	X
Controls for health			X			X			X			X			X
behaviors in 2016?
R-squared	0.237	0.274	0.298	0.243	0.262	0.293	0.281	0.331	0.406	0.449	0.492	0.644	0.018	0.123	0.259
AIC	8,043	7,994	7,968	8,229	8,208	8,170	8,359	8,280	8,148	7,178	7,087	6,660	−925	−1,054	−1,248
N	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251

⁺p < 0.10, * p < 0.05, ** p < 0.01, *** p <0.001. Results are pooled across 10 imputations. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Table A3. Joint association between current occupation category and self-reported working conditions in 2010 and epigenetic age measures.

	PCHorvath			PCHannum			PCPhenoAge			PCGrimAge			DunedinPACE
	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3
Managerial/ Professional (reference)
Sales/ Clerical	1.036**	0.813*	0.789*	1.298**	1.078*	1.079**	1.57***	0.731^+	0.736^+	0.948***	0.334	0.300	0.033**	0.008	0.007
p-value	0.005	0.034	0.039	0.001	0.010	0.009	<0.001	0.090	0.074	<0.001	0.214	0.192	0.001	0.424	0.446
Service	0.439	0.677	0.449	0.781	0.867^+	0.62	2.489***	1.209*	0.831	1.626***	0.968**	0.659*	0.049***	0.001	−0.006
p-value	0.332	0.158	0.346	0.109	0.096	0.230	<0.001	0.025	0.105	<0.001	0.004	0.020	<0.001	0.951	0.636
Manual	0.335	−0.013	−0.079	0.33	−0.034	0.000	1.963***	0.678	0.678	2.087***	1.166***	0.440	0.052***	0.008	−0.002
p-value	0.443	0.978	0.868	0.484	0.947	1.000	<0.001	0.204	0.188	<0.001	<0.001	0.124	<0.001	0.522	0.899
High stress	−0.089	−0.249	−0.374	−0.166	−0.279	−0.386	0.075	0.127	−0.130	0.744**	0.691**	0.235	0.02*	0.023*	0.009
p-value	0.804	0.479	0.288	0.669	0.467	0.313	0.852	0.749	0.730	0.003	0.005	0.258	0.049	0.015	0.336
High physical effort	−0.013	0.094	0.18	0.403	0.467	0.55	−0.177	−0.548	−0.448	0.009	−0.171	−0.126	0.013	0.000	−0.001
p-value	0.975	0.813	0.647	0.352	0.280	0.196	0.693	0.216	0.287	0.974	0.535	0.585	0.253	0.981	0.891
Long working hours	0.335	0.161	0.152	0.35	0.222	0.189	−0.039	0.181	0.185	0.516	0.572+	0.251	0.001	0.009	0.002
p-value	0.490	0.736	0.751	0.506	0.673	0.716	0.943	0.734	0.717	0.125	0.084	0.379	0.923	0.480	0.858
Controls for age & sex?	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Race & education?		X	X		X	X		X	X		X	X		X	X
Health behaviours in 2016?			X			X			X			X			X
R-squared	0.242	0.277	0.301	0.249	0.267	0.298	0.301	0.335	0.408	0.475	0.499	0.646	0.040	0.124	0.260
AIC	8,042	7,994	7,969	8,224	8,205	8,168	8,330	8,280	8,149	7,125	7,078	6,660	−947	−1,049	−1,244
N	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251	1,251

⁺p < 0.10, * p < 0.05, ** p < 0.01, *** p <0.001. Results are pooled across 10 imputations. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Table B1. Summary statistics of longest-reported occupational category in analytic sample.

Number of respondents	1,246
Longest-reported occupational category in 2010^a
Professional/managerial	44.0%
Sales/clerical	24.6%
Service	11.6%
Manual	19.8%
Military	<1%

Note: Data are from the Health and Retirement Study. The sample is restricted to participants aged 51–64 in 2010 who reported working for pay in the 2010 HRS core wave and who subsequently participated in the 2016 Venous Blood Study (VBS) and were part of the DNA methylation assay subsample. Individual earnings are constructed by RAND and already contain imputations for missing values. Summary statistics are weighted with VBS DNA methylation sample weights. 5.4% of respondents were missing longest-reported occupation category. Missing values were imputed using multiple imputation before running the regression analyses. Respondents reporting a military occupation as their longest-held job were excluded because military occupations do not neatly fall into these categories and there are too few respondents with a history of military work for it to be a distinct category.

Table B2. Current occupational category by longest-reported occupational category.

	Longest-reported occupation category in 2010 wave
	Managerial/professional	Sales/clerical	Service	Manual
Current occupation category in 2010 wave
Managerial/professional	9.1	10.2	4.2	5.5
Sales/clerical	5.6	81.0	10.3	6.9
Service	1.8	6.6	77.7	11.7
Manual	2.4	2.2	7.8	76.0
Total	100%	100%	100%	100%

Table B3. Associations between longest-reported occupational category and epigenetic age measures.

	PCHorvath			PCHannum			PCPhenoAge			PCGrimAge			DunedinPACE
	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3
Managerial/
Professional (reference)
Sales/Clerical	0.612	0.508	0.421	1.121**	0.993*	0.934*	1.784***	1.026*	0.957*	0.669*	0.171	0.097	0.031**	0.007	0.007
Service	0.339	0.598	0.322	1.084*	1.171*	0.916^+	2.792***	1.504*	1.052^+	1.765***	1.138**	0.57^+	0.058***	0.008	−0.003
Manual	0.204	−0.143	−0.348	0.272	−0.121	−0.304	2.24***	0.956^+	0.584	1.968***	1.068**	0.33	0.066***	0.019	0.002
Controls for age & sex?	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Controls for race & education?		X	X		X	X		X	X		X	X		X	X
Controls for health			X			X			X			X			X
behaviors in 2016?
N	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246	1,246

⁺p < 0.10, * p < 0.05, ** p < 0.01, *** p <0.001. Results are pooled across 10 imputations. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Table C1. Summary of job demands and job control questionnaire items.

Variable	Question wording	Possible responses	Coding
Job demands items
Time pressure	I am under constant time pressure due to a heavy workload.	1 = strongly disagree 2 = disagree 3 = agree 4 = strongly agree	Kept original coding scale
Working fast	Considering the things I have to do at work, I have to work very fast.	1 = strongly disagree 2 = disagree 3 = agree 4 = strongly agree	Kept original coding scale
Conflicting demands	In my work I am free from conflicting demands that others make.	1 = strongly disagree 2 = disagree 3 = agree 4 = strongly agree	Reverse coded such that 1 = strongly agree and 4 = strongly disagree.
Job control items
Freedom over work	I have very little freedom to decide how I do my work.	1 = strongly disagree 2 = disagree 3 = agree 4 = strongly agree	Reverse coded such that 1 = strongly agree and 4 = strongly disagree.
Control over what happens	At work, I feel I have control over what happens in most situations.	1 = strongly disagree 2 = disagree 3 = agree 4 = strongly agree	Kept original coding scale
Opportunity to develop skills	I have the opportunity to develop new skills.	1 = strongly disagree 2 = disagree 3 = agree 4 = strongly agree	Kept original coding scale

Table C2. Associations between self-reported working conditions and epigenetic age measures in subsample that responded to psychosocial leave-behind questionnaire: High job strain as measure of job stress.

	PCHorvath			PCHannum			PCPhenoAge			PCGrimAge			DunedinPACE
	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3	M1	M2	M3
High job strain	0.427	0.334	0.268	0.54	0.468	0.42	0.834*	0.864*	0.715^+	0.469*	0.461*	0.251	0.023*	0.025**	0.018*
High physical effort	0.603	0.607	0.669	1.256*	1.209*	1.315*	0.766	−0.092	−0.032	0.757*	0.334	0.243	0.018	−0.009	−0.011
Long work hours	0.237	0.084	−0.015	−0.011	−0.112	−0.176	−0.822	−0.392	−0.353	0.216	0.37	−0.059	−0.008	0.006	−0.002
Controls for age & sex?	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Controls for race & education?		X	X		X	X		X	X		X	X		X	X
Controls for health			X			X			X			X			X
behaviors in 2016?
N	886	886	886	886	886	886	886	886	886	886	886	886	886	886	886

⁺p < 0.10, * p < 0.05, ** p < 0.01, *** p <0.001. Results are pooled across 10 imputations. Model 1 adjusts for age and sex. Model 2 additionally includes race/ethnicity and educational attainment. Model 3 additionally includes smoking status, alcohol consumption, physical inactivity, and Body Mass Index (BMI) category.

Data availability statement

The data that support the findings of this study are available from the Health and Retirement Study (HRS). These data are available to registered users who meet security requirements and agree to data use conditions specified by the HRS at https://hrsdata.isr.umich.edu/data-products/conditions-of-use