A healthy diet in women is associated with less facial wrinkles in a large Dutch population-based cohort

Abstract

Background:
Little is known about the effects of different dietary patterns on facial wrinkling.

Objective:
We aimed to investigate the association between diet and facial wrinkles in a population-based cohort of 2,753 elderly participants of the Rotterdam Study.

Methods:
Wrinkles were digitally quantified as the area in facial photographs they occupied as a percentage of the total skin area. Diet was assessed using the Food Frequency Questionnaire. Adherence to the Dutch Healthy Diet Index (DHDI) was calculated. In addition, we used Principal Component Analysis (PCA) to extract relevant food patterns in men and women separately. All food patterns and the DHDI were analyzed for association with wrinkle severity using multivariable linear regression.

Results:
Better adherence to the Dutch guidelines was significantly associated with less wrinkles among women but not in men. In women, a red meat and snack dominant PCA pattern was associated with more facial wrinkles whereas a fruit dominant PCA pattern associated with fewer wrinkles.

Limitations:
Due to the cross-sectional design of our study, no causation can be proved. Health-conscious behavior could have influenced our results.

Conclusion:
Dietary habits associate with facial wrinkling in women. Global disease prevention strategies might benefit from emphasizing that a healthy diet is also linked to less facial wrinkling.

Full text

A healthy diet in women is associated
with less facial wrinkles in a large Dutch
population-based cohort
Selma Meki
c, MD,
a
LeonieC.Jacobs,MD,PhD,
a
MerelA.Hamer,MD,
a
M. Arfan Ikram, MD, PhD,
b
Josje D. Schoufour, PhD,
b
David A. Gunn, PhD,
c
Jessica C. Kiefte-de Jong, PhD,
b,d
and Tamar Nijsten, MD, PhD
a
Rotterdam and The Hague, The Netherlands, and Bedfordshire, United Kingdom
Background: Little is known about the effects of different dietary patterns on facial wrinkling.
Objective: We aimed to investigate the association between diet and facial wrinkles in a population-based
cohort of 2753 elderly participants of the Rotterdam study.
Methods: Wrinkles were measured in facial photographs by digitally quantifying the area wrinkles occupied
as a percentage of total skin area. Diet was assessed by the Food Frequency Questionnaire. Adherence to the
Dutch Healthy Diet Index (DHDI) was calculated. In addition, we used principal component analysis (PCA)
to extract relevant food patterns in men and women separately. All food patterns and the DHDI were
analyzed for an association with wrinkle severity using multivariable linear regression.
Results: Better adherence to the Dutch guidelines was significantly associated with less wrinkles among
women but not in men. In women, a red meat and snackedominant PCA pattern was associated with more
facial wrinkles, whereas a fruit-dominant PCA pattern was associated with fewer wrinkles.
Limitations: Due to the cross-sectional design of our study, causation could not be proven. Other
health-conscious behaviors of study participants could have influenced the results.
Conclusion: Dietary habits are associated with facial wrinkling in women. Global disease prevention
strategies might benefit from emphasizing that a healthy diet is also linked to less facial wrinkling. ( J Am
Acad Dermatol 2019;80:1358-63.)
Key words: diet; Dutch Healthy Diet Index; facial wrinkling; healthy lifestyle; nutrition; principal
component analysis; Rotterdam study; skin aging.
Maintaining a healthy body and youthful
appearance is increasingly becoming
popular because the longevity and wealth
of the global population is still rising. The rise of
functional foods claiming various skin benefits
suggests that certain nutrients could help to prevent
skin aging and enhance cosmesis.
1
While several small studies have investigated the
effects of dietary supplements on skin aging,
2-4
large
nutritional studies on this topic are lacking. To our
From the Departments of Dermatology
a
and Epidemiology,
b
Erasmus MC University Medical Center, Rotterdam; Unilever
Research and Development, Colworth Science Park, Bedford-
shire
c
; and Leiden University College, The Hague.
d
Funding sources: Supported by restricted research grant from
Unilever (to Dr Nijsten and Drs Meki
c). The funders had no role
in the study design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Conflicts of interest: Although no products were tested, it is
possible that this manuscript could promote products or foods
that reduce the appearance of wrinkles, which could lead to
financial gain for Unilever of whom Dr Gunn is an employee.
Accepted for publication March 15, 2018.
Correspondence to: Tamar Nijsten, MD, PhD, Department of
Dermatology, Erasmus MC University Medical Center, PO Box
2040, 3000 CA, Rotterdam, The Netherlands. E-mail: t.nijsten@
erasmusmc.nl.
Reprint requests: Selma Meki
c, MD, Department of Dermatology,
Erasmus MC University Medical Center, PO Box 2040, 3000 CA,
Rotterdam, The Netherlands. E-mail: s.mekic@erasmusmc.nl.
Published online March 27, 2018.
0190-9622/$36.00
Ó2018 by the American Academy of Dermatology, Inc. Published
by Elsevier Inc. All rights reserved.
https://doi.org/10.1016/j.jaad.2018.03.033
1358

knowledge, only 3 previous studies have investi-
gated features of skin aging in association with diet.
5-7
In these studies, intake of vegetables, foods high in
carotenoids and vitamin C, olive oil, linoleic acid,
and fish were associated with less photoaging and
intake of saturated fats and sugar with more
wrinkling.
On the basis of these
observations, a healthy diet
appears associated with less
skin aging. However, in
these previous studies, re-
searchers investigated sepa-
rate nutrients or food groups
that were prone to false-
positive associations because
of co-linearity with the
causative nutrient and the
interaction between single
nutrients. Also, the effect
sizes of single nutrients are
often small, making it diffi-
cult to discover associations. Studying complete
dietary patterns in epidemiologic nutritional
research, therefore, can be preferred over studying
single nutrients.
8
Dietary pattern analysis can be
conducted a priori, in which the healthiest pattern is
predefined using existent guidelines, eg, the Dutch
Healthy Diet Index (DHDI).
9
However, in case of
little prior knowledge, an a posteriori approach, in
which formed patterns were data driven, could be
more appropriate, eg, using a principal component
analysis (PCA).
8
In our study, we investigated the association
between digitally quantified facial wrinkling, dietary
patterns, and healthy lifestyle parameters in a large
population-based cohort of 2753 elderly participants
of the Rotterdam study using both an a priori and an
a posteriori approach.
METHODS
Study population
Participants were selected from the Rotterdam study,
a prospective population-based cohort study in
Rotterdam, the Netherlands. The Rotterdam study was
approved by the institutional review board (Medical
Ethics Committee) of the Erasmus Medical Center and
by the review board of The Netherlands Ministry of
Health, Welfare, and Sports. Objectives and details of
the study design have been described elsewhere.
10
During 2010-2014, standardized high-resolution
digital facial photographs were taken of 4649
participants by trained physicians. From these pic-
tures, we obtained wrinkle data for 3831 participants,
and nutrition data was available for 2813 of these
participants. We excluded 60 of the 2813 participants
because of unrealistic caloric intakes (\500 and
[5000 kcal/day). The remaining 2753 participants
were included in our analysis.
Wrinkles
Using full-face photographs, we digitally quanti-
fied the area detected as
wrinkles as a percentage of
the total facial skin area using
a semi-automated script
in MATLAB (MathWorks,
Natick, MA). Our wrinkle
data has been validated and
utilized in other analyses.
11-13
Dietary intake and food
pattern analysis
Dietary intake was assessed
using a validated semi-
quantitative Food Frequency
Questionnaire.
9
We defined
the a priori healthiness of the diet in our population
using the DHDI.
14,15
For the a posteriori approach, we
used a PCA (Supplementary Appendix; available at
http://www.jaad.org).
Statistical analysis
For all analyses, we used a basic (age adjusted) or
a multivariable linear regression model stratified by
sex because men and women have different risk
factors for wrinkling.
11
First, we tested the effect of
known and possible new risk factors (physical
activity, daily energy intake) on facial wrinkling.
Second, all PCA patterns and DHDI were used
to test associations between wrinkles and diet
(Supplementary Appendix).
Additional analysis
Ultraviolet (UV) exposure data was missing for
45% of the study participants. In a complete case
analysis in women (N = 849), we adjusted our main
analysis for UV exposure variables (Supplementary
Appendix). Association tests with physical
activity were additionally adjusted for UV to
reduce residual confounding. Also, we tested single
food groups separately in women to understand
which food groups drive the association of a
food pattern (Supplemental Table I; available at
http://www.jaad.org).
RESULTS
The study population consisted of 2753 middle-
aged and elderly Dutch men (41%) and women
(59%), with a median age of 67.3 (interquartile range
CAPSULE SUMMARY
dLittle is known about the impact of
nutrition on youthful appearance.
dAdherence to the recommended healthy
diet was associated with less wrinkling
and the unhealthy diet with more facial
wrinkling in women.
dDietary recommendations for skin aging
preventing strategies could in addition
help to improve overall health.
JAMACAD DERMA TOL
VOLUME 80, NUMBER 5
Meki
cetal 1359

[IQR] 62.6-72.3) years. Known risk factors, such as
age, sex, body mass index, and smoking status,
showed a significant association with wrinkle area
in both the basic and the multivariable model. Of the
newly investigated risk factors, daily energy intake
did not have an effect on wrinkles, even when not
adjusting for body mass index. Strikingly, more
physical activity resulted in more wrinkles in the
multivariable model in men and in women (Table I).
Adherence to the predefined healthy diet, shown
by higher DHDI scores, resulted in significantly less
facial wrinkling in women (4.19%, 95% confidence
interval [CI] 7.30 to 1.08; Table II) but not men. In
women and in men, we extracted 4 and 3 food
patterns, respectively, using PCA of the 34 food
groups. The first 3 PCA patterns in women and men
were comparable. The first pattern consisted of high
consumption of mainly healthy food groups
(including vegetables, fish and poultry, nuts and
seeds, and mineral water) and wine. The second
pattern was an unhealthy pattern consisting of
consumption of mainly meat, grains, snacks, soft
drinks, coffee, and other alcoholic drinks. The third
pattern was an intermediate mix of healthy and
unhealthy foods that resembled a typical Dutch diet,
which included a high intake of cheese, potatoes,
grains, and fats (Table III). The fourth PCA pattern,
which was seen in women, was a diet high in
fruit, supplemented with yogurt, milk, and some
vegetables (Table III).
In men, no a posteriorly defined food pattern was
associated with increased or decreased wrinkling,
but in women, the unhealthy pattern was
significantly associated with more wrinkling
(3.32%, 95% CI 0.06 to 6.68) and the fruit pattern
was significantly associated with less facial wrinkling
(3.20%; 95% CI 6.25 to 0.06) (Table II). We also
calculated the same fruit PCA pattern in men, but
there was no significant protective effect on wrinkles
for this food pattern (0.41%, 95% CI 3.67 to 2.96).
UV exposure and physical activity did not
significantly alter effect size of food patterns in our
sensitivity analysis (data not shown). The single food
group analysis detected single food groups
associated with facial wrinkling (Supplemental
Table I).
DISCUSSION
We found a healthy diet to be associated with less
facial wrinkling in women, shown by both the
predefined DHDI and the healthy fruit pattern in
women. In addition, the unhealthy food pattern was
associated with more facial wrinkling in the same
group, providing more evidence of the link between
a healthy diet and wrinkling. These observations are
in-line with previous studies showing that high
intake of animal source products, fats, and
carbohydrates increased skin aging
5,6
and vitamin
C and carotenoids decreased winkles.
7
Both a healthy diet preventing wrinkles and an
unhealthy diet aggravating wrinkles were found in
women but not in men. Men and women are known
to show distinct wrinkling patterns and different
dietary habits, which could help explain the sex
differences in the wrinkle associations.
11,16
Although, no wrinkle-protecting effect was found
when applying the fruit-based food pattern to the
male subgroup, this difference might be explained
by men consuming less fruits than women, making
an association harder to detect.
The posteriorly defined healthy food PCA pattern
was not associated with less facial wrinkling in
women. The single food group analysis showed
that of nutrients in the healthy PCA pattern, yellow
vegetables and soy were significantly associated with
less wrinkling and wine was significantly associated
with more wrinkling. Thus, this common food
pattern includes food groups that associate with
both less and greater wrinkling, and is therefore not
associated with wrinkling overall.
Examining patterns of nutrient intake can be
valuable in the interpretation of nutrition associa-
tions. For example, although processed meat does
not associate with wrinkling in the single nutrient
analysis, it does associate via the unhealthy PCA
pattern in women, which suggests that in concert
with other (unhealthy) nutrients, processed meat
could be promoting skin wrinkling.
The biologic mechanism responsible for the as-
sociation between skin wrinkling and unhealthy diet
could be increased oxidative stress load,
17
an upre-
gulated inflammatory state,
18
or the effect of
advanced glycation endproducts, which can disrupt
cell metabolism and weaken antioxidant defense.
19
In contrast, vitamins and flavonoids in a healthy
diet provide protection from photoaging and
stimulate collagen production and DNA repair
mechanisms.
20,21
We found more physical activity to be associated
with more facial wrinkling in both sexes. As many
sports are practiced outside, UV exposure could play
Abbreviations used:
CI: confidence interval
DHDI: Dutch Healthy Diet Index
IQR: interquartile range
PCA: principal component analysis
UV: ultraviolet
JAMACAD DERMATOL
MAY 2019
1360 Meki
cetal

Table I. Population characteristics of women (N = 1613) and men (N = 1140) included in study
Characteristic Value
Wrinkle % change,
univariable analysis 95% CI*
Wrinkle % change,
multivariable analysis 95% CI
y
Women
Wrinkle % change, median (IQR) 3.7 (2.3-5.8)
Age, y, median (IQR) 67.1 (62.5-72.0) 4.524 4.102 to 4.948 4.726 4.294 to 5.159
Daily energy intake, Kcal, mean (SD)
z
2027 (636) 0.0004 0.004 to 0.005 0.001 0.005 to 0.004
Physical activity (MET), hr/wk, median (IQR)
x
46.6 (18.8-87.4) 0.093 0.029 to 0.157 0.082 0.019 to 0.144
BMI, kg/m
2
, mean (SD) 27.4 (4.8) 2.138 2.726 to 1.546 2.057 2.652 to 1.458
Smoking, %
Never 634 (39) Referent eReferent e
Former 739 (46) 1.392 6.992 to 4.545 10.413 3.798 to 17.450
Current 237 (15) 32.692 22.300 to 43.967 37.473 25.978 to 50.016
Education,
jj
%
Low 140 (9) Referent eReferent e
Medium 1087 (67) 2.235 4.018 to 8.894 3.142 12.156 to 6.798
High 370 (23) 5.111 11.610 to 1.865 9.263 18.754 to 1.337
Men
Wrinkle % change, median (IQR) 4.6 (3.1-6.5)
Age, y, median (IQR) 67.7 (62.7-72.7) 2.395 1.980 to 2.811 2.555 2.113 to 2.999
Daily energy intake, Kcal, mean (SD)
z
2312 (706) 0.005 0.0003 to 0.009 0.004 0.001 to 0.008
Physical activity (MET), hr/wk, median (IQR)
x
40.7 (17.9-73.3) 0.111 0.035 to 0.186 0.108 0.033 to 0.183
BMI, kg/m
2
, mean (SD) 27.4 (3.5) 1.898 2.757 to 1.031 1.816 2.680 to 0.946
Smoking, %
Never 242 (21.2) Referent eReferent e
Former 679 (59.6) 6.149 11.978 to 0.066 0.524 6.935 to 8.620
Current 218 (19.1) 13.610 5.102 to 22.807 15.279 4.898 to 26.686
Education,
jj
%
Low 67 (5.9) Referent eReferent e
Medium 630 (55.3) 0.858 5.131 to 7.225 9.157 3.092 to 23.017
High 428 (37.5) 1.194 5.008 to 7.800 8.463 4.140 to 22.721
Significant results (P#.05) are bold.
BMI, Body mass index; CI, confidence interval; IQR, interquartile range; MET, metabolic equivalent of task; SD, standard deviation.
*Univariable analysis was adjusted for technical variation and age.
y
Multivariable analysis was adjusted for technical variation, age, daily energy intake, physical activity, BMI, smoking status, and education.
z
Daily energy intake in kilocalories. 1 Kcal = 4184 joules.
x
MET hours per week is a physiologic measure expressing the energy cost of physical activity.
jj
Categories of education were low (primary education), medium (lower vocational education, lower secondary education, intermediate vocational education), and high (general secondary
education, higher vocational education, university).
JAMACAD DERMA TOL
VOLUME 80, NUMBER 5
Meki
cetal 1361

Table II. Association of Dutch Healthy Diet Index and dietary patterns with facial wrinkles
y
Dietary pattern
Women, N = 1613 Men, N = 1140
Wrinkle % change*95% CI PWrinkle % change*95% CI P
A priori
Dutch Healthy Diet Index 4.48 7.58 to 1.36 .005 0.61 2.79 to 4.03 .724
A posteriori
Healthy 0.56 3.55 to 2.54 .723 0.76 2.62 to 4.26 .664
Unhealthy 3.32 0.06 to 6.68 .046 2.72 0.58 to 6.12 .107
Intermediate 1.84 5.39 to 1.83 .322 0.67 4.79 to 3.63 .755
Fruit 3.20 6.25 to 0.06 .046 ---
Pvalues \.05 were considered significant and are presented in bold.
CI, Confidence interval.
*Increase or decrease in percentage of wrinkle area percentage (D%) per 10 points increase on the DHDI, when committing to a dietary pattern.
y
Adjusted for technical variation, age, physical activity, body mass index, daily energy intake, smoking, and education level.
Table III. Dietary patterns (eigenvalue $1.5) with factor loadings of the contributing food groups in women
and men
Category
Women, N = 1613 Men, N = 1140
Healthy Unhealthy Intermediate Fruit Healthy Unhealthy Intermediate
Citrus fruits - - - 0.836 0.448 0.418 -
Other fruits - - - 0.826 0.515 0.455 -
Yellow vegetables 0.754 - - 0.204 0.718 --
Green leafy vegetables 0.739 ---0.691 --
Other vegetables 0.688 ---0.653 --
Pulses 0.227 - - - - - -
Milk - - - 0.251 - - 0.323
Yoghurt - - - 0.208 0.267 0.230 -
Cheese - 0.370 - - - - 0.343
Soy 0.297 0.277 - - - - -
Nuts and seeds - - - - 0.317 - -
Eggs - 0.350 - - - 0.228 -
Poultry 0.262 0.252 - - 0.271 0.251 -
Unprocessed meat - 0.546 ---0.471 -
Processed meat - 0.575 ---0.505 0.306
Lean fish 0.394 0.227 - - 0.339 0.289 -
Fatty Fish 0.469 - - - 0.291 0.248 -
Shellfish 0.272 0.275 - - - 0.378 -
Whole grains - - 0.365 - - - 0.374
Refined grains - 0.338 0.246 - - 0.425 0.210
Potatoes - 0.204 0.417 - - 0.097 0.394
Soups and sauces - - 0.343 - - 0.181 0.397
Savory snacks - 0.446 ---0.454 -
Sweets - 0.229 0.441 0.285 - - 0.521
Soft drinks - 0.286 - - - 0.288 0.237
Wine 0.246 0.233 - - 0.246 - 0.299
Other alcoholic drinks - - - - - 0.404 -
Mineral water 0.303 - - - 0.289 0.100 -
Herb tea 0.299 0.314 - - 0.214 - -
Black tea - - 0.209 - - 0.259 0.235
Coffee - 0.338 - - - 0.328 -
Olive oil - - 0.494 - - - 0.266
Healthy fats - - 0.619 -- - 0.532
Unhealthy fats - - 0.589 -- - 0.507
Eigenvalues 3.051 2.302 1.753 1.588 3.140 2.326 1.776
Explained variance, % 8.973 6.770 5.155 4.671 9.236 6.842 5.223
The food groups presented, with factor loadings $0.2 or #0.2, were considered to have an important association with at least 1 of the
dietary patterns. Factor loadings [0.4 (bold) represent the highest and most explanatory for the specific pattern.
-, Weak associations.
JAMACAD DERMATOL
MAY 2019
1362 Meki
cetal

a role through residual confounding. However, the
effect was independent of UV exposure in our
sensitivity analysis.
The main strength of our study is that we used 2
validated methods to capture dietary patterns
associated with facial wrinkles in a large
population-based cohort. Also, wrinkles were
digitally quantified in a standardized and validated
way, reducing interobserver bias and measurement
error.
However, nutrition intake is difficult to capture
accurately and our Food Frequency Questionnaire
data correlates less with intake of vegetables and
better with snacks in a validation study.
9
We tried to
reduce confounding by adjusting for possible and
known confounders in our analyses. Nonetheless,
there are other possible residual confounders, which
were not available in our data set, such as stress and
hours of sleep per night. Another possible
confounder is health-conscious behavior, as it is
possible that people who eat healthy also tend to use
sunscreen more often. Although our sensitivity
analysis excluded confounding by UV protection
behaviors, 45% of the data was missing, giving some
uncertainty to the accuracy of this analysis. Finally,
due to the cross-sectional design of our study, we
cannot exclude reverse causality.
In conclusion, our findings imply that type of diet
influences the severity of facial wrinkles in women,
where an unhealthy diet significantly increases
wrinkling and a healthy diet decreases facial
wrinkling. This creates opportunities to stimulate
adherence to a healthy dietary pattern in women
who want to maintain a youthful appearance, which
simultaneously could improve overall health and
decrease mortality risk.
22
REFERENCES
1. Cho S. The role of functional foods in cutaneous anti-aging.
J Lifestyle Med. 2014;4:8-16.
2. Yoon H-S, Kim JR, Park GY, et al. Cocoa flavanol
supplementation influences skin conditions of photo-aged
women: a 24-week double-blind, randomized, controlled trial.
The J Nutr. 2016;146:46-50.
3. Schwartz S, Frank E, Gierhart D, Simpson P, Frumento R.
Zeaxanthin-based dietary supplement and topical serum
improve hydration and reduce wrinkle count in female
subjects. J Cosmet Dermatol. 2016;15:e13-e20.
4.

Zmitek K, Poga
cnik T, Mervic L,

Zmitek J, Pravst I. The effect of
dietary intake of coenzyme Q10 on skin parameters and
condition: results of a randomised, placebo-controlled,
double-blind study. BioFactors. 2017;43:132-140.
5. Purba MB, Kouris-Blazos A, Wattanapenpaiboon N, et al. Skin
wrinkling: can food make a difference? J Am Coll Nutr. 2001;20:
71-80.
6. Cosgrove MC, Franco OH, Granger SP, Murray PG, Mayes AE.
Dietary nutrient intakes and skin-aging appearance among
middle-aged American women. Am J Clin Nutr. 2007;86:
1225-1231.
7. Nagata C, Nakamura K, Wada K, et al. Association of dietary fat,
vegetables and antioxidant micronutrients with skin ageing in
Japanese women. Br J Nutr. 2010;103:1493-1498.
8. Hu FB. Dietary pattern analysis: a new direction in nutritional
epidemiology. Curr Opin Lipidol. 2002;13:3-9.
9. Goldbohm RA, van den Brandt PA, Brants HA, et al. Validation
of a dietary questionnaire used in a large-scale prospective
cohort study on diet and cancer. Eur J Clin Nutr. 1994;48:
253-265.
10. Ikram MA, Brusselle GGO, Murad SD, et al. The Rotterdam
study: 2018 update on objectives, design and main results. Eur
J Epidemiol. 2017;32:807-850.
11. Hamer MA, Pardo LM, Jacobs LC, et al. Lifestyle and
physiological factors associated with facial wrinkling in men
and women. J Invest Dermatol. 2017;137:1692-1699.
12. Hamer MA, Jacobs LC, Lall JS, et al. Validation of image
analysis techniques to measure skin aging features from facial
photographs. Skin Res Technology. 2015;21:392-402.
13. Liu F, Hamer Merel A, Deelen J, et al. The MC1R gene and
youthful looks. Curr Biol. 2016;26:1213-1220.
14. van Lee L, Feskens EJ, Meijboom S, et al. Evaluation of a
screener to assess diet quality in the Netherlands. Br J Nutr.
2016;115:517-526.
15. van Lee L, Geelen A, van Huysduynen EJ, de Vries JH, van’t
Veer P, Feskens EJ. The Dutch Healthy Diet index (DHD-Index):
an instrument to measure adherence to the Dutch guidelines
for a healthy diet. Nutr J. 2012;11:49.
16. Wardle J, Haase AM, Steptoe A, Nillapun M, Jonwutiwes K,
Bellisle F. Gender differences in food choice: the contribution
of health beliefs and dieting. Ann Behav Med. 2004;27:107-116.
17. Rinnerthaler M, Bischof J, Streubel MK, Trost A, Richter K.
Oxidative stress in aging human skin. Biomolecules. 2015;5:
545-589.
18. Baylis D, Bartlett DB, Patel HP, Roberts HC. Understanding how
we age: insights into inflammaging. Longev Healthspan. 2013;
2:8.
19. Clarke RE, Dordevic AL, Tan SM, Ryan L, Coughlan MT. Dietary
advanced glycation end products and risk factors for chronic
disease: a systematic review of randomised controlled trials.
Nutrients. 2016;8:125.
20. Schagen SK, Zampeli VA, Makrantonaki E, Zouboulis CC.
Discovering the link between nutrition and skin aging.
Dermatoendocrinol. 2012;4:298-307.
21. Rahman K. Studies on free radicals, antioxidants, and
co-factors. Clin Interv Aging. 2007;2:219-236.
22. Gunn DA, Larsen LA, Lall JS, Rexbye H, Christensen K. Mortality
is written on the face. J Gerontol A Biol Sci Med Sci. 2016;71:
72-77.
JAMACAD DERMA TOL
VOLUME 80, NUMBER 5
Meki
cetal 1363

MATERIALS AND METHODS
Covariate selection
Sex and age were collected from the database.
Education level, smoking habits, ultraviolet (UV)
exposure and physical activity were retrieved from
the interviews. Body mass index (BMI) was
calculated from weight and height measured at the
research center. Total energy intake per day in
kilocalories was calculated using the Dutch Food
Composition Table (NEVO) of 2006.
S1
Food Frequency Questionnaire (FFQ)
The FFQ gives information on the consumption
frequency and the average consumed amounts of
389 food items. The FFQ score was validated
against dietary records over a 3-day period
(4-5 months apart) in another Dutch population
aged 55-69 years.
S2
A priori food pattern
The DHDI is a composed measure of healthy
nutritional behavior taking the Dutch governmental
guidelines of a healthy diet into account, where a
higher score correlates with the highest diet
quality.
S3,S4
Physical activity and fish, fruit,
vegetable, and fiber consumption are adequacy
components, and saturated fatty acids, trans-fatty
acids, number of consumption occasions of acidic
drinks and foods, sodium, and alcohol are
moderation components.
S4
In our study, we
calculated the DHDI from the nutritional data out
of the FFQ, leaving the physical activity component
out, resulting in a healthiness grade 0-90.
A posteriori food patterns
The 389 food items were first subdivided into 34
food groups by a nutritionist (Supplemental Table I)
on the basis of their nutritional characteristics and
hypothesized association with skin aging.
S1
The
principal component analysis (PCA) with varimax
rotation extracted food patterns from the 34 food
groups, explaining the maximum variation of food
intake in women and men separately, since men and
women tend to eat differently.
S5
Food patterns were
considered relevant when showing an eigenvalue
$1.5.
Statistical analysis
Associations between dietary intake and wrinkle
area percentage were assessed using linear regression.
Wrinkle area percentage was natural logarithme
transformed to normalize the distribution. For a
more intuitive interpretation of the betas, we used
the formula (exp
b
-1) 3100%, which results in a
wrinkle percentage change. This is the percentage
increase or decrease in wrinkle area per unit increase
of the tested variable. All analyses were adjusted for
technical variation, explained by 2 variables, which
accounted for variations in resolution and flash light,
asdescribedindetailpreviously.
S6
We tested the
association of the wrinkle percentage in both a basic
model (adjusted for technical variation and age) and a
multivariable model including all covariates (age, sex,
BMI, daily energy intake, physical activity, smoking
habit, education level). We tested for effect
modification by BMI, which did not alter our results.
All covariates had \8% missing data, which we
replaced with multiple imputation. In our main
analysis, we tested the association of DHDI (per 10
points increase in DHDI) and the relevant nutritional
patterns from the PCA with wrinkle area adjusted for
all covariates. The extra relevant food pattern in
women was also tested in men. All analyses were
conducted using IBM SPSS Statistics for Windows
version 21.0.
UV variables
UV exposure variables included tanning bed use,
hibernating in a sunny country, sunburn tendency,
outdoor work, and UV protection behavior.
SUPPLEMENTAL REFERENCES
S1. Voedingscentrum. Nederlands Voedingsstoffenbestand
2006/Stichting Nederlands Voedingsstoffenbestand. The
Hague, the Netherlands: RIVM; 2006.
S2. Goldbohm RA, van den Brandt PA, Brants HA, et al. Validation
of a dietary questionnaire used in a large-scale prospective
cohort study on diet and cancer. Eur J Clin Nutr. 1994;48:
253-265.
S3. van Lee L, Feskens EJ, Meijboom S, et al. Evaluation of a
screener to assess diet quality in the Netherlands. Br J Nutr.
2016;115:517-526.
S4. van Lee L, Geelen A, van Huysduynen EJ, de Vries JH, van’t
Veer P, Feskens EJ. The Dutch Healthy Diet index (DHD-
index): an instrument to measure adherence to the Dutch
guidelines for a healthy diet. Nutr J. 2012;11:49.
S5. Wardle J, Haase AM, Steptoe A, Nillapun M, Jonwutiwes K,
Bellisle F. Gender differences in food choice: the contribution
of health beliefs and dieting. Ann Behav Med. 2004;27:
107-116.
S6. Hamer MA, Pardo LM, Jacobs LC, et al. Lifestyle and
physiological factors associated with facial wrinkling in men
and women. J Invest Dermatol. 2017;137:1692-1699.
JAMACAD DERMATOL
MAY 2019
1363.e1 Meki
cetal

Supplemental Table I. Single food group analysis
in women
Food group B 95% LB 95% UB P
Citrus fruits 0.022 0.051 0.007 .14
All other fruits 0.009 0.021 0.003 .14
Total fruit*0.007 0.017 0.002 .11
Green leafy vegetables 0.028 0.081 0.024 .29
Yellow vegetables
y
0.061 0.112 0.009 .02
All other vegetables 0.005 0.036 0.026 .76
Total vegetables*0.011 0.028 0.006 .20
Pulses 0.048 0.046 0.142 .32
Milk 0.008 0.007 0.022 .29
Yogurt 0.015 0.007 0.037 .19
Cheese
z
0.165 0.038 0.293 .01
Soy products
y
0.064 0.119 0.009 .02
Refined grains 0.060 0.129 0.010 .09
Whole grains 0.002 0.038 0.042 .91
Soft drinks 0.002 0.026 0.030 .87
Eggs
z
0.187 0.013 0.361 .04
Unprocessed meat 0.014 0.091 0.118 .80
Processed meat 0.114 0.295 0.067 .22
Poultry
y
0.196 0.382 0.009 .04
Fatty fish 0.053 0.110 0.216 .53
Lean fish 0.043 0.143 0.230 .65
Shellfish 0.548 0.137 1.238 .12
Total fish*0.045 0.056 0.147 .38
Savory snacks 0.115 0.023 0.254 .10
Sweets 0.048 0.121 0.024 .19
Nuts and seeds 0.053 0.156 0.262 .62
Coffee
z
0.016 0.003 0.029 .02
Black tea 0.002 0.016 0.012 .79
Herbal tea 0.005 0.026 0.017 .68
Mineral water 0.009 0.002 0.019 .11
Alcoholic drinks
other than wine
0.015 0.018 0.048 .37
Wine
z
0.041 0.011 0.070 .01
Soups and sauces
y
0.048 0.095 0.002 .04
Potatoes 0.002 0.048 0.053 .93
Olive oil 0.060 0.467 0.589 .82
Healthy fats 0.015 0.184 0.214 .88
Unhealthy fats
z
0.237 0.043 0.432 .02
Total fats*0.103 0.019 0.226 .10
Wrinkle percent change per 100 g intake of food group was
calculated (N = 1613). Multivariable linear regression was adjusted
for technical variation, age, body mass index, energy intake,
physical activity, smoking, and education.
B, Beta; LB, lower bound; UB, upper bound.
*Because of their different nutritional characteristics, some of the
food groups are presented both as a subgroup defined by a
nutritionist (eg, citrus fruits that are high in vitamin C) and as a
total together with all other fruits. Total fruits = citrus fruits 1all
other fruits. Total vegetables = green leafy vegetables 1yellow
vegetables 1all other vegetables. Total fish = fatty fish 1lean
fish 1shellfish. Total fats = olive oil 1healthy fats 1unhealthy
fats.
y
Significant association (P\.05) with less wrinkling.
z
Significant association (P\.05) with more wrinkling.
JAMACAD DERMA TOL
VOLUME 80, NUMBER 5
Meki
cetal 1363.e2