Changes in fruit and vegetable consumption in relation to changes in sleep characteristics over a 3-month period among young adults

Abstract

Objectives:

To evaluate whether increases in fruit and vegetable (FV) consumption were associated with concomitant changes in insomnia symptoms, sleep duration, and quality.

Design:

Secondary longitudinal analysis of a randomized trial, baseline to 3 months

Setting:

Integrated health care systems in Detroit, Michigan and Danville, Pennsylvania

Participants:

1165 young adults who were low consumers of FV (<3 servings/day) at baseline

Intervention:

Online 3-arm program designed to increase FV consumption

Measurements:

We categorized FV changes into 4 categories: no change or decrease, 1 serving increase, 2 serving increase, and 3 or more serving increase. We then compared the changes in chronic insomnia classification (yes or no), sleep duration, quality, and time to fall asleep (all self-reported) across the FV change categories. Analyses were both overall and stratified by gender, adjusting for potential confounders (depression, physical activity, education, children, and study site).

Results:

Average age ± SD was 26 ± 2.8 years (71% women). At 3-month follow-up, participants on average increased FV intake by 1.2 ± 1.4 servings. Women who increased FV intake by 3+ servings showed improvements in insomnia symptoms (2-fold higher odds of improvement; 95% CI 1.1 to 3.6), sleep quality (0.2-point higher sleep quality score; 95% CI −0.01, 0.3), and time to fall asleep (4.2 minutes; 95% CI −8, 0) compared to women who did not change or decreased their FV intake. Associations were not as apparent among men.

Conclusion:

Young women with low consumption of FV may experience improvements in insomnia-related sleep difficulties by increasing their consumption of FV.

INTRODUCTION

Young adults in the US have a high prevalence of poor sleep quality and short sleep duration. In one study of US adults, 67% of those aged 19–26 felt they did not obtain sufficient sleep to function properly¹. A nationally representative survey of US adults showed that about half of those aged 19–32 years experienced some self-reported problems with sleep, difficulty falling asleep, non-refreshing sleep or poor perceived sleep quality². In a recent study of midwestern young adults, over one-third of participants met the criteria for chronic insomnia³. These sleep statistics are a public health concern since sufficient and high-quality sleep is highly important for maintaining physical and mental health^4,5. Further, lifestyles established in young adulthood, including diet and sleep habits, may persist over time and impact chronic disease risk⁶.

Many lifestyle and socioeconomic factors contribute to poor sleep and especially those prevalent in young adulthood, including lack of a consistent bedtimes⁷, exposure to light at night⁸ (e.g. through electronics), work, school, and family-related constraints and economic stressors³. Recently, diet has been highlighted as another lifestyle factor that could affect sleep, although it has rarely been examined specifically in the young adulthood period. Examples of findings among middle age and older adults include the MESA study of adults aged 45 to 84 years where adherence to a Mediterranean dietary pattern was related to fewer insomnia symptoms over a 10-year period⁹, although the findings did not show symptoms by gender. In addition, a greater consumption of healthy fruit and vegetable diets in Mexican women aged 41 years at baseline was related to higher sleep quality after a 6 year follow-up.¹⁰ Another small study of <100 adults (average age 47) found participants who followed dietary advice to consume more anti-inflammatory diets over a 3-month period experienced higher sleep efficiency and shorter wake after sleep onset compared to those whose diets became more pro-inflammatory¹¹. Despite the small sample size, the study design helped ascertain whether positive dietary changes may improve sleep.

The MENU GenY study, a randomized dietary intervention study, provides a similar opportunity to examine a relationship between sleep and FV consumption among a large sample of young adults. In this study, young adults with less than recommended daily FV intake (<5 servings of FV/day at baseline) participated in an online program designed to increase their daily intake of fruits and vegetables. Information on sleep was also gathered at baseline and at follow-up surveys to evaluate whether participants who increased their FV intake also experienced improvements in their sleep compared to participants with no change or decrease in their FV intake. We hypothesized participants who reported an improvement in their FV intake would also experience improvements in their sleep, and that this relationship may depend on gender, based on our previous cross-sectional baseline findings regarding sleep and FV intake.³

METHODS

A secondary longitudinal analysis of data was conducted utilizing the baseline and 3-month follow-up data collection from a randomized trial (Encouraging Young Adults to Make Effective Nutrition Choices: MENU GenY, R01HD067314; ClinicalTrials.gov Identifier NCT01979809)) that sought to improve FV intake using a tailored online intervention. Participants were recruited in 2013 to 2014 by mailed invitation from an urban center (Henry Ford Hospital, Detroit, MI) and a mainly rural location (Geisinger Health System, Danville, PA). Eligibility criteria included young adults ages 21–30 years old who received any medical care (including routine yearly examinations) from Geisinger Health System or Henry Ford Health System within the previous year, and report of eating less than 5 servings/cups of FV per day. Young adults between the ages of 18–21 were not included because the young adult literature typically considers 21 years as the beginning of young adulthood. No exclusions were made based on health conditions other than for mental or cognitive conditions that may impact the ability to answer the questionnaire. Following online informed consent approved by each institution’s IRB, participants were randomized into one of three intervention arms. The control arm was an untailored web-based program to encourage higher FV consumption. The second arm was an age-targeted tailored web-based program, and the third arm additionally included personalized e-coaching support. At baseline and 3 months follow-up, participants completed questionnaires regarding socio-demographic and lifestyle characteristics including validated assessments of fruit and vegetable consumption and sleep characteristics. Of the 1444 participants enrolled in the study, 1165 participants reported their FV intake at baseline and 3-months and at least one of the sleep characteristics at both points.

Fruit and vegetable intake

A two-item instrument previously developed and validated with a description of serving sizes¹² was used to ascertain the participant self report of the number of servings of fruits and vegetables typically consumed per day including juice. Numeric response options were none, 1, 2, 3 or ≥4 servings/day. Responses to these two questions were summed (a response of ≥4 was imputed as 4), yielding each participant’s total FV intake in servings/day. We then calculated the change in FV intake from baseline to 3 months, and categorized the change into four categories: decrease or no change, or improvement of 1 serving/day change, 2 servings/day change, or 3+ servings/day change.

Sleep measures

At baseline and 3 month follow-up, participants were asked a subgroup of questions pertaining to sleep from the Pittsburgh Sleep Quality Index¹³, a validated instrument in young adult populations¹⁴. The sleep questions used included number of hours of actual sleep typically obtained (separately for weekday and weekends), a rating of the typical quality (ranging from poor (=1) to excellent (=5) on a 5-point Likert-type scale), and the average amount of time it took to fall asleep within the past month (sleep latency; ranging from immediately to ≥2 hours in 15-minute increments). Participants were also asked a single question of whether they had experienced a difficult time getting to sleep, staying asleep, or difficulty with non-refreshing sleep at least 3 nights per week over a period of 3 months within the past year (a clinical marker of chronic insomnia¹⁵). If they responded yes, participants were asked to report the average number of days per week they typically experienced these problems, range 1–7 days. Insomnia symptoms were categorized as never to <3 days per week versus 3 or more days per week, consistent with the criteria for chronic insomnia (insomnia symptoms at least 3 nights per week over a period of 3 months within the past year¹⁶). The initial justification for including these four sleep constructs was to investigate insomnia within young adults.

To evaluate change in sleep characteristics, the baseline sleep duration, quality, and amount of time to fall asleep was subtracted from the corresponding 3-month sleep value. For insomnia, three categories were created: no change, improvement (classified as having insomnia at baseline but no insomnia at 3-month follow-up), and decline (not classified as having insomnia at baseline but meeting the criteria for insomnia at the 3-month follow-up).

Baseline covariates

The baseline survey asked participants to report socio-demographic information as well as lifestyle factors. For the present analysis based on previously-reported associations with diet and sleep³ the following demographic and lifestyle characteristic variables included: gender, age, race/ethnicity, highest education attained, highest education level of childhood primary guardian (proxy of young adult SES^17,18), current education and work status, marital status, living arrangements (live alone, with roommate, parents, partner/spouse), number of children present in the home, height and weight (used to estimate BMI in kg/m²), depression symptoms, perceived stress, typical physical activity, smoking and alcohol habits, and study site. Depression symptoms were measured with four items from the PROMIS Emotional Distress- Depression Short Form (www.nihpromise.org/measures/samplequestions), with questions assessing feelings of worthlessness, helplessness, depression, and hopelessness (responses range from never to always on a 5-point Likert-type scale; Cronbach’s alpha= 0.92). The individual components were added and divided by 4 to create a cumulative depression symptom score. Perceived stress was measured with a shortened version (5 questions) of the Perceived Stress Scale¹⁹ that had responses from never (=1) to fairly often (=5) on a 5-point Likert-type scale (Cronbach’s alpha=0.77). The scores were summed, divided by 5, and used as a composite stress measure. Typical physical activity was assessed with the International Physical Activity Questionnaire,²⁰ a 6-item scale that estimates average minutes of sedentary, light, moderate, and vigorous activity. Responses from these questions were used to calculate average metabolic equivalents (METS)/day using standard equations. For smoking data, participants were asked to self report if they smoked at least 100 cigarettes in their lifetime and if yes, asked if currently smoking. For analysis, responses were coded as never, former or current smoking²¹. Drinking alcohol was collected by asking how often they had an alcoholic beverage in the past year, ranging from never, monthly, weekly to ≥6 times per week²² and categorized as 0, <3, and 3 or more drinks/week. Other covariates are categorized as shown in Table 1.

Table 1.

Sociodemographic and lifestyle characteristics in relation to change in sleep over 3 months

	N (Percentage)	Change in Weekday sleep duration (hours/night), Mean ± SD	Change in Weekend sleep duration (hours/night), Mean ± SD	Change in Perceived sleep quality (1=poor, 5=excellent) Mean ± SD	Change in Time to fall asleep (minutes) Mean ± SD	Change in insomnia symptoms (difficult time falling asleep, staying asleep, or non-refreshing sleep)
						Stayed same	Insomnia symptoms worsened	Insomnia symptoms improved
Overall	1165	0.12±1.53	0.02±1.52	0.13±0.87	−1±21	76.29%	9.97%	13.75%
Sex
Women	827(70.99%)	0.17±1.53	−0.01±1.48	0.09±0.86	−1±21	75.00%	10.44%	14.56%
Men	338(29.01%)	0.01±1.51	0.07±1.61	0.23±0.87	0±21	79.41%	8.82%	11.76%
P value1		0.056	0.582	0.012	0.302	0.271
Age
21–23 years	345(29.61%)	0.14±1.63	0.02±1.65	0.13±0.91	−2±23	77.33%	9.59%	13.08%
24–26 years	348(29.87%)	0.12±1.61	−0.01±1.52	0.17±0.87	−2±21	76.66%	8.93%	14.41%
27–30 years	472(40.52%)	0.12±1.38	0.02±1.42	0.10±0.83	1±19	75.26%	10.99%	13.74%
P value		0.988	0.857	0.616	0.375	0.868
Study Site
Henry Ford	607(52.1%)	0.05±1.52	0.02±1.53	0.16±0.90	−1±21	78.09%	13.84%	8.07%
Geisinger	558(47.9%)	0.20±1.53	0.01±1.51	0.09±0.83	−1±20	74.33%	13.64%	12.03%
P value		0.071	0.883	0.082	0.87	0.078
Race/Ethnicity
White	924(79.31%)	0.14±1.51	0.05±1.47	0.11±0.87	−1±20	75.73%	13.54%	10.73%
African American	174(14.94%)	0.15±1.71	−0.11±1.73	0.19±0.85	−3±28	76.16%	16.28%	7.56%
Other	67(5.75%)	−0.17±1.17	−0.14 ±1.60	0.22±0.95	0±15	84.06%	10.14%	5.80%
P value		0.304	0.309	0.241	0.687	0.308
Highest education attained
High school grad or less	87(7.48%)	−0.11±1.97	−0.02±2.06	0.08±0.86	−1±32	70.11%	17.24%	10.73%
Some college	430(36.97%)	0.19±1.89	0.13±1.67	0.18±0.90	−1±22	75.00%	14.72%	7.56%
College grad or higher	646(55.55%)	0.11±1.14	−0.06±1.31	0.11±0.85	−1±16	77.90%	12.67%	5.80%
P value		0.45	0.128	0.4	0.792	0.524
Parent’s education
High school grad or less	273(23.62%)	0.07±1.74	0.03±1.57	0.07±0.87	0±24	74.55%	14.55%	10.91%
Some college	345(29.84%)	0.22±1.69	0.03±1.48	0.13±0.90	0±22	79.18%	11.44%	9.38%
College grad or higher	538(46.54%)	0.08 ±1.30	0.00±1.52	0.16±0.85	−1±18	75.14%	14.84%	10.02%
P value		0.291	0.953	0.569	0.504	0.584
Education status
Enrolled, not started	31(2.67%)	0.36±1.68	0.42±1.56	0.42±0.72	−4±26	74.19%	12.90%	12.90%
Part time student	127(10.93%)	−0.10±1.59	0.18±1.60	0.20±0.80	−2±20	74.60%	12.70%	12.70%
Full-time student	236(20.31%)	0.23±1.68	−0.03±1.60	0.21±0.92	1±21	80.08%	11.44%	8.47%
Not in school	768(66.09%)	0.12±1.46	−0.01±1.47	0.08±0.87	−1±20	75.39%	14.71%	9.90%
P value		0.191	0.047	0.014	0.524	0.693
Work status
Full time	678(58.4%)	0.10 ±1.24	0.01±1.42	0.09±0.88	0±17	76.44%	13.93%	9.63%
Part time	275(23.69%)	0.10±1.67	−0.03±1.42	0.18±0.81	−1±24	77.62%	11.55%	10.83%
Unemployed	188(16.19%)	0.35±1.94	0.12±1.89	0.19±0.92	−3±27	72.34%	17.02%	10.64%
Unable	20(1.72%)	−0.68±3.11	−0.20±2.16	−0.05±0.69	−2±19	85.00%	10.00%	5.00%
P value		0.217	0.479	0.289	0.256	0.656
Marital status
Casual	81(6.98%)	0.40±1.85	0.21±1.91	0.15±0.84	−1±19	67.90%	19.75%	12.35%
Committed	722(62.19%)	0.05±1.34	0.00±1.40	0.13±0.86	−1±18	76.45%	13.02%	10.53%
Separated, div., widow	13(1.12%)	0.58±1.26	0.33±1.19	0.00±0.71	−2±34	69.23%	30.77%	0.00%
Single, never married	345(29.72%)	0.18±1.79	−0.01±1.66	0.12±0.91	−1±25	78.20%	13.37%	8.43%
P value		0.366	0.527	0.939	0.972	0.175
Living arrangement
Live alone/other	194(16.71%)	0.23±1.77	−0.01±1.64	0.23±0.85	−1±21	73.58%	16.06%	10.36%
Friend/roommate	128(11.03%)	0.18±1.47	−0.16±1.63	0.10±0.90	1±20	80.62%	6.98%	12.40%
Parents	387(33.33%)	0.14±1.75	0.05±1.63	0.08 ±0.88	−2±25	76.62%	15.06%	8.31%
Partner/spouse	452(38.93%)	0.05±1.20	0.05±1.33	0.13±0.85	0±17	75.94%	13.47%	10.60%
P value		0.768	0.238	0.191	0.876	0.229
Children in home
0	820(70.57%)	0.15±1.45	0.01±1.50	0.11±0.87	0±19	78.97%	12.47%	8.56%
1	203(17.47%)	0.07±1.69	−0.02±1.55	0.18±0.86	−3±24	68.47%	15.76%	15.76%
>1	139(11.96%)	0.07±1.73	0.15±1.58	0.16±0.84	−2±26	71.43%	18.57%	10.00%
P value		0.54	0.113	0.806	0.178	0.004
BMI
Normal	511(43.86%)	0.10±1.38	0.02±1.34	0.15±0.84	−1±18	77.73%	12.70%	9.57%
Overweight	291(24.98%)	0.09± 1.17	0.05±1.41	0.16±0.85	−2±19	76.82%	13.49%	9.69%
Obese	363(31.16%)	0.19±1.93	−0.02±1.81	0.07±0.91	0±26	73.83%	15.43%	10.74%
P value		0.74	0.806	0.319	0.901	0.748
Depression symptoms2
Q1, 0 to 1.2	384(33.22%)	0.02±1.20	−0.04±1.33	0.10±0.84	−1±17	82.03%	9.38%	8.59%
Q2, >1.2 to 2	276(23.88%)	0.12±1.56	−0.01±1.46	0.13±0.89	0±20	81.09%	11.64%	7.27%
Q3, >2 to 2.8	256(22.15%)	0.11±1.19	0.01±1.41	0.15±0.86	−1±21	70.20%	18.43%	11.37%
Q4, >2.8	240(20.76%)	0.24±2.09	0.13±1.94	0.16±0.89	−2±27	67.63%	18.26%	14.11%
P value		0.704	0.528	0.7	0.472	< 0.001
Perceived stress3
Q1, 0 to 2.4	341(29.6%)	0.05±1.13	−0.05±1.34	0.13±0.90	0±18	83.28%	10.26%	6.45%
Q2, >2.4 to 2.9	237(20.57%)	0.08±1.31	0.14±1.25	0.07±0.82	1±17	79.75%	10.97%	9.28%
Q3, >2.9 to 3.4	351(30.47%)	0.23±1.63	−0.05±1.50	0.15±0.83	−2±23	70.86%	16.57%	12.57%
Q4, >3.4	223(19.36%)	0.10±2.06	0.09±1.99	0.14±0.90	−1±25	69.96%	17.94%	12.11%
P value		0.515	0.176	0.92	0.553	0.001
Physical activity, METS per day
Q1, 0	276(24.67%)	0.05±1.68	0.06±1.81	0.21±0.90	0±25	75.09%	15.38%	9.52%
Q2, >0 to 12	325(29.04%)	0.06±1.24	−0.08±1.35	0.05±0.88	−1±20	73.15%	16.36%	10.49%
Q3, >12 to 32	261(23.32%)	0.12±1.32	0.12±1.28	0.12±0.81	1±18	79.17%	12.50%	8.33%
Q4, >32	257(22.97%)	0.23±1.70	−0.04±1.61	0.12±0.86	−4±19	78.29%	10.08%	11.63%
P value		0.704	0.156	0.183	0.031	0.285
Smoking status
Never	946(81.27%)	0.11±1.44	−0.04±1.45	0.12±0.85	−1±20	76.72%	13.44%	9.84%
Former	102(8.76%)	0.11±1.67	0.12±1.38	0.05±0.89	−3±24	78.43%	12.75%	8.82%
Current	116(9.97%)	0.23±1.99	0.42±2.07	0.30±0.98	−1±24	71.55%	17.24%	11.21%
P value		0.494	0.148	0.156	0.598	0.751
Alcohol consumption
0 drinks/week	214(18.46%)	0.09±1.88	−0.08±1.62	0.06±0.88	−2±22	78.40%	11.74%	9.86%
<2 drinks/week	500(43.14%)	0.15±1.23	0.06±1.33	0.13±0.90	−1±21	75.45%	14.37%	10.18%
2 or more drinks/week	445(38.4%)	0.11±1.64	−0.01±1.66	0.16±0.83	0±21	76.35%	13.74%	9.91%
P value		0.524	0.184	0.279	0.859	0.916

¹P values are from Kruskal Wallis for continuous variables and Chi-square tests for categorical variables

²From the PROMIS Emotional Distress-Depression Short Form, ranging from 0 to 5 (sum of responses to 4 questions, divided by 4)

³From the Perceived Stress Scale, ranging from 0 to 5 (sum of responses to 5 questions divided by 5)

Statistical Analysis

To evaluate potential confounders, means ± standard deviations (SD) of changes in sleep duration, perceived sleep quality and time to fall asleep were estimated according to categories of socio-demographic and lifestyle characteristics. For changes in insomnia symptoms, proportions within each category were computed according to the socio-demographic and lifestyle characteristics. For continuous sleep variables, P values were computed using Kruskal Wallis tests, while Chi-square tests were computed for the dichotomous insomnia symptoms.

To evaluate the primary research question, means ± SD of changes in sleep duration, perceived sleep quality and time to fall asleep were estimated according to categories of FV intake change. The proportions of insomnia change categories were computed according the FV change categories. In adjusted analysis, the continuous sleep change outcomes were analyzed with linear regression models, using the “no change or decrease in FV” category as the reference and adjusting for baseline sleep, gender, stress, children in the home, and study site. Insomnia change categories were analyzed with multinomial logistic regression and were adjusted for gender, stress, children in the home, and study site. All analyses were run non-stratified and stratified by gender, to evaluate the presence of gender-specific associations, which have been found in prior work³. Preliminary analysis showed that associations did not depend on intervention arm, so final models were not stratified by intervention. Further, adjustment for treatment arm did not substantially alter estimates. Analyses were run using R statistical software version 3.3.1 (R Project for Statistical Computing, Vienna, Austria).

RESULTS

The average age of the young adults in this sample was 26 ± 2.8, 71% were women and 79% white. Average sleep duration was 7.24 ± 1.31 on weekdays, and 8.16 ± 1.47 on weekends (Supplemental Table 1). Thirty-five percent met the criteria for insomnia and it took on average 27 ± 24 minutes to fall asleep. Over the 3-month period and across all subjects, sleep duration, sleep quality and time to fall asleep, on average, remained fairly constant, while presence of insomnia symptoms remained the same for approximately 75%, and worsened and/or improved for the remaining 25% (Table 1). Women experienced greater changes in weekday sleep duration, while men had higher improvement in sleep quality. Young adults who were enrolled in post-secondary school (not yet started) experienced greater improvements in weekend sleep duration and sleep quality than currently-enrolled students or those not in school. Participants with one or more children in the home, and those with higher depression and stress were more likely to experience changes in insomnia symptoms (i.e. either worsening of insomnia symptoms or improvements). Finally, those with the highest baseline physical activity had improvements in the amount of time to fall asleep compared to those with no physical activity.

Participants on average increased their FV intake by 1.2 ± 1.4 servings, although 23% did not change their FV intake at all, and 8% reduced their intake over the 3 months. Although not a primary outcome of the study, there were essentially no changes in BMI over this period (median [Q25, Q75]= 0 [−0.66, 0.37]). Changes in FV intake were not associated with changes in sleep duration overall or in gender-stratified analysis, except that males who increased their FV intake by 3 servings or more had 0.5 hour shorter sleep duration (95% CI −1.0 to −0.01) on weekends than men whose FV intake did not change or decreased (Supplemental Table 2). However, change in FV intake was associated with improvements in time to fall asleep (Table 2) and reduction of insomnia symptoms (Table 3). Gender-stratified analyses revealed that these differences were found primarily among women, and that there was an additional positive association with increased FV intake and sleep quality only among women. To illustrate, women who increased intake by 3 FV servings/day at 3 months were 2 times more likely to report improved insomnia symptoms (i.e. to change from meeting the threshold for chronic insomnia at baseline to not meeting the threshold 3 months later; 95% CI 1.1 to 3.6) after adjusting for depression, children in the home, and study site (Figure 1 and Table 2). Further, women who increased their FV intake by 3 servings reported, on average, a 4-minute reduction in time to fall asleep (95% CI −8 to 0; P=0.04) after adjusting for baseline sleep quality, depression symptoms, education status, physical activity and study site. They also experienced a 0.2-point higher sleep quality score (95% CI −0.01 to 0.3; P=0.06), and were 80% more likely to change from having poor to good/excellent sleep quality (95% CI 6% to 208%), compared to women who did not change or reduced their FV intake. In sensitivity analyses where we dichotomized the exposure variable into those who changed their FV intake to meet the target level of 5 servings/day versus those who remained below this threshold, we found that the associations were in the same direction but were not statistically significant.

Table 2.

Change in FV intake in relation to change in sleep characteristics among MENU GenY participants

	N	Change in sleep quality score (SD)	Adjusted change, sleep quality score (95% CI)1	Change in time to fall asleep (SD), minutes	Adjusted change, minutes (95% CI)1
Change in FV intake
No change or decrease	352	0.08(0.9)	Reference	1(19)	Reference
Increased by 1 serving	363	0.17(0.82)	0.08 (−0.04,0.20)	−1(21)	−1 (−4, 1)
Increased by 2 servings	252	0.04(0.84)	−0.03 (−0.15,0.09)	−1(23)	−3 (−6, 0)
Increased by 3+	193	0.26(0.92)	0.09 (-0.05,0.23)	−4(20)	−4 (−7, −1)*
MEN
Change in FV intake
No change or decrease	111	0.23(0.98)	Reference	1(17)	Reference
Increased by 1 serving	108	0.31(0.74)	0.07 (−0.15,0.29)	−1(19)	−2 (−7, 3)
Increased by 2 servings	69	0.09(0.85)	−0.15 (−0.39,0.09)	−1(26)	−4 (−10, 1)
Increased by 3+	50	0.26(0.9)	−0.05 (−0.32,0.22)	−2(23)	−5 (−11, 1)
WOMEN
Change in FV intake
No change or decrease	241	0.01(0.85)	Reference	1(20)	Reference
Increased by 1 serving	255	0.12(0.85)	0.1 (−0.04,0.24)	−2(22)	−1 (−5,2)
Increased by 2 servings	183	0.02(0.84)	0.01 (−0.15,0.17)	−1(21)	−2 (−6,1)
Increased by 3+	143	0.26(0.92)	0.15 (−0.01,0.31)	−5(20)	−4 (−8,0)*

¹Adjusted for baseline sleep quality or time to fall asleep, gender (only in unstratified models), depression symptoms, education status, physical activity, and study site

^*Statistically significant at P<0.05

Table 3.

Change in FV intake in relation to change in insomnia symptoms among MENU GenY participants

	N	Odds of improvement in insomnia symptoms1	Odds of worsening insomnia symptoms1
Change in FV intake
No change or decrease	355	Reference	Reference
Increased by 1 serving	362	1.16 (0.73,1.87)	1.25 (0.77,2.06)
Increased by 2 servings	253	1.54 (0.95,2.50)	0.9 (0.51,1.61)
Increased by 3+	193	1.78 (1.07,2.97)*	1.07 (0.57, 1.98)
MEN
Change in FV intake
No change or decrease	113	Reference	Reference
Increased by 1 serving	108	1.70 (0.71,4.05)	0.71 (0.26, 1.95)
Increased by 2 servings	69	1.18 (0.41,3.38)	0.84 (0.29,2.45)
Increased by 3+	50	1.03 (0.32,3.28)	1.30 (0.43,3.94)
WOMEN
Change in FV intake
No change or decrease	242	Reference	Reference
Increased by 1 serving	254	1.03 (0.58,1.84)	1.50 (0.84,2.69)
Increased by 2 servings	184	1.73 (0.99,3.02)	0.89 (0.45,1.78)
Increased by 3+	143	1.98 (1.10,3.56)*	0.98 (0.46,2.08)

¹Adjusted for gender (only in unstratified models), depression, children in the home, and study site

^*Statistically significant at P<0.05

Figure 1.

Bivariate associations between change in FV intake and change in insomnia symptoms

DISCUSSION

This study conducted a unique exploration of the sleeping quality and patterns among a large and diverse group of young adults who enrolled in an online intervention study to improve FV consumption. The sample of 71% women and 29% men allowed for stratification analyses by gender. Comparing baseline to the 3-month follow-up report, men and students who were enrolled but not yet started reported improvements in sleep quality, and young adults who reported highest physical activity also reported a reduction in time to fall asleep.

Young adults who increased their FV consumption by 3 or more servings experienced modest improvements in sleep latency and insomnia over a 3-month period compared to participants with no change or smaller increases in FV intake, although no differences in sleep duration were noted (data not shown). Analyses stratified by gender revealed that these associations were primarily found among women. To highlight, women who increased their FV intake by 3 or more servings reported a 4-minute shorter time, on average, to fall asleep at follow-up, and 2-fold higher odds of improvement in insomnia symptoms.

These findings provide evidence that increasing consumption of fruits and vegetables could improve sleep quality and insomnia symptoms, results that are in line with recent evidence. Studies in both the US and Mexico have shown that healthy diet patterns are prospectively related to better sleep quality, fewer insomnia symptoms, and earlier timing of sleep.^9,10,23 Similarly, the present study represents an extension of this prior work in a rarely studied population of young adults, with empirical evidence that directly links changes in FV consumption with changes in sleep. To our knowledge, only a few other studies have used a similar design. In one study of older adults, Wirth et al showed that adults who were advised to consume a more anti-inflammatory diet, composed primarily of plant-based foods and seafood, experienced improvements in sleep quality but not sleep duration.¹¹ Most likely due to a limited sample size, it was unclear from the Wirth study whether gender differences were apparent. Another study, a randomized trial of 49 men with insomnia and overweight/obesity, found that men in the dietary intervention arm (300–500 kcal reduction/day) had shorter sleep onset latency by the end of the 6-month follow-up than controls.²⁴ The role of individual components of the diet on sleep was not assessed.

There are a few mechanisms that could be responsible for a link between FV consumption and improvements in sleep. In line with the Wirth study, fruits and vegetables are important constituents of an anti-inflammatory diet, and anti-inflammatory diets may help improve sleep by promoting production of melatonin and other neurotransmitters involved in onset and maintenance of sleep.²⁵ Further, increased consumption of FV may lead to improvements in other aspects of the diet, such as reductions in processed foods, meat, and/or snacking, which have each been related to lower quality of sleep.^26–28

In gender-stratified analysis, it was evident that the associations between FV changes and improvements in time to fall asleep and insomnia symptoms, overall, were found among women and not men. Thus, these findings could have particular relevance for women with chronic insomnia. Although a difference of 5 minutes in the amount of time to fall asleep is pretty modest, these results may be conservative since they were self-reported. Further, insomnia symptoms encompass not only the amount of time to fall asleep, but also the inability to maintain sleep and/or awaken too early. Indeed, a 2-fold difference in experience of chronic insomnia (which includes all of the symptoms) among women with the highest changes in FV consumption indicates potential for a clinically meaningful effect.

Reasons for the gender difference are not entirely clear, although it is noteworthy that women have a higher prevalence of insomnia compared to men.²⁹ Thus, it may be that insomnia-related symptoms are more sensitive to change among women. Although the study population comprised only women, a study of Mexican teachers found that those who consumed a diet rich in fruits and vegetables at baseline had higher overall sleep quality as well as lower sleep onset latency, higher sleep efficiency, fewer sleep disturbances, and less daytime dysfunction after 6 years of follow-up.¹⁰ Another potential reason for gender differences observed in our study could be that women who increased their FV intake also made other beneficial changes in their diet (e.g. fewer processed foods or lower total calories) while men who increased their FV intake did not make substantial changes in other aspects of their diet. A final potential reason could be related to magnesium intake, a nutrient found in leafy green vegetables. Magnesium intake has been associated with improvements in insomnia symptoms³⁰, and its daily requirements differ by gender.³¹

There are both strengths and limitations of this study to consider. One of the primary strengths was the study design, which allowed us to directly correlate changes in FV consumption with concomitant changes in sleep. We could also examine potential confounding by a large number of baseline variables within the understudied population of young adults. In addition, we had a large sample size, allowing us to evaluate gender-specific associations. Nonetheless, we acknowledge that the sample was more heavily female and thus the male-specific models may be underpowered. There are some other limitations as well. First, sleep characteristics were self-reported, contributing to measurement error (resulting both in underestimation and overestimation). However, we do not expect this measurement error to be differential with respect to FV intake and therefore should not be a source of bias. Another limitation of the sleep assessment is that we did not administer the full Pittsburgh Sleep Quality Index to avoid survey burden. The participants investigated in this study were enrollees in an online health promotion, year-long study so they may not represent a generalizable subsample of young adults. In addition, the study population included young adults who were low-consumers of FV (<5 servings/day). FV intake was also self-reported at both time points, which could have resulted in an overestimation of consumption, although the validation exercises of this FV intake instrument suggest responses represent accurate daily intake. Again, we do not expect that any misreporting of FV intake would have been differential with respect to sleep characteristics. Another limitation related to dietary reporting was that we did not have a comprehensive assessment of the overall diet, which did not allow us to assess how other components of the diet changed at the same time as FV consumption changed. As with any observational study, there could be unmeasured confounders associated with both the changes in FV consumption and changes in sleep, such as an underlying health condition. Finally, there always exists the possibility that results were due to a chance finding, although the fact that we found associations across multiple sleep outcomes and not only one means this is less likely.

CONCLUSIONS

In summary, within a sample of young adults taking part in a FV intervention study, women who increased their FV intake by 3 or more servings experienced concomitant improvements in insomnia-related symptoms compared to women who did not change their FV intake (or who slightly decreased their intake). These findings highlight the potential for dietary improvement as an additional therapeutic recommendation for women experiencing insomnia.