Abstract
Objectives:
The opioid neuropeptide pro-enkephalin A (PENK-A) may be a circulating marker of cardiovascular risk, with prior findings relevant to heart failure, kidney disease, and vascular dementia. Despite these findings, the association of PENK-A with ischemic stroke is unknown, so we examined this association in a prospective cohort study and analyzed differences by race and sex.
Materials and Methods:
The REasons for Geographic and Racial Differences in Stroke study (REGARDS) is a prospective cohort study of 30,239 Black and White adults. Plasma PENK-A was measured in 473 participants that developed first-time ischemic stroke over 5.9 years and 899 randomly selected participants. Cox models adjusted for demographics and stroke risk factors were used to calculate hazard ratios (HRs) of stroke by baseline PENK-A.
Results:
PENK-A was higher with increasing age, female sex, White race, lower body mass index, and antihypertensive medication use. Each SD higher increment of PENK-A was associated with an adjusted HR of 1.20 (95% CI 1.01–1.42) for stroke, with minimal confounding by stroke risk factors. Spline plots suggested a U-shaped relationship, particularly in White men, with an adjusted HR 3.88 (95% CI 1.94–7.77) for the 95th versus 50th percentile of PENK-A in White men.
Conclusions:
Higher baseline plasma PENK-A was independently associated with future stroke risk in REGARDS. This association was most apparent among White men. There was little confounding by established stroke risk factors, suggesting a possible causal role in stroke etiology. Further research is needed to understand the role of endogenous opioids in stroke pathogenesis.
Keywords: pro-enkephalin, stroke risk, REGARDS, cohort, epidemiology, risk factors
INTRODUCTION
Ischemic stroke is a leading cause of death and disability.1 Stroke care has seen vast improvements in diagnosis and treatment in recent years, yet efforts to identify novel biomarkers of increased risk for future stroke have yielded few results.2,3 The early detection of individuals at high stroke risk is important in implementing preventative measures.4
Insights on stroke mechanism can be gained by studying biomarkers known to relate to cardiovascular disease in general. Plasma measurement of the endogenous neuropeptide fragment pro-enkephalin A (PENK-A) has previously shown promise as a cardiovascular disease biomarker. Prior studies implicate higher circulating PENK-A as a risk biomarker for vascular dementia, kidney function decline, and heart failure.5–10 Pro-enkephalin is a precursor peptide to the enkephalin opioid family which are produced in the central nervous system, heart, adrenal glands, and kidneys.11,12 The enkephalins are involved in several physiologic processes, including analgesia, immunity, and autonomic regulation, with the latter two being important causal determinants of cardiovascular disease.13–16 Endogenous opioid receptors on cardiac myocytes increase cardiac contractility through interactions with β-adrenergic receptor signaling, and circulating enkephalins are important protective mediators against oxidative stress.17–20 Oxidative stress is highly relevant to ischemic stroke pathogenesis.21–23 However, endogenous opioid pathways are complex and the cardiovascular role of the enkephalins remain incompletely understood.
Despite these prior cardiovascular-related findings, the association between baseline plasma PENK-A and stroke risk is unknown. Given the importance of cardiovascular health to stroke pathogenesis, we assessed the association of PENK-A and incident ischemic stroke in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study. Additionally, in line with the goal of REGARDS to discover potential mediators of health disparities, we assessed whether the association between PENK-A and stroke risk varies by race and/or sex due to the increased stroke burden experienced by Black and female individuals.
MATERIALS AND METHODS
Study Population
The REGARDS study is a prospective, population-based cohort study of 30,239 Black and White individuals aged 45 years and older. Participants were recruited from January 2003 to October 2007 through a combination of mail and telephone contacts. Black individuals and those living in the US Southeastern “stroke belt” were oversampled. Examiners conducted computer-assisted telephone interviews to collect participant information. Written informed consent, blood pressure, physical measurements, electrocardiogram (ECG) testing, and medication inventory were obtained during a subsequent in-home examination.24 Institutional review boards at each participating site reviewed and approved the study methods.
Measurements and Definitions
Hypertension was defined as systolic blood pressure ≥140 mm Hg, diastolic pressure ≥90 mm Hg, or self-reported use of antihypertensive medication. Diabetes mellitus was defined by fasting blood glucose ≥126 mg/dL (or a non-fasting glucose ≥200 mg/dL among those failing to fast) or self-reported use of antidiabetic medication. Atrial fibrillation was defined as self-report or based on ECG finding. Coronary artery disease was defined as self-reported physician diagnosis of a myocardial infarction, bypass, percutaneous coronary intervention, or based on ECG detection. Left ventricular hypertrophy was based on ECG finding. Pre-baseline stroke and TIA were defined based on self-report of a physician diagnosis.
Stroke Ascertainment
Participants were contacted by telephone every six months to collect health information. Medical records were obtained where there was a suspected of stroke or TIA, or in the event of death. Following pre-review by a trained stroke clinician, medical records associated with suspected stroke events were reviewed and adjudicated by a committee of stroke physicians. Ischemic stroke was defined as focal neurological symptoms lasting ≥24 hours, or non-focal symptoms with positive imaging for ischemic stroke.25 Ischemic strokes were sub-classified according to the TOAST criteria.26 Stroke severity, as defined through medical record review, was defined as the National Institutes of Health Stroke Scale (NIHSS) score recorded by an attending neurologist upon hospital admission. Hemorrhagic stroke was also ascertained but not studied here. Stroke severity was defined using the clinically determined NIHSS score at hospital admission: mild, ≤ 5; moderate, 6–11; severe, ≥12.
Case-Cohort Sample
We used a case-cohort study design with complete follow up to evaluate associations of biomarkers with stroke risk.27,28 The case-cohort design was selected to efficiently study PENK-A and its relationship to stroke, and to provide results that would approximate PENK-A measurement in the entire cohort. Median follow-up time was 5.9 years (interquartile range [IQR] 4.3–7.0). Cases were 518 participants without pre-baseline stroke or TIA that developed ischemic stroke during follow-up. The cohort random sample (N = 1127) was selected using age-, sex-, and race-based stratified sampling (50% Black, 50% White, 50% women, 50% men, age groups: 20% 45–54, 20% 55–64, 25% 65–74, 25% 75–84, and 10% ≥85 years). 152 participants in the random sample (13.5%) were excluded due to pre-baseline stroke or TIA (Supplemental Figure 1). Of the remaining 975 included participants in the random sample, 74 (7.9%) developed ischemic stroke during follow-up.
Laboratory Methods
Baseline morning blood samples were obtained from participants using standardized methods. Samples were processed within 120 minutes of phlebotomy, serum and plasma were shipped overnight on ice to a central biorepository, and stored at −80°C until batch analysis.29 Plasma PENK-A was measured using a chemiluminometric sandwich immunoassay. This assay detects endogenous PENK-A fragments using a stable solid phase and tracer antibodies. PENK-A measurement is the current clinically preferred assay of enkephalin activity due to its superior plasma stability compared to other volatile cleavage products like met- and leu-enkephalin.30 The analytical coefficient of variation range was 0.6–7.4%. Assays were not completed in 45 case participants and 76 cohort random sample participants due to lack of available plasma (Supplemental Figure 1).
Statistical Methods
A skewed distribution of PENK-A was observed in the cohort random sample by Q-Q plot visualization. Therefore, values were natural log transformed for all continuous analyses. Potential differences in ln(PENK-A) levels by baseline fasting status or concomitant opioid use were examined using Mann-Whitney tests in the cohort random sample.
Baseline characteristics are presented as percent and median (IQR) across quartiles of PENK-A in the cohort random sample. Differences in risk factors across PENK quartiles were assessed by Pearson or Kruskal-Wallis tests. For parsimonious evaluation, stroke risk factors that varied across PENK-A quartiles (P < 0.05) were further evaluated for independent correlation with continuous PENK-A using linear regression models.
We examined the association of PENK-A and ischemic stroke risk using Cox proportional hazard models. Inverse probability weighting was applied to all models to account for the case-cohort study design and reflect the entire REGARDS cohort.31 PENK-A was divided into quartiles, with the lowest quartile serving as the referent group. The first model was unadjusted. The second model was minimally adjusted for race, sex, age, and age-by-sex interaction. A third model added geographic region of residence, annual income, education, and Framingham stroke risk factors: smoking status, diabetes mellitus, antihypertensive medication use, systolic blood pressure, atrial fibrillation, left ventricular hypertrophy, and coronary artery disease. The final model added baseline use of statins, NSAIDs, and aspirin. Participants in the cohort random sample that developed non-ischemic stroke during follow-up (N = 5) were censored at the date of non-ischemic stroke. While prior studies reported an inverse relationship between PENK-A and renal function (as measured by estimated glomerular filtration rate [eGFR]),7,8 we did not adjust for baseline renal function because previous work in REGARDS did not demonstrate an association between lower eGFR and stroke risk.32 Therefore, renal function as measured by eGFR could not be a confounder in these analyses.
Differences in association by race and sex was examined by adding either a race-by-ln(PENK-A) or sex-by-ln(PENK-A) interaction term. Significance for interaction was defined as P <0.10. Non-linear associations between PENK and stroke risk were examined using Cox proportional hazard models with restricted cubic splines with five knots. Sequential multivariable adjustments matched the models above.
In analyses restricted to those with incident stroke events, we evaluated the association of baseline ln(PENK-A) with characteristics of the strokes. Differences in baseline ln(PENK-A) concentration by incident stroke subtype were examined using a Kruskal-Wallis test. Additionally, ordinal logistic regression was performed with stroke severity as the outcome to assess the association between baseline PENK-A and stroke severity. Only stroke cases with an available NIHSS score (N = 339) were included in this analysis. Multivariable adjustments matched the models above.
All analyses were completed in R, with use of the survival package (version 3.6.3, R Foundation).33,34
Data Availability
Data used in these analyses are not freely publicly available due to legal and ethical restrictions. However, qualified investigators may request access to de-identified data from the University of Alabama at Birmingham (regardsadmin@uab.edu).
RESULTS
Participant Characteristics
During a median follow up of 5.9 years (interquartile range [IQR] 4.4–7.2), 547 participants developed ischemic stroke. There were 473 participants with measured PENK-A in the stroke case group, and 899 in the cohort random sample (Supplemental Figure 1). The median time to stroke was 2.7 years (IQR 1.4–4.1). There were no appreciable differences between participants with and without PENK-A measurement due to lack of available plasma. There was no difference in PENK-A concentration by fasting status or baseline opioid medication use in the cohort random sample (data not shown).
Cross-Sectional Association of PENK-A with Stroke Risk Factors
Table 1 shows the cross-sectional association of PENK-A quartiles with participant characteristics in the cohort random sample. In bivariate analyses, higher PENK-A was associated with older age, female sex, White race, lower body mass index, lower annual income, non-smoking status, coronary artery disease, antihypertensive medication use (but not systolic blood pressure), and statin use. Considering these variables, in a multivariable model, PENK-A remained positively correlated with older age, female sex, White race, lower body mass index, and antihypertensive use (Table 2).
Table 1:
Baseline characteristics of the cohort random sample by PENK-A quartile.
| Characteristic | Plasma PENK-A in cohort random sample | ||||
|---|---|---|---|---|---|
| Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | P-value | |
| N = 227 | N = 224 | N = 223 | N = 225 | ||
| <48.2 | 48.2–59.3 pmol/L | 59.4–74.1 pmol/L | ≥74.2 | ||
| pmol/L | pmol/L | ||||
| Demographics | |||||
| Age, years – median (IQR) | 59 (52–68) | 64 (55–72) | 70 (61–77) | 76 (67–84) | <0.001 |
| Female sex – no. (%) | 103 (45.4) | 109 (48.7) | 105 (47.1) | 137 (60.9) | 0.004 |
| Black race – no. (%) | 147 (64.8) | 113 (50.4) | 98 (43.9) | 82 (36.4) | <0.001 |
| Body mass index, kg/m2 – median (IQR) | 31 (28–36) | 28 (25–32) | 27 (24–31) | 26 (23–29) | <0.001 |
| Waist circumference, cm – median (IQR) | 101 (91–113) | 95 (86–103) | 93 (86–102) | 90 (81–99) | <0.001 |
| Region – no. (%) | 0.8 | ||||
| Stroke buckle | 39 (17.2) | 41 (18.3) | 40 (17.9) | 39 (17.3) | |
| Other stroke belt | 83 (36.6) | 80 (35.7) | 77 (34.5) | 67 (29.8) | |
| Other US regions | 105 (46.2) | 103 (46.0) | 106 (47.6) | 119 (52.9) | |
| Education Level – no. (%) | 0.07 | ||||
| Less than high school | 32 (14.1) | 25 (11.2) | 35 (15.7) | 35 (15.6) | |
| High school graduate | 59 (26.0) | 50 (22.3) | 40 (17.9) | 57 (25.3) | |
| Some college | 70 (30.8) | 52 (23.2) | 66 (29.6) | 58 (25.8) | |
| College graduate | 66 (29.1) | 97 (43.3) | 82 (36.8) | 75 (33.3) | |
| Annual household income* – no. (%) | 0.01 | ||||
| <$20,000 | 34 (15.0) | 39 (17.4) | 32 (14.3) | 58 (25.8) | |
| $20,000–34,999 | 52 (22.9) | 56 (25.0) | 50 (22.4) | 55 (24.4) | |
| $35,000–74,999 | 71 (31.3) | 65 (29.0) | 61 (27.4) | 63 (28.0) | |
| ≥$75,000 | 35 (15.4) | 41 (18.3) | 43 (19.3) | 19 (8.4) | |
| Comorbidities – no. (%) | |||||
| Diabetes mellitus | 64 (28.3) | 33 (14.7) | 40 (17.9) | 46 (20.4) | 0.003 |
| Hypertension | 170 (74.9) | 151 (67.4) | 168 (75.3) | 170 (75.6) | 0.3 |
| Systolic Blood Pressure, mmHg – median (IQR) | 125 (118–137) | 125 (117–136) | 128 (119–140) | 125 (117–137) | 0.2 |
| Coronary Artery Disease | 26 (11.5) | 32 (14.3) | 34 (15.2) | 47 (20.9) | 0.04 |
| Left ventricular hypertrophy | 20 (8.8) | 16 (7.1) | 19 (8.5) | 24 (10.7) | 0.6 |
| Atrial fibrillation | 17 (7.5) | 17 (7.6) | 17 (7.6) | 24 (10.7) | 0.5 |
| Alcohol use | 0.002 | ||||
| Never | 59 (26.0) | 56 (25.0) | 69 (30.9) | 79 (35.1) | |
| Past | 59 (26.0) | 35 (15.6) | 31 (13.9) | 42 (18.7) | |
| Current | 109 (48.0) | 133 (59.4) | 123 (55.2) | 104 (46.2) | |
| Smoking status* | 0.01 | ||||
| Never | 95 (41.9) | 111 (49.6) | 116 (52.0) | 110 (48.9) | |
| Past | 84 (37.0) | 88 (39.3) | 71 (31.8) | 91 (40.4) | |
| Current | 48 (21.1) | 24 (10.7) | 34 (15.2) | 24 (10.7) | |
| Physical activity frequency | 0.2 | ||||
| None | 81 (35.8) | 65 (29.4) | 72 (32.6) | 92 (41.4) | |
| 1–3 per week | 77 (34.1) | 90 (40.7) | 90 (36.2) | 71 (32.0) | |
| ≥4 or more per week | 68 (30.1) | 66 (29.9) | 69 (31.2) | 59 (26.6) | |
| Medication use – no. (%) | |||||
| Statins | 55 (24.2) | 50 (22.3) | 75 (33.6) | 78 (34.7) | 0.004 |
| Antihypertensives | 106 (46.7) | 89 (39.7) | 118 (52.9) | 133 (59.1) | <0.001 |
| NSAIDs | 39 (17.2) | 25 (11.2) | 23 (10.3) | 35 (15.6) | 0.09 |
| Lipids – median (IQR) | |||||
| Cholesterol, mg/dL | 186 (163–214) | 196 (172–220) | 186 (159–205) | 188 (158–213) | 0.002 |
| HDL, mg/dL | 48 (39–57) | 50 (42–62) | 49 (40–63) | 50 (39–63) | 0.3 |
| LDL, mg/dL | 112 (90–138) | 117 (96–140) | 106 (88–129) | 106 (85–130) | 0.004 |
Abbreviations: HDL, high density lipoprotein; LDL, low density lipoprotein; NSAIDs, non-steroidal anti-inflammatory drugs.
Column percentages may not add to 100% due to missing participant data.
Table 2:
Multivariable associations of participant characteristics with baseline PENK-A*.
| Characteristic | Adjusted difference in | 95% CI |
|---|---|---|
| ln(PENK-A), | ||
| SD ln(pmol/L) | ||
| Age, per 10-year increase | 0.31 | 0.24, 1.45 |
| Female vs. male sex | 0.31 | 0.24, 0.39 |
| White vs. Black race | 0.4 | 0.23, 0.60 |
| BMI, per 1 kg/m2 increase | −0.05 | −0.06, −0.04 |
| Annual income, thousands | ||
| <$20 vs. ≥$75 | 0.1 | −0.12, 0.36 |
| $20–34 vs. ≥$75 | 0.08 | −0.11, 0.32 |
| $35–74 vs. ≥$75 | −0.04 | −0.19, 0.16 |
| Coronary artery disease | 0.02 | −0.15, 0.22 |
| Current smoker | −0.01 | −0.18, 0.19 |
| Antihypertensive use | 0.18 | 0.04, 0.36 |
| Statin use | 0.08 | −0.07, 0.25 |
Variables included in this model were univariately associated with baseline PENK-A in the cohort random sample. Differences are expressed as standard deviation units of natural-log transformed PENK-A. The mean ln(PENK-A) value in the cohort random sample was 4.11 with a standard deviation of 0.37 ln(pmol/L).
Associations of PENK-A with Incident Ischemic Stroke
Table 3 shows the hazard ratios (HR) of stroke by baseline PENK-A quartile. In an unadjusted model, stroke risk increased with increasing PENK-A such that the fourth vs. first quartile had a 2.24-fold increased risk of stroke (95% CI 1.58–3.16). After adjusting for demographics, this HR was attenuated to only a 35% increased risk and was no longer statistically significant; most of this attenuation was due to confounding by age (not shown). Further adjustment for both established stroke risk factors and medications minimally affected this relationship. Findings were similar in race and sex subgroups, and when race/sex specific quartile definitions were used (Supplemental Tables 1–4). Because of the large attenuation by age, we tested for statistical interaction of age and PENK-A, each considered continuously in a model adjusting only for age, and the P value for interaction was 0.7.
Table 3:
Cox proportional hazard models for incident ischemic stroke.
| Hazard ratio (95% CI) for ischemic stroke* | ||||||
|---|---|---|---|---|---|---|
| Model | Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | P trend | Hazard ratio per 1 SD higher |
| N = 313 | N = 342 | N = 355 | N = 394 | |||
| 1 | 1 (ref) | 1.21 (0.84, 1.72) | 1.31 (0.92, 1.85) | 2.24 (1.58, 3.16) | <0.001 | 1.32 (1.16, 1.49) |
| 2 | 1 (ref) | 1.05 (0.73, 1.51) | 0.96 (0.66, 1.38) | 1.35 (0.92, 1.99) | 0.2 | 1.14 (0.99, 1.34) |
| 3 | 1 (ref) | 1.04 (0.65, 1.65) | 0.92 (0.58, 1.46) | 1.40 (0.86, 2.27) | 0.2 | 1.19 (1.00, 1.41) |
| 4 | 1 (ref) | 1.08 (0.68, 1.73) | 1.00 (0.62, 1.60) | 1.44 (0.88, 2.35) | 0.2 | 1.20 (1.01, 1.42) |
Model 1 was unadjusted. Model 2 was adjusted for age, sex, race, and age-by-sex interaction. Model 3 added geographic region of residence, annual income, education, and Framingham stroke risk factors to Model 2. Model 4 added baseline use of statins, NSAIDs, or aspirin to Model 3. All models were weighted to reflect the entire REGARDS cohort. SD, standard deviation (0.37 ln(pmol/L)).
PENK-A was then considered continuously after natural log transformation (Table 3). Higher PENK-A was similarly associated with increased stroke risk in an unadjusted model (HR 1.32 per 1 SD ln(PENK-A) higher, 95% CI 1.16–1.49). As seen in the categorical analysis, adjustment for demographics attenuated this association (HR 1.15, 95% CI 0.99–1.34). Further adjustment for stroke risk factors accentuated the association (HR 1.20, 95% CI 1.01–1.42), and there was no impact of added adjustment for baseline medications. Interaction terms for race or sex with PENK-A were not significant in any of the models (all P >0.5).
To address the discrepancy between results of the categorical and continuous adjusted models, we tested for non-linear association between PENK-A and incident stroke using restricted cubic spline plots (Figure 1). An unadjusted spline model showed that high PENK-A was associated with increased stroke risk with up to an estimated hazard ratio of 2.01 (95% CI 1.54–2.64) for the PENK-A concentration corresponding to the 95th percentile, compared to the median value (Figure 1A). Adjustment for demographic factors slightly attenuated this risk (Figure 1B). This attenuation was again largely due to adjustment for age (Supplemental Figure 2). In this analysis a possible U-shaped association was revealed, with a trend towards increased stroke risk at low levels of PENK-A (HR 1.23 [95% CI 0.86–1.77] for the PENK-A value corresponding to the 5th percentile, compared to the median, and a P value for non-linear relationship =0.06). Further adjustment for stroke risk factors (Figure 1C) and medications (Figure 1D) minimally impacted stroke risk for both high and low PENK-A levels (fully adjusted HR 1.48 [95% CI 1.07–2.05] for the PENK-A value corresponding to the 95th percentile, compared to the median, and P for a non-linear relationship = 0.1). These associations were robust to varying spline knot sets.
Figure 1. Non-linear associations of PENK-A with stroke risk.
Incident stroke risk (line) was modeled as a restricted cubic spline term and plotted as hazard relative to the median PENK-A value. All PENK-A values are log transformed and relative hazards are plotted on a log scale. Vertical ticks on the curve represent quartile bounds shading represents 95% confidence intervals, and the lower curve lines represent PENK-A distribution. Model 1 is unadjusted. Model 2 is adjusted for age, sex, race, and age-by-sex interaction. Model 3 adds geographic region of residence, annual income, education, and Framingham stroke risk factors. Model 4 adds baseline use of statins, NSAIDs, and aspirin.
Sex- and race-specific models using restricted cubic splines were also plotted (Figure 2A–D). There was no overall difference in the association by race considering any of the adjusted models (all P interaction >0.2). An interaction term for sex and PENK-A was significant in the minimally adjusted non-linear model (Model 1, P = 0.02), but not after further adjustments (P = 0.2 in Model 4). When considering only White participants, the sex-by-ln(PENK-A) interaction was significant in all models (P = 0.03 for fully adjusted model) Figure 2A–D shows that the reason for these interactions was a U-shaped association apparent in White men and Black women (P for non-linear association 0.003 and 0.04, respectively), which was not seen in Black men and White women. The association of higher PENK-A with stroke was largest in White men (fully adjusted HR 3.88 [95% CI 1.94–7.77] for 95th percentile versus median concentration.
Figure 2. Non-linear association of PENK-A with stroke risk in subgroups.
Incident stroke risk (line) was modeled as a restricted cubic spline term and plotted as hazard relative to the median PENK-A value after adjustment for a full set of covariates. All PENK-A values are log transformed and relative hazards are plotted on a log scale. Vertical ticks on the curve represent subgroup specific quartile bounds, shading represents 95% confidence intervals, and the lower curve lines represent relative PENK-A distribution.
Associations of PENK-A with Stroke Subtype and Severity
Figure 3 shows that baseline ln(PENK-A) concentration did not differ by incident stroke subtype (Kruskal-Wallis test P = 0.8).
Figure 3. Association of PENK-A concentration and stroke subtype.
Bars represent 5th-95th percentiles with outliers plotted. Units: natural-log transformed pmol/L.
Of the 547 participants that developed stroke, 339 (62.0%) had an available admission NIHSS score; 268 (79.0%) developed mild stroke, 49 (14.5%) developed moderate stroke, and 22 (6.5%) developed severe stroke. Ordinal logistic regression models did not reveal an association between PENK-A and higher severity category (unadjusted odds ratio of having one higher stroke severity level was 1.07 per 1 SD increment of ln(PENK-A), 95% CI 0.85–1.33). This was similar with multivariable adjustment. Results did not differ by race or sex (data not shown).
DISCUSSION
This study of community-dwelling Black and White participants showed an independent association between baseline concentration of the endogenous opioid precursor PENK-A with risk of future ischemic stroke. The relationship appeared to be non-linear with increased risk at high and low plasma concentrations. While this relationship was partly attenuated by age, the shape and magnitude of PENK-A’s association with incident stroke was not appreciably influenced by adjustment for traditional stroke risk factors. Similarly, PENK-A was not independently associated with traditional stroke risk factors in cross-sectional analyses. Among race-sex subgroups, PENK-A appeared to be most relevant among White men, where a U-shaped relationship was apparent.
The role of PENK-A as a causal or surrogate biomarker of any disease is unclear, with prior studies positing that high plasma concentrations reflect general cardiovascular risk.5,8–10 Our findings are consistent with this hypothesis and expand upon it through our finding of a potential U-shaped association with stroke risk. Though linear models per log-transformed standard deviation indicated an independent association with stroke risk with higher PENK-A (adjusted OR 1.20, 95% CI 1.01–1.42; Table 3), results from spline plots may indicate non-linear consideration is more informative of PENK-A’s biological role (Figures 1 and 2). While two cohort studies reported cross-sectional associations of PENK-A with diabetes and hypertension,8,10 leading to the possibility that this might explain associations with cardiovascular disease, in this study there was no independent association of diabetes or hypertension with higher plasma PENK-A concentration. No established stroke risk factor was independently associated with higher baseline PENK-A (Table 2). Moreover, multivariable adjustment for cardiovascular risk factors minimally impacted the non-linear association between PENK-A and stroke risk. PENK-A levels in REGARDS were comparable to those in the Malmo Preventive Project, another large population-based cohort that was not racially diverse, (mean 61 (SD 45) in REGARDS vs 64 (SD 27) pmol/L in Malmo Preventive Project).5 This does not speak to the generalizability of our results as the Malmo cohort did not assess stroke and was not racially diverse. Our findings are, however, generalizable to US Black and White adults aged >45 years. These results suggest that the reason for PENK-A’s cardiovascular role is unknown and further work is needed to characterize its cardiovascular impact.
Though the role of PENK-A in human disease pathogenesis is incompletely understood, animal studies describe the enkephalins as important mediators against oxidative damage,12,17,19,35 particularly in cardiac and neuronal tissue.20,36,37 Similarly, endothelial opioid receptor activation is relevant to nitric oxide production, endothelin release, proliferation, and inflammation.38–41 The association between oxidative stress and human cardiovascular disease is well described,23,42 though randomized trials of antioxidant prophylaxis have yielded mixed results in ischemic stroke prevention.22,43,44 Higher baseline plasma PENK-A levels likely represent a physiological response to oxidative stress, which may lead to stroke. In contrast, the non-linear relationship of lower PENK-A with higher stroke risk was not expected. It is possible this association reflects increased susceptibility to oxidative damage with lower PENK-A production, though additional research is needed to contextualize these findings.
While our cross-sectional results indicate that PENK-A likely increases with age, it is currently unclear whether PENK-A concentration may rise temporally leading up to a stroke. Results here of increased stroke risk with higher baseline PENK-A align with prior studies that plasma PENK-A concentration increases with acute ischemic stroke. Doehner et al. reported a median PENK-A concentration of 123.8 (IQR 93.0–160.5) pmol/L at time of ischemic stroke,45 higher than the baseline median of 59.4 (IQR 48.2–74.1) pmol/L in our cohort random sample. Gruber et al. similarly found high PENK-A concentrations at time of stroke.46 While higher PENK-A in acute stroke is likely due to acute effects of stroke, it would be desirable to understand whether PENK-A rises over time before a stroke event, and if these changes reflect potential mechanism(s) by which PENK-A relates to stroke.
Taken together, our quartile-based, linear, and non-linear models revealed that extreme high and low plasma PENK-A concentrations may be most relevant to stroke risk. Many participants had baseline log-transformed concentrations clustered near the median (Figure 1), which likely contributed to the different results across modeling techniques. While a null association was evident in our quartile analysis after multivariable adjustment, our non-linear models indeed show that non-linear consideration of PENK-A was justified to capture the independent risk associated with extreme PENK-A values. No prior studies have examined PENK-A’s association with stroke risk or have examined PENK-A in a non-linear fashion. It is possible that these findings may be spurious; additional research is needed to confirm these findings and to explain the observed non-linear relationship.
Considering these findings and prior cardiovascular results, these data represent a first step towards a potential clinical use for PENK-A as a stroke risk stratification tool. Our cross-sectional and longitudinal findings together may indicate that endogenous opioid signaling marked by PENK-A represents unrecognized pathways to stroke. Clinical testing may inform care providers when abnormalities are noted, but it is unclear which interventions may be taken to attenuate risk as mechanisms are yet unknown. Much additional research is needed to characterize the prognostic characteristics of PENK-A testing and to determine potential therapeutic targets.
The sex and race subgroup analyses presented above were performed a priori in line with the goal of REGARDS investigators to discover causes of stroke risk disparities. As PENK-A’s association with stroke risk appeared most relevant to White men, these findings suggest PENK-A is an unlikely mediator of these disparities. Yet, the apparent differences in non-linear association by race and sex are scientifically intriguing and merit further investigation. Potential mechanisms underlying the differential associations observed between White men and women are unclear, though this difference was statistically significant (fully adjusted P interaction = 0.03). Additionally, it is also unclear why a similar difference in non-linear association was not observed between Black men and women. It is also possible that these results are due to chance as potential differences in peripheral opioid signaling pathways by either race or sex are poorly characterized. Limited evidence suggest sex differences may exist.47–49 We present these findings to more completely describe PENK-A’s non-linear association with stroke risk and to encourage future research. However, these results should be interpreted cautiously.
This study had several strengths. It had a prospective design and enrolled a large cohort of community-dwelling Black and White adults across the United States. All strokes were rigorously evaluated by a panel of experts. Participant retention was high allowing for substantial follow-up. Additionally, this is the first study to examine PENK-A in a biracial cohort, to test the association of plasma PENK-A concentration with stroke risk, and to evaluate PENK-A in a non-linear fashion. There are several limitations to consider. Stroke classification and severity measures relied upon data obtained from many hospitals, and stroke care documentation was not standardized. Measurement of PENK-A at a single timepoint may not accurately reflect baseline plasma levels, though blood sampling was conducted in a standardized manner for all participants. Finally, these findings may not be generalizable to non-White, non-Black racial groups and may be subject to residual unmeasured confounding.
CONCLUSIONS
In the REGARDS cohort, baseline plasma PENK-A concentration was higher with older age, in White people, and in women. PENK-A was not associated with traditional stroke risk factors. Higher baseline plasma PENK-A concentration was independently associated with stroke risk, and there was also a trend toward increased risk with lower PENK-A. This non-linear association may be most relevant to White men. Future work should focus on understanding mechanisms for the race and sex differences of PENK-A as well as the mechanisms by which PENK-A is associated with stroke risk.
Supplementary Material
HIGHLIGHTS.
Higher circulating PENK-A, an endogenous opioid neuropeptide, was independently associated with ischemic stroke risk in REGARDS
PENK-A’s association with stroke was non-linear
Adjustment for established stroke risk factors did not alter the relationship between PENK-A and stroke
PENK-A’s association with stroke may be most relevant to White men
ACKNOWLEGEMENTS
The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at: https://www.uab.edu/soph/regardsstudy/. This research project is supported by cooperative agreement U01 NS041588 co-funded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Service. Additional support was from the National Institute of General Medical Sciences, P20 GM135007 (MC) and the Cardiovascular Research Institute of Vermont at the University of Vermont (Burlington, VT; SS). This content is the sole responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institutes of Health. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis, or interpretation of the data. Assay measurement was provided by SphingoTec GmbH (Hennigsdorf, Germany). Janin Schulte is employed by SphingoTec GmbH, the manufacturer of the PENK-A immunoassay. The remaining authors have nothing to disclose.
Glossary
- CI
confidence interval
- ECG
electrocardiogram
- eGFR
estimated glomerular filtration rate
- HR
hazard ratio
- IQR
interquartile range
- NIHSS
National Institutes of Health Stroke Scale
- PENK-A
pro-enkephalin A
- SD
standard deviation
- REGARDS
REasons for Geographic and Racial Differences in Stroke study
- TOAST
Trial of ORG 10172 in Acute Stroke Treatment
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest
None.
REFERENCES
- 1.Katan M, Luft A. Global Burden of Stroke. Semin Neurol. 2018;38(02):208–211. doi: 10.1055/s-0038-1649503 [DOI] [PubMed] [Google Scholar]
- 2.Hijazi Z, Lindbäck J, Alexander JH, et al. The ABC (age, biomarkers, clinical history) stroke risk score: a biomarker-based risk score for predicting stroke in atrial fibrillation. Eur Heart J. 2016;37(20):1582–1590. doi: 10.1093/eurheartj/ehw054 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kamin Mukaz Debora, Zakai Neil A, Salvador Cruz-Flores, McCullough Louise D, Cushman Mary. Identifying Genetic and Biological Determinants of Race-Ethnic Disparities in Stroke in the United States. Stroke. 2020;51(11):3417–3424. doi: 10.1161/STROKEAHA.120.030425 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Pandian JD, Gall SL, Kate MP, et al. Prevention of stroke: a global perspective. The Lancet. 2018;392(10154):1269–1278. doi: 10.1016/S0140-6736(18)31269-8 [DOI] [PubMed] [Google Scholar]
- 5.Holm H, Nägga K, Nilsson ED, et al. High circulating levels of midregional proenkephalin A predict vascular dementia: a population-based prospective study. Sci Rep. 2020;10(1):8027. doi: 10.1038/s41598-020-64998-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Beunders R, Struck J, Wu AHB, et al. Proenkephalin (PENK) as a Novel Biomarker for Kidney Function. J Appl Lab Med. 2017;2(3):400–412. doi: 10.1373/jalm.2017.023598 [DOI] [PubMed] [Google Scholar]
- 7.Khorashadi M, Beunders R, Pickkers P, Legrand M. Proenkephalin: A New Biomarker for Glomerular Filtration Rate and Acute Kidney Injury. Nephron. 2020;144(12):655–661. doi: 10.1159/000509352 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Schulz C-A, Christensson A, Ericson U, et al. High Level of Fasting Plasma Proenkephalin-A Predicts Deterioration of Kidney Function and Incidence of CKD. J Am Soc Nephrol. 2017;28(1):291–303. doi: 10.1681/ASN.2015101177 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ng LL, Squire IB, Jones DJL, et al. Proenkephalin, Renal Dysfunction, and Prognosis in Patients With Acute Heart Failure: A GREAT Network Study. J Am Coll Cardiol. 2017;69(1):56–69. doi: 10.1016/j.jacc.2016.10.038 [DOI] [PubMed] [Google Scholar]
- 10.Arbit B, Marston N, Shah K, et al. Prognostic Usefulness of Proenkephalin in Stable Ambulatory Patients With Heart Failure. Am J Cardiol. 2016;117(8):1310–1314. doi: 10.1016/j.amjcard.2016.01.042 [DOI] [PubMed] [Google Scholar]
- 11.Hughes J, Kosterlitz HW, Smith TW. The distribution of methionine-enkephalin and leucine-enkephalin in the brain and peripheral tissues. Br J Pharmacol. 1977;61(4):639–647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Denning GM, Ackermann LW, Barna TJ, et al. Proenkephalin expression and encephalin release are widely observed in non-neuronal tissues. Peptides. 2008;29(1):83–92. doi: 10.1016/j.peptides.2007.11.004 [DOI] [PubMed] [Google Scholar]
- 13.Drolet G, Dumont EC, Gosselin I, Kinkead R, Laforest S, Trottier JF. Role of endogenous opioid system in the regulation of the stress response. Prog Neuropsychopharmacol Biol Psychiatry. 2001;25(4):729–741. doi: 10.1016/s0278-5846(01)00161-0 [DOI] [PubMed] [Google Scholar]
- 14.Przewłocki R, Hassan AH, Lason W, Epplen C, Herz A, Stein C. Gene expression and localization of opioid peptides in immune cells of inflamed tissue: functional role in antinociception. Neuroscience. 1992;48(2):491–500. doi: 10.1016/0306-4522(92)90509-z [DOI] [PubMed] [Google Scholar]
- 15.Salzet M, Tasiemski A. Involvement of pro-enkephalin-derived peptides in immunity. Dev Comp Immunol. 2001;25(3):177–185. doi: 10.1016/s0145-305x(00)00047-1 [DOI] [PubMed] [Google Scholar]
- 16.Moore RH, Dowling DA. Effects of intravenously administered Leu- or Met-enkephalin on arterial blood pressure. Regul Pept. 1980;1(2):77–87. doi: 10.1016/0167-0115(80)90012-9 [DOI] [PubMed] [Google Scholar]
- 17.Calero A, Villanueva H, Giri P, Tischler AS, Powers JF, Evinger M. Acute Hypoxia Induces Enkephalin Production and Release in an Adrenergic Cell Line Model of Neonatal Chromaffin Cell Responses to Hypoxic Stress. Am J Perinatol. 2018;35(11):1100–1106. doi: 10.1055/s-0038-1641169 [DOI] [PubMed] [Google Scholar]
- 18.van den Brink OWV, Delbridge LM, Rosenfeldt FL, et al. Endogenous cardiac opioids: Enkephalins in adaptation and protection of the heart. Heart Lung Circ. 2003;12(3):178–187. doi: 10.1046/j.1444-2892.2003.00240.x [DOI] [PubMed] [Google Scholar]
- 19.Webster KA, Discher DJ, Bishopric NH. Induction and nuclear accumulation of fos and jun proto-oncogenes in hypoxic cardiac myocytes. J Biol Chem. 1993;268(22):16852–16858. [PubMed] [Google Scholar]
- 20.Takasaki Y, Wolff RA, Chien GL, van Winkle DM. Met 5 -enkephalin protects isolated adult rabbit cardiomyocytes via δ-opioid receptors. Am J Physiol-Heart Circ Physiol. 1999;277(6):H2442–H2450. doi: 10.1152/ajpheart.1999.277.6.H2442 [DOI] [PubMed] [Google Scholar]
- 21.Allen CL, Bayraktutan U. Oxidative Stress and Its Role in the Pathogenesis of Ischaemic Stroke. Int J Stroke. 2009;4(6):461–470. doi: 10.1111/j.1747-4949.2009.00387.x [DOI] [PubMed] [Google Scholar]
- 22.Fearon IM, Faux SP. Oxidative stress and cardiovascular disease: Novel tools give (free) radical insight. J Mol Cell Cardiol. 2009;47(3):372–381. doi: 10.1016/j.yjmcc.2009.05.013 [DOI] [PubMed] [Google Scholar]
- 23.Thomas Heitzer, Titus Schlinzig, Karoline Krohn, Thomas Meinertz, Thomas Münzel. Endothelial Dysfunction, Oxidative Stress, and Risk of Cardiovascular Events in Patients With Coronary Artery Disease. Circulation. 2001;104(22):2673–2678. doi: 10.1161/hc4601.099485 [DOI] [PubMed] [Google Scholar]
- 24.Howard VJ, Cushman M, Pulley L, et al. The reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology. 2005;25(3):135–143. doi: 10.1159/000086678 [DOI] [PubMed] [Google Scholar]
- 25.Howard VJ, Kleindorfer DO, Judd SE, et al. Disparities in stroke incidence contributing to disparities in stroke mortality. Ann Neurol. 2011;69(4):619–627. doi: 10.1002/ana.22385 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Adams HP, Biller J. Classification of Subtypes of Ischemic Stroke: History of the Trial of Org 10 172 in Acute Stroke Treatment Classification. Stroke. 2015;46(5). doi: 10.1161/STROKEAHA.114.007773 [DOI] [PubMed] [Google Scholar]
- 27.Jenny NS, Callas PW, Judd SE, et al. Inflammatory cytokines and ischemic stroke risk: The REGARDS cohort. Neurology. 2019;92(20):e2375–e2384. doi: 10.1212/WNL.0000000000007416 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Arora P, Agarwal Z, Venkatraman A, et al. Galectin-3 and risk of ischaemic stroke: Reasons for Geographic and Racial Differences in Stroke cohort. Eur J Neurol. 2017;24(12):1464–1470. doi: 10.1111/ene.13440 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Gillett SR, Boyle RH, Zakai NA, McClure LA, Jenny NS, Cushman M. Validating laboratory results in a national observational cohort study without field centers: The Reasons for Geographic and Racial Differences in Stroke cohort. Clin Biochem. 2014;47(16):243–246. doi: 10.1016/j.clinbiochem.2014.08.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ernst A, Köhrle J, Bergmann A. Proenkephalin A 119–159, a stable proenkephalin A precursor fragment identified in human circulation. Peptides. 2006;27(7):1835–1840. doi: 10.1016/j.peptides.2006.03.008 [DOI] [PubMed] [Google Scholar]
- 31.Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol. 1999;52(12):1165–1172. doi: 10.1016/s0895-4356(99)00102-x [DOI] [PubMed] [Google Scholar]
- 32.Gutierrez OM, Judd SE, Muntner P, et al. Racial differences in albuminuria, kidney function, and risk of stroke. Neurology. 2012;79(16):1686–1692. doi: 10.1212/WNL.0b013e31826e9af8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Therneau TM, until 2009) TL (original S->R port and R maintainer, Elizabeth A, Cynthia C. Survival: Survival Analysis.; 2020. Accessed December 12, 2020. https://CRAN.R-project.org/package=survival
- 34.R: The R Project for Statistical Computing. Accessed December 12, 2020. https://www.r-project.org/
- 35.Rosenberger J, Petrovics G, Buzas B. Oxidative stress induces proorphanin FQ and proenkephalin gene expression in astrocytes through p38- and ERK-MAP kinases and NF-κB. J Neurochem. 2001;79(1):35–44. doi: 10.1046/j.1471-4159.2001.00520.x [DOI] [PubMed] [Google Scholar]
- 36.Islam MR, Yang L, Lee YS, Hruby VJ, Karamyan VT, Abbruscato TJ. Enkephalin-Fentanyl Multifunctional Opioids as Potential Neuroprotectants for Ischemic Stroke Treatment. Curr Pharm Des. 2016;22(42):6459–6468. doi: 10.2174/1381612822666160720170124 [DOI] [PubMed] [Google Scholar]
- 37.Wang S, Cao X, Duan Y, Zhang G. Delta Opioid Peptide [d-Ala2, d-Leu5] Enkephalin (DADLE) Exerts a Cytoprotective Effect in Astrocytes Exposed to Oxygen-Glucose Deprivation by Inducing Autophagy. Cell Transplant. 2019;28(6):775–782. doi: 10.1177/0963689719825619 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Stefano GB, Salzet M, Hughes TK, Bilfinger TV. δ2 opioid receptor subtype on human vascular endothelium uncouples morphine stimulated nitric oxide release. Int J Cardiol. 1998;64:S43–S51. doi: 10.1016/S0167-5273(98)00035-7 [DOI] [PubMed] [Google Scholar]
- 39.Arendt RM, Schmoeckel M, Wilbert-Lampen U, Plasse A, Heucke L, Werdan K. Bidirectional effects of endogenous opioid peptides on endothelin release rates in porcine aortic endothelial cell culture: mediation by delta opioid receptor and opioid receptor antagonist-insensitive mechanisms. J Pharmacol Exp Ther. 1995;272(1):1–7. [PubMed] [Google Scholar]
- 40.Balasubramanian S, Ramakrishnan S, Charboneau R, Wang J, Barke RA, Roy S. Morphine Sulfate Inhibits Hypoxia-induced Vascular Endothelial Growth Factor Expression in Endothelial Cells and Cardiac Myocytes. J Mol Cell Cardiol. 2001;33(12):2179–2187. doi: 10.1006/jmcc.2001.1480 [DOI] [PubMed] [Google Scholar]
- 41.Vidal EL, Patel NA, Wu G, Fiala M, Chang SL. Interleukin-1 induces the expression of μ opioid receptors in endothelial cells. Immunopharmacology. 1998;38(3):261–266. doi: 10.1016/S0162-3109(97)00085-4 [DOI] [PubMed] [Google Scholar]
- 42.Siti HN, Kamisah Y, Kamsiah J. The role of oxidative stress, antioxidants and vascular inflammation in cardiovascular disease (a review). Vascul Pharmacol. 2015;71:40–56. doi: 10.1016/j.vph.2015.03.005 [DOI] [PubMed] [Google Scholar]
- 43.Sesso HD, Buring JE, Christen WG, et al. Vitamins E and C in the Prevention of Cardiovascular Disease in Men: The Physicians’ Health Study II Randomized Trial. JAMA J Am Med Assoc. 2008;300(18):2123–2133. doi: 10.1001/jama.2008.600 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Cook NR, Albert CM, Gaziano JM, et al. A Randomized Factorial Trial of Vitamins C, E, and Beta-Carotene in the Secondary Prevention of Cardiovascular Events in Women. Arch Intern Med. 2007;167(15):1610–1618. doi: 10.1001/archinte.167.15.1610 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Doehner W, von Haehling S, Suhr J, et al. Elevated Plasma Levels of Neuropeptide Proenkephalin A Predict Mortality and Functional Outcome in Ischemic Stroke. J Am Coll Cardiol. 2012;60(4):346–354. doi: 10.1016/j.jacc.2012.04.024 [DOI] [PubMed] [Google Scholar]
- 46.Gruber P, Fluri F, Schweizer J, et al. Proenkephalin A Adds No Incremental Prognostic Value After Acute Ischemic Stroke. Clin Appl Thromb. 2020;26:1076029619895318. doi: 10.1177/1076029619895318 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Bai X, Zhang X, Li Y, Lu L, Li B, He X. Sex differences in peripheral mu-opioid receptor mediated analgesia in rat orofacial persistent pain model. PloS One. 2015;10(3):e0122924. doi: 10.1371/journal.pone.0122924 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Saloman JL, Niu KY, Ro JY. Activation of peripheral delta-opioid receptors leads to anti-hyperalgesic responses in the masseter muscle of male and female rats. Neuroscience. 2011;190:379–385. doi: 10.1016/j.neuroscience.2011.05.062 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Sarton E, Teppema L, Dahan A. Sex Differences in Morphine-induced Ventilatory Depression Reside within the Peripheral Chemoreflex Loop. Anesthesiology. 1999;90(5):1329–1338. doi: 10.1097/00000542-199905000-00017 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data used in these analyses are not freely publicly available due to legal and ethical restrictions. However, qualified investigators may request access to de-identified data from the University of Alabama at Birmingham (regardsadmin@uab.edu).