fpsyg-08-01361
August 7, 2017
Time: 15:7
# 4
Rozenblat et al.
5-HTTLPR and Disordered Eating
was coded as per the di-allelic model into s-present (s/s or s/l
genotype) or s-absent (l/l genotype) groups, as the s-allele is
believed to operate in a genetically dominant manner (
Lesch
et al., 1996
).
Potential Confounding Factors
Age, height, and weight were self-reported at age 15–16, with the
latter two figures used to calculate participant BMI. SES status
was measured according to maternal and paternal education and
occupation as reported by parents in the first survey in 1983.
Data Analysis
The main and interaction effects of
5-HTTLPR and the
two psychological stress exposures (depression and emotional
control), as well as the three social stress exposures (sexual abuse,
mild-to-moderate parental punishment, and severe parental
physical punishment), were assessed using separate linear
regression models. Outcome variables were Drive for Thinness
and Bulimia scores. G×E models were adjusted for sex and
BMI, as per
Keller (2014)
, by including all the covariate × gene
and covariate × environment interaction terms in the regression
models. Prior to analyses, missing data (23.5%) for the BMI
variable were imputed using multiple imputation in IBM SPSS
Version 21, with no systematic patterns of missingness observed.
A total of 10 tests were conducted with
p-values adjusted
accordingly (adjusted
p-value = 0.05/10, corrected p = 0.005),
to correct for multiple-testing, a frequent limitation of genetic
association studies (
Munafò and Flint, 2009
). Standardized effect
sizes are reported.
RESULTS
5-HTTLPR genotype distribution (l/l = 197, s/l = 341, and
s/s = 112) for the overall sample met the Hardy–Weinberg
Equilibrium,
χ
2
=
2.96,
df = 1, p
> 0.05. 5-HTTLPR genotype
TABLE 1 | Sociodemographic details of participants included in the present sample.
Full sample
Females
Males
N
N
%
N
%
N
%
Sex
650
Males
312
49.8
Females
338
50.2
SES Quartile
632
Highest
231
36.6
116
36.6
115
36.5
Medium-High
197
31.2
103
32.5
94
29.8
Medium-Low
129
20.4
59
18.6
70
22.2
Lowest
75
11.9
39
12.3
36
11.4
Parent marital status
645
Married/De facto
538
83.4
264
81.5
274
85.4
Separated/Divorced
73
11.3
42
13.0
31
9.7
Single/Widowed
22
3.4
10
3.1
12
3.7
Remarried
12
1.8
8
2.5
4
1.2
Father’s occupation
612
Professional
172
28.1
85
29.1
87
29.6
Managerial
126
20.6
61
20.9
65
22.1
Semi-skilled
200
32.7
105
36.0
95
32.3
Unemployed/Pensioner/House duties
114
18.6
41
14.0
47
16.0
Mother’s occupation
637
Professional
175
27.3
80
25.4
95
30.4
Managerial
54
8.5
31
9.8
23
7.3
Semi-skilled
193
30.3
104
33.0
89
28.4
Unemployed/Pensioner/House duties
215
33.8
100
31.7
106
33.9
Father’s education
594
Tertiary
176
29.6
86
28.7
90
30.6
Diploma/Apprenticeship
103
17.3
49
16.3
54
18.4
Year 11/12
167
28.1
87
29.0
80
27.2
Year 10 or less
148
24.9
78
26.0
70
23.8
Mother’s education
633
Tertiary
133
21.0
66
21.0
67
21.1
Diploma/Apprenticeship
102
16.1
55
17.5
47
14.8
Year 11/12
227
35.9
116
36.8
111
34.9
Year 10 or less
171
27.0
78
24.8
93
29.2
Frontiers in Psychology | www.frontiersin.org
4
August 2017 | Volume 8 | Article 1361
fpsyg-08-01361
August 7, 2017
Time: 15:7
# 5
Rozenblat et al.
5-HTTLPR and Disordered Eating
distribution was not in Hardy–Weinberg Equilibrium for the
subsample providing retrospective reports of sexual and physical
stress (l/l = 139, s/l = 255, and s/s = 73;
χ
2
=
6.11,
df = 1,
p = 0.014). Further descriptive statistics are presented in Table 2.
Across all regression models, female sex predicted Drive for
Thinness and Bulimia scores, while BMI predicted Drive for
Thinness (all
p
< 0.001). Tables pertaining to each regression
model discussed below are contained in the Supplementary
Materials.
Depression
There was a significant positive association between depressive
symptoms and Drive for Thinness (
β = 0.24, p < 0.001), as
well as Bulimia scores (
β = 0.41, p < 0.001); however, there
was no evidence of genetic moderation by
5-HTTLPR. There
was a significant interaction between depression and sex, with
depression associated with greater Drive for Thinness (
β = 0.46,
p = 0.001), and to a lesser extent, Bulimia (
β = 0.36, p = 0.018),
for females only. There were no other significant effects.
Emotional Control
Lower emotional control was significantly associated with greater
Drive for Thinness (
β = −0.22, p < 0.001) and Bulimia
(
β = −0.29, p < 0.001) scores; however, there was no evidence
of genetic moderation by
5-HTTLPR. There was a significant
interaction between sex and emotional control, with lower
emotional control associated with greater Drive for Thinness
(
β = −0.45, p = 0.001) and Bulimia (β = −0.67, p < 0.001)
for females to a greater extent than for males. However, amongst
those with the highest levels of emotional control, females
displayed lower levels of Bulimia than did males.
Sexual Abuse and Parental Physical
Punishment
Of the 467 participants (59.1% female) who provided data on
sexual abuse and parental physical punishment in the 14th
survey, 22 (4.7%) reported sexual abuse, 180 (38.5%) reported
mild to moderate parental physical punishment, and 27 (5.8%)
reported severed parental physical punishment. See Table 3 for
further descriptive statistics. Predictor and outcome variables for
this sub-sample did not differ from the overall sample (
t-tests all
p
> 0.05).
TABLE 2 | Descriptive statistics for mean values of continuous predictor and
outcome variables in the overall sample (N = 650), in females (N = 326), and in
males (N = 324).
Mean (SD)
Variable
Full sample
Females
Males
Age (years)
15
.72 (0.16)
15
.74 (0.14)
15
.72 (0.17)
EDI-2 Bulimia
1
.78 (0.65)
1
.94 (0.73)
1
.62 (0.53)
EDI-2 drive for thinness
2
.23 (1.15)
2
.81 (1.22)
1
.64 (0.70)
Emotional control
3
.74 (0.63)
3
.62 (0.64)
3
.85 (0.59)
Depression
0
.49 (0.34)
0
.59 (0.36)
0
.39 (0.28)
BMI
21
.27 (3.28)
21
.20 (3.10)
21
.34 (3.44)
There was a direct effect of severe parental physical
punishment in predicting Bulimia (
β = 0.14, p = 0.001), but
not Drive for Thinness scores. There were no direct effects of
sexual abuse or mild-to-moderate parental physical punishment
on either disordered eating outcome, although there was an
interaction between sexual abuse and sex (
β = 0.28, p = 0.039),
with males who reported experiencing sexual abuse tending to
display lower Drive for Thinness than those who did not report
sexual abuse. This pattern was not evident in females. However,
this result did not withstand
p-value adjustment for multiple
testing.
There was also statistically significant moderation of the
relationship between severe parental punishment and bulimia
scores by
5-HTTLPR, with greater punishment related to higher
Bulimia scores for those with the s-allele only (
β = 0.22,
p = 0.0048; see Figure 1). This finding remained after Bonferroni
correction for multiple testing was applied. However, this result
was based on a sample where genotypic frequencies were not in
the Hardy–Weinberg Equilibrium,
χ
2
=
6.11,
df = 1, p = 0.014.
No other G×E interactions were significant.
DISCUSSION
This study affirmed the central relationship between depression,
emotional control, and physical abuse and adolescent bulimic
behaviors
and
attitudes
regarding
thinness.
Conversely,
5-HTTLPR did not directly predict any pattern of disordered
eating, nor was there conclusive evidence that 5-
HTTLPR
moderated any risk factor for disordered eating. A statistically
significant interaction between
5-HTTLPR and retrospectively
reported parental physical punishment was observed; however,
genotypes in this subsample were not in Hardy–Weinberg
Equilibrium so cautious interpretation and independent
replication is needed.
Findings from this study support a key role for depression and
compromised emotional control in adolescent drive for thinness
and bulimic behaviors. The relationships differed by sex in most
cases. Female sex predicted greater overall drive for thinness and
bulimia, and sex differences were present in the relationships
between depression, emotional control, and sexual abuse with
disordered eating symptoms. Parental physical punishment was
the only variable that showed no sex differences in its relationship
TABLE 3 | Descriptive statistics for mean values of continuous predictor and
outcome variables in the subsample (N = 467), in females (N = 199), and in males
(N = 157).
Mean (SD)
Variable
Full sample
Females
Males
Age (years)
15
.72 (0.15)
15
.73 (0.15)
15
.71 (0.14)
EDI-2 Bulimia
1
.81 (0.67)
1
.92 (0.68)
1
.62 (0.51)
EDI-2 drive for thinness
2
.37 (1.20)
2
.79 (1.22)
1
.69 (0.74)
Emotional control
3
.74 (0.64)
3
.64 (0.65)
3
.88 (0.59)
Depression
0
.51 (0.34)
0
.58 (0.02)
0
.40 (0.28)
BMI
21
.19 (3.13)
21
.17 (2.97)
21
.21 (3.37)
Frontiers in Psychology | www.frontiersin.org
5
August 2017 | Volume 8 | Article 1361