The link between allergic disease and depression in young adults: A structural equation modelling analysis

Background: The co-occurrence of multiple interacting medical and psychological disorders is extremely common. A signifi cant association between allergic disease (ADz e.g., asthma, allergic rhinitis, allergic dermatitis) and depression has been reported. Objective: Path analysis models using cross-sectional data were constructed to evaluate the inter-relationships among underlying co-occurring factors known to impact depression in young adults. Method: 885 college students (21.80 ± 3.22 years old) completed an online survey assessing current ADz, sleep quality, pain/discomfort, Body Mass Index, stress, substance use, Internet addiction, physical activity, social support and depression. Results: Our fi ndings suggest that ADz is only one of multiple factors modulating depression. Six factors had signifi cant direct effects (all p < .05) on depression: female gender, sleep quality, pain/discomfort, stress, Internet addiction, and social support. ADz, Body Mass Index, and substance use had only indirect effects (all p < .05). Social support was associated with better sleep, less depression and less stress. Physical activity had no measurable effect on depression. Conclusions: This study is the fi rst attempt to evaluate the complex interrelations between ADz, depression and co-occurring factors. The complex interactions between variables in the model highlight the need to study these factors conjointly. Future longitudinal studies are needed to determine the temporal order and to validate causal pathways. Research Article The link between allergic disease and depression in young adults: A structural equation modelling analysis Perla A Vargas1*, Agnes Bucko2 and Elias Robles3 1Associate Professor, Honors Faculty, School of Social and Behavioral Sciences, Arizona State University, 4701 W Thunderbird Rd, MC 3051, Glendale, AZ 85306, USA 2Graduate Research Assistant, Children’s Physical Activity Research Group, Department of Exercise Science, Division of Health Aspects of Physical Activity, Arnold School of Public Health, University of South Carolina, 921 Assembly St.Columbia, SC 29208, USA 3School of Social and Behavioral Sciences, Arizona State University, 4701 W Thunderbird Rd, MC 3051,Glendale, AZ 85306, USA Received: 08 August, 2018 Accepted: 08 October, 2018 Published: 09 October, 2018 *Corresponding author: Perla A Vargas, Associate Professor, Honors Faculty, School of Social and Behavioral Sciences, Arizona State University, 4701 W Thunderbird Rd, MC 3051, Glendale, AZ 85306, USA, Tel: 602-543-8224; E-mail:


Introduction
Depression is one of the most common mental health conditions [1]. It is characterized by a loss of positive affect with persistent symptoms including lack of self-care, poor concentration, anxiety, lack of energy, and loss of interest [2].
In 2011, 20.1% of U.S. adults reported depressive symptoms [3], with women more likely than men to suffer it (1.5 -3 times) [1,[4][5][6]. In 2016, adults aged 18-25 had the highest prevalence of major depressive episodes (10.9%) [1]. The economic burden of depression exceeds $210 billion per year, with less than half (45-47%) attributable to direct costs [7]. But, this cost can be seriously underestimated if the long-term personal burden of depression, such as reduced educational attainment, lower earning potential, increased chance of teenage childbearing, higher unemployment, and increased work disability are not taken into account [8]. Such long term personal burden has a more damaging impact on the young [9,10], (18-34 years) who experience the highest incidence (24%) [11] and highest cumulative prevalence of depression [12]. Further, the high incidence of depression combined with its current high relapse rate place emerging adults at particularly high-risk, with poorer psychosocial functioning, more severe chronic episodes, comorbid anxiety, and suicidality than any other age group [11]. Furthermore, although most patients with depression will not commit suicide, one third will make an attempt sometime during adolescence or young adulthood, and, 2.5% -7% will commit suicide by young adulthood [13]. Depression increases the risk for suicide by more than 15-fold [14]; further, increased depression severity increases risk for suicidality [15]. Suicide is a leading cause of death in the US; suicide rates increased by more than 30% from 1999 through 2016 [16]. Of great concern Depress Anxiety 4(2): 040-055. DOI: http://doi.org/10.17352/2455-5460.000033 is the increase in suicide rate among young people (< 34 years of age) [17,18]. In 2013, suicide was the second leading cause of death for people ages 15-24 years, the second for ages 25-34 years, the fourth for ages 35-54 years, the eighth for those aged 55-64 years, and the seventeenth for those ≥ 65 years [18]. It is speculated that the upsurge in depression rates over the past 10 -15 years [19][20][21], is the result of major societal and economic changes creating signifi cant stressors for emerging adults [22], such as lack of structure, poor time and disease management, instability of personal relationships, struggle for economic independence, and uncertainty about the future and one's competence to handle it [23][24][25]. According to the Cooperative Institutional Research Program Freshman Survey, college students' self-rated emotional health has been declining [26]. In 2017, 40.2% reported feeling so depressed "that it was diffi cult to function," two-thirds (62%) reported experiencing "overwhelming anxiety," nearly half (44.8%) reported more than average stress, and 11.6% reported tremendous stress [27]. Electronic media and the Internet in particular, have exacerbated stressors for young people. Internet usage in young adults and college graduates is pervasive [28]. Some researchers blame Internet use for the rise in depression [29][30][31][32] and suicide [33,34], arguing that spending too much time using electronic devices prevents young people from engaging in face-to-face social activities that might help protect against depression [35][36][37]. The increase in Internet use has been linked to clinical symptoms of dependence that interfere with a person's daily life, similar to gambling and drug dependence, leading to the recognition of Internet addiction as a psychiatric disorder associated with depression [38]. The link between Internet addiction and depression may be mediated by sleep disturbance or insomnia [30,39]. These fi ndings are pertinent here because college students experience insuffi cient and poorquality sleep, as well as growing levels dissatisfaction with their sleep [40][41][42]. Poor-quality sleep among young adults may be reaching epidemic proportions, with 73% of college students reporting sleep problems that include insuffi cient sleep, diffi culty falling asleep, sleep disturbances, and excessive daytime sleepiness [43,44]. Insomnia and sleep disturbances are considered to be core symptoms of depression; 2 however, recent studies suggest a different causal pathway [45][46][47].
Specifi cally, the evidence shows a detrimental effect of poor sleep on emotional functioning (i.e., daytime mood, emotional reactivity and regulation) [48], suggesting that sleep diffi culties precede and trigger the symptoms of depression [45][46][47]49,50]. Depression is also associated with psychiatric and medical comorbidity, chronic pain, insomnia [49,51,52], impaired functioning, and suicidality [53,54]. For example, patients with persistent pain experiencing sleep problems at baseline had more than three times the risk of developing depression [51]. Untreated conditions such as allergy, pain, and insomnia have a tremendous impact on a person's quality of life. Yet, there is evidence that emerging adults, particularly young men, access health services less often than any other age group: 59% of young adults report having a usual source of care, and those without a regular source of care had signifi cantly lower utilization of outpatient services (except emergency room -ER visits), than those who have a regular source of care [55].
Further, poor-health practices and risky behaviors such as substance use increase during emerging adulthood [15,56], and have been clearly identifi ed as major risk factors for depression [57][58][59] and suicidality [60].
Most theoretical accounts of depression propose a diathesisstress model in which stressors may activate a diathesis or vulnerability, triggering a depressive event [61][62][63][64][65]. According to this view, there is a synergistic relation between the diathesis and stress which results in depressive symptomatology being greater than would be expected by their independent effects; that is, the effects are multiplicative. According to these models, vulnerabilities --temperament, physiological, cognitive or genetic characteristics --may negatively potentiate the effect of environmental stressors [64,65]. However, progress in understanding and treating depression as accounted by these models has been slow and challenged by methodological concerns [66]. Rather than vulnerabilities (e.g., temperament, attributional style, and specifi c genes), some argue, individual characteristics may function as plasticity factors, impacting how susceptible individuals might be to environmental infl uences [66]. Recently, a network perspective was proposed as a theoretical alternative to understand depression [67][68][69]. A network perspective assumes that depression is the result of direct interrelations among sets of symptoms. For example, symptoms (such as wheezing, runny nose, post-nasal drip) trigger other symptoms (like trouble with sleep, fatigue, and poor concentration) which, over time, lead to depressive symptoms and suicidal behaviors that further exacerbate stress, pain/ discomfort, and sleep. At the heart of the theory is the notion that symptoms of co-morbid conditions are causally connected through multiple mechanisms [69]. The presence of multiple interacting medical and psychological disorders is extremely common, and the link between mood disorders, particularly depression, and allergic disease (ADze.g., asthma, allergic rhinitis, and atopic dermatitis) is widely accepted [70][71][72][73][74][75][76]; however, the mechanisms underlying such associations are still not understood.
ADz is a major cause of disability, health resource utilization, and poor quality of life for those who are affected [77]. Asthma is the most common chronic disease among children and young adults [77]. In 2011, 15.6% of Americans 15-34 years reported lifetime asthma [78]. The condition may worsen due to allergic rhinitis (AR), a common comorbid, underdiagnosed, and often undertreated condition [79,80]. Approximately 80% of all patients with asthma have co-morbid AR [81]. Compared to those with asthma alone, patients with asthma and comorbid AR tend to have more asthma exacerbations, more visits to their primary care provider, more ER visits, and more hospitalizations [81]. It has also been shown that AR signifi cantly impacts all dimensions of sleep [81][82][83][84]; indeed, patients are considered to have moderate/severe AR if they experience sleep disturbance and/or impairment of daily activities [81][82][83][84][85]. An association between ADz and insomnia has also been reported. For example, Meltzer [86], found that adolescents with severe asthma get insuffi cient sleep, have poor sleep hygiene, and experience clinically signifi cant insomnia. The data revealed a dose response pattern where adolescents with severe asthma were more likely to report insomnia than all other comparison groups: adolescents with no asthma, lifetime-asthma, mild, or moderate asthma [86]. In addition, compared to patients without asthma and/or AR, patients with asthma [87][88][89][90] and patients with AR are more likely to report symptoms of depression [70,72,74,[81][82][83][84][85][91][92][93][94][95].
In recent years, numerous cross-sectional studies have shown a link between obesity and asthma [96][97][98][99]. Obesity is one of the most ubiquitous conditions worldwide [100]. In 2017, 33.6% of college students in the US were overweight or obese [27]. Further, obesity has been associated with reduced quality of life and multiple other comorbidities including insomnia / sleep disturbances [97,101,102,103], pain / discomfort [97,104] and depression [97,101,105,106], but not atopy [97,99]. Depression rates among obese individuals are 1.5-2 times higher than normo-weight individuals [105,107,108]; this increased rate is more evident among females [105]. There is evidence suggesting that depression may be an underlying mechanism linking increased Body Mass Index (BMI ≥ 25) and poor asthma control [98,109,110]. Given its high prevalence and impact on quality of life, identifying factors that may prevent or reduce the impact of depression is paramount; evidence suggests that social support [36,37,111,112] and physical activity (PA) [113][114][115], are two promising protective factors in the prevention of future depression.
The evidence suggests that depression is a multifactorial disorder. However, we found no published studies looking at the inter-relationships among factors known to concurrently impact depression such as ADz, sleep quality, pain/discomfort, BMI, stress, substance use, Internet addiction, physical activity, social support, and depression among emerging adults. The purpose of this exploratory study was to investigate these relationships utilizing multivariate modeling that allows for the simultaneous statistical examination of multiple relationships [116]. To address the objective of this study, we constructed seven quantitative models and tested them using multivariate techniques (i.e., structural equation modeling -SEM), which afforded us the opportunity to examine both direct and indirect effects of variables infl uencing depression. We hypothesized that 1) ADz is signifi cantly associated with depression only indirectly through its negative impact on sleep quality; 2) the impact of ADz on depression is also mediated by stress and pain / discomfort); 3) BMI is only indirectly associated with depression through its negative impact on sleep quality, increased pain / discomfort, and stress; 4) higher stress is directly and indirectly associated with higher depression through its direct negative impact on sleep quality and increased pain / discomfort; 5) substance use is directly and indirectly associated with depression through its negative impact on sleep quality, increased pain / discomfort, and stress; 6) Internet addiction is directly and indirectly associated with depression through its negative impact on sleep quality and stress; 7) higher levels of physical activity and social support are directly and indirectly associated with better sleep quality and less stress, pain / discomfort, and depression.

Participants
A sample of 985 college students was recruited from a large university in the southwestern U.S. who completed the study to satisfy a class requirement. Participants were told the purpose of the survey was to study the lifestyle of college students. Thirty-nine non-traditional students ≥ 34 years of age were excluded, 46 records were excluded due to missing data to evaluate sleep, and 14 records were excluded for other missing data, yielding a total sample of 885 complete records. The sample's average age was 21.80 (SD=3.22) years, mostly females (79.4%). The study was approved by the University Committee on Human Subjects. No personal identifi ers were collected and, after reading a description of the study, participants gave passive consent by proceeding to complete an online survey.

Measures
Depression was measured by the Center for Epidemiologic Studies Depression Scale (CESD) [117]. To prevent overlap and the potential infl ation of correlations, item 11 (My sleep was restless) was removed; thus scores in the CESD 19 ranged from 0 to 52. The CESD has good test-retest reliability and has been validated in various populations, including adolescents [118], (Cronbach's  for the CESD 19 was .917). A CESD score ≥ 16 identifi es individuals with clinically relevant symptoms [117].
Allergic disease was ascertained by several questions: fi rst, students were asked "have you ever received a physician's diagnosis of asthma (No or Yes); students who answered yes were asked if they "have been taking any prescribed medications or treatment for asthma" (No or Yes). Those who answered both questions positively were classifi ed as having current asthma (0-No or 1-Yes). Then, they were asked if they had been diagnosed with food allergies or allergies to airborne allergens such as dust mites, pet dander, mold, or pollen. Current allergy identifi cation was based on doctor-diagnosis of other allergies (i.e., dust mites, pet dander, mold, pollen, etc.) and a positive answer to either "Have you been prescribed medications to treat your allergy (ies)?" or "In the past month, did you take any over-the-counter allergy medication (such as Allegra, Benadryl, Claritin, Clarinex, Sudafed, and Zyrtec)?" (0-No or 1-Yes). A binary variable ADz (0-No or 1-Yes) was created by combining having current asthma and/or having current allergy.
Sleep was measured with the Pittsburgh Sleep Quality Index (PSQI). The PSQI is a standardized, questionnaire composed of 19 questions grouped into seven components (1: Subjective sleep quality, 2. Sleep latency, 3. Sleep duration, 4. Sleep effi ciency, 5. Sleep disturbance, 6. Use of sleep medication, and 7. Daytime dysfunction) with high levels of consistency, reliability, and validity (Cronbach's  for this study of .683) [119]. The PSQI was scored following the authors' proposed algorithm [119] and the global score (0 -21) was treated as a continuous variable (higher scores indicating worsening problems). A cut-off score > 5 has been recommended for clinical use as indicative of poor quality sleep [119].
Pain / discomfort was assessed by a question from the Canadian Community Health Survey, "Are you usually free of pain or discomfort?" [120] (0-No or 1-Yes). Those who answered "No" were classifi ed as experiencing pain / discomfort. successfully identifi ed 91% of the high-risk college drinkers [127]; compared to the traditional cutoff score of ≥8 that identifi ed only 82%.
Smoking and nicotine dependence were assessed with the Fagerström Test for Nicotine Dependence (FTND). The FTND is highly reliable, with scores ranging from 0 to 10; the higher the score, the more intense the physical dependence on nicotine (Cronbach's  for this study of .70) [128][129].
Internet addiction was measured using Young's Internet Addiction Test (YIAT) [14]. The YIAT consists of 12 questions to assess the effects of daily Internet use using a 5-point Likert scale (Cronbach's  for this study of .90) [14]. A total score, ranging from 12 to 60 is calculated by adding all items. A score > 30 is suggested as indicative of problematic use, and one > 37 as indicative of pathological use [14].
The International Physical Activity Questionnaire -short form (IPAQ) was used to assess physical activity (PA); it is a 9-item instrument consisting of six items reporting the number of days (frequency) and the number of minutes per day (duration) participants engage in PA during the last seven days. Due to a glitch, data on walking and sitting were lost. Thus, the PA score was calculated for vigorous and moderate activity as Metabolic Equivalent of Task (MET) minutes per week applying guidelines of the IPAQ research committee to the available data [130].
According to the scoring method, vigorous PA is equal to weekly PA expenditure multiplied by 8 METs, and moderate PA equals weekly expenditure multiplied by 4. A total MET weekly score was calculated by adding the METs for moderate and vigorous PA. This score has been used to classify individuals according to volume of activity as low (< 600 MET/min), moderate (600-1,499 MET/min), and high (≥ 1,500 MET/min) [131,132].
Satisfaction with social support was measured with the 6-item Social Support Questionnaire [133]. Participants were asked to indicate the number of supporters, and to rate their level of satisfaction with the support provided using a 6-point Likert scale. The Satisfaction Score (SSSQ-6) was calculated by adding all ratings (range 6 -36). Scores ≤ 33 are considered indicative of low satisfaction with available social support [35,134,135].

Procedure
All statistical analyses were performed using SPSS 24 and

Mediation models
Based on evidence from the literature, seven mediation path models were constructed using AMOS 24.0 [136,137] [4][5][6], stress and sleep [138][139] and the relationship between sleep and BMI [138][139][140][141]. Mediation analysis assumes both causal and temporal relations and it has been used to attain a deep understanding of disease pathology and the mechanisms through which variables infl uence the condition [142]. When performed using strong evidence, prior theory, and with appropriate context, mediation analysis may provide the basis for future intervention research  [142]. Bootstrapping was used to assess the direct, indirect, and total effects of the predictors on outcomes [143]. The models goodness of fi t was ascertained using  2 (p > .05, suggesting the absence of meaningful unexplained variance); however,  2 is sensitive to sample size, thus we also used the ratio of  2 to df ( 2 / df -where a ratio smaller than 3 is considered indicative of an acceptable data model fi t). Furthermore, the Comparative Fit Index (CFI) and Standardized Root Mean Square Residuals (SRMR) were also used as evidence of model fi t. Values greater than .95 for CFI, and lower than .08 for SRMR are considered evidence of a well-fi tted model [144].

Results
Descriptive statistics for the variables in this study are shown in table 1. The sample was comparable to the student population in the psychology program, being mostly females (79.4% vs. 75.3% respectively), and Caucasian (58.1% vs. 55.9% respectively). One-hundred eleven (12.5%) students reported having current asthma, and 121 (13.7%) current allergy. Forty percent of those with "current asthma" also reported "current allergy" ( = .336; p < .001). About one fi fth of the students (20.7%) had either current asthma or allergy, and 5.5% had both. One fi fth (20.7%) of the students also reported pain / discomfort. Of the students who reported pain / discomfort (n=183), 27.3% had current ADz. Compared to students without ADz, signifi cantly more students with ADz reported pain / discomfort (18.9% vs. 27.3%, p = .013) and had higher PSQI scores (7.22 vs. 8.00, p = .002) which is indicative of poorer sleep quality. About one third of the students (35.3%) were either overweight or obese (BMI ≥ 25); and 13.6% were obese (BMI ≥ 30). The overall mean CESD score (20-item scale) was 16 [121][122][123][124]. The top quartile of the IDH scores ranged from 109 -184.
They were also more likely to be female, have reported lower levels of PA (i.e., MET min / week) and being less satisfi ed with their social support (all p < .05). Similarly, compared to their peers without the conditions, students with current ADz, and those who experienced pain and discomfort had signifi cantly poorer sleep quality, higher level of depression (CESD 19 ), and stress (all p < .05). In addition, students who experienced pain and discomfort were signifi cantly older, with higher BMI, and reported signifi cantly less satisfaction with their social support (p < .05).
Mediation Analysis. Several alternative models were compared in terms of fi t and effi cacy in accounting for interrelationships among ADz, sleep quality, pain/discomfort, BMI, stress, substance use, Internet addiction, SSSQ-6, PA, and depression in emerging adults. Gender was entered as a control variable in all models. As shown in table 3, all models ( Figure   1) showed acceptable fi t (all  2 p > .05, the ratio of  2 to df all < 3, CFI all > .95 and SRMR all < .08) [144]; the differences in fi t indices were likely due to a greater parsimony (i.e., fewer parameters) of Models 1 -6 relative to Model 7 (See Table 3).

Hypothesis 2: (the impact of ADz on depression is also mediated by stress and pain / discomfort). This hypothesis was supported;
ADz had a direct impact on pain / discomfort ( = .068, p < .05) and stress ( = .060, p < .05) that resulted in a signifi cant indirect effect on depression ( = .052, p < .05).

Hypothesis 3: (BMI is only indirectly associated with depression
through its negative impact on sleep quality, increased pain / discomfort, and stress). This hypothesis was supported: the impact of BMI on sleep ( = .056, p < .001), and depression ( = .065, p < .05) was only indirect. Also as predicted, BMI showed signifi cant effects on pain / discomfort (total effect of  = .131, p < .001, direct effect of  = .117, p < .001, and indirect effects  = .014, p < .05) and a small, but signifi cant effect on stress ( = .036, p < .05).

Hypothesis 4: (higher stress is directly and indirectly associated
with higher depression through its direct negative impact on sleep quality and increased pain / discomfort). This hypothesis was also supported, stress had a direct negative impact on sleep quality ( = .387, p < .001) and pain / discomfort ( = .161, p < .001). In addition, stress was directly ( = .512, p < .001) and indirectly associated with depression ( = .082, p < .001).

Hypothesis 5: (substance use is directly and indirectly associated
with depression through its negative impact on sleep quality, increased pain / discomfort, and stress). This hypothesis was only partially supported. We found that high-risk drinking was indirectly associated to sleep ( = .036, p < .05), pain / discomfort ( = .017, p < .001) and depression ( = .064, p < .001) by having a negative impact on stress (direct effect  = .105, p < .001). However, contrary to our hypothesis, there was no direct impact of high-risk drinking on depression, sleep, or pain / discomfort (all paths tested, p > .05). Further, tobacco use was not signifi cantly associated to depression or any of the mediating variables (all paths tested, p > .05) but the small direct and indirect effects of tobacco use added up to a small signifi cant total effect ( = .072, p < .05) on depression.

Hypothesis 6: (Internet addiction is directly and indirectly associated with depression through its negative impact on sleep quality and stress). This hypothesis was only partially supported.
Internet addiction had no direct impact on sleep quality, but had signifi cant indirect and total effects on sleep quality (total effect of  = .070, p < .05, direct effect of  = -.031, p >.05, and indirect effects  = .101, p < .001). As expected, Internet addiction had a signifi cant impact on stress ( = .287, p < .001) and depression (total effect of  = .237, p < .001, direct effect of  = .087, p = .002, and indirect effects  = .150, p < .001). In addition, Internet addiction had an unexpected positive direct impact on pain ( = -.102, p = .005), that was neutralized (total effect of  = -.056, p = .107) by the indirect effects mediated by stress ( = .046, p < .001).

Hypothesis 7:
(higher levels of PA and SSSQ-6 are directly and indirectly associated with better sleep quality and less stress, pain / discomfort, and depression). This hypothesis was also partially supported. Contrary to our predictions, PA --as measured in this study--had no measurable effect on sleep, pain / discomfort, or depression (all paths tested, p > .05). On the other hand, as predicted, SSSQ-6 was directly and indirectly associated with better sleep quality (total effect of  = -.225, p < .001, direct effect  = -.

Discussion
The co-occurrence of multiple interacting medical and psychological disorders is extremely common, and the link between mood disorders, particularly depression, and allergic disease is widely accepted. Our goal was to explore the interrelations among the co-morbid conditions that previous research has linked to depression which include ADz, BMI, sleep Table 4: Regression coeffi cients for Model 7 including standardized direct, indirect, and total effects for the predictor variables (i.e., ADz, BMI, alcohol and tobacco use, Internet addiction, PA as MET min/ week and Social Support), the mediating variables (i.e., sleep quality; pain/ discomfort, and stress) and the outcome variable (i.e., depression). Signifi cant effects are in bold. Also unexpectedly, PA --as measured in this study--had no effect on sleep, pain / discomfort, or depression. Finally, consistent with earlier literature, SSSQ-6 was associated with better sleep quality, lower stress, and less depression.

Standardized Regression Coeffi cients: Direct, Indirect and Total Effects
Because the interaction between participating factors is complex, studying the individual relationships between these variables is insuffi cient. For example, the bivariate analysis suggests that a signifi cant correlation exists between ADz and depression. Also, the SEM models 1 and 2 (Fig 1)

Multimorbidity among emerging adults
The prevalence of multimorbidity has increased recently in the general population. Both, the proportion of people affected and the number of co-occurring morbidities tend to increase with age. It has been estimated that 11.3% of adults ages 25 to 44 have at least two co-morbid physical conditions, and that 5.7% suffer co-morbid physical and mental health disorders [149]. There is evidence that co-occurring psychiatric and substance use disorders, for example, are associated with more functional impairment than either type of disorder alone [150]. Moreover, multimorbidity has a stronger negative impact in younger people [151]. During the transition into adulthood emerging adults must assume responsibilities known to impact their own well-being such as meal preparation, self-regulation of substance use, management of work and sleep routines, and in the case of students with multimorbidity, appropriate health care management. Managing multimorbidity is particularly challenging for young adults in view of the complexity and possible interactions between conditions, as well as the students' constraints of time and resources. This is of great concern given that over one third of our students were overweight or obese and one fi fth of them reported suffering either asthma and/or allergy. In addition, 42.8% of the students were classifi ed as having symptoms of depression (CESD ≥ 16). The high rate of self-reported symptoms of depression in our students is similar to that reported by children and adolescents (42.8% vs. 20 -50%) in epidemiologic studies [152] and similar to estimates using the CESD in college populations (36.8%, 95% CI, 35.2-38.4) [153]. Further, more than two-thirds of the students reported experiencing poor sleep quality (PSQI > 5), and only 25.2% reported the recommended eight or more daily hours of sleep. These fi ndings are consistent with evidence from the literature showing changes in sleep architecture during adolescence and early adulthood resulting from both social and biological factors [140]. The evidence suggests that during this period of susceptibility, inadequate sleep can trigger a cascade of negative health events, including depression. College counseling centers report a signifi cant increase in the number of students seeking mental health services, as well as worsening of the severity of their conditions [26,154,155]. There is evidence that the self-rated emotional health of college students has been declining [26]. For instance, about half of the college students report feeling so depressed "that it was diffi cult to function," and experiencing "overwhelming anxiety," and high levels of stress [27]. Finally, students with ADz experienced more pain / discomfort, worst sleep quality, and more depression.
In addition, alcohol consumption has continued to increase in the US among all age groups [156], but young adults have consistently had the highest prevalence of alcohol use (80.1%), high-risk drinking, and dependence than any other group [156,157]. It is estimated that roughly one fi fth of alcohol users ages 18-29 meet DSM -IV criteria for alcohol use disorder Depress Anxiety 4(2): 040-055. DOI: http://doi.org/10.17352/2455-5460.000033 [156,157]. For college students, drinking is a rite of passage, part of the college experience [158]. Findings in our study show drinking patterns comparable to those obtained in national surveys, with most students reporting past-month drinking (70.2% vs. 63.2%), forty percent (40.2% vs. 37.9%) past-month binge drinking [159], and roughly one fi fth indicating possible alcohol dependence [156,157]. Positive alcohol expectancies are associated to increased drinking [160], particularly in a college setting [158,161]. For example, college drinkers reported both positive and negative consequences from drinking, but most (96.6%) reported at least one positive consequence such as "having a good time" and "feeling less stressed", while 82% of drinkers reported at least one negative consequence [161]. The authors noted that positive consequences were associated with lower levels of drinking, a pattern not necessarily perceived by the students [161]. This mixed experience among drinkers may explain why drinking in this population does not appear to be directly related with depression. Cigarette smoking has changed signifi cantly in the last decades [162]. Consumption of cigarettes has decreased signifi cantly overall, and light smoking and smoking e-cigarettes have become the predominant forms of tobacco use among young adults [163]. The prevalence of smoking among our study participants was quite low (5.4%) and most smokers were classifi ed as light smokers (i.e., < 10 cigarettes a day). Although occasional smoking and light smoking were previously considered transient behaviors, they are now a common social practice for college students [164]. There is some evidence suggesting that social-lightsmoking (i.e., smoking primarily in social contexts) may be growing among college students [164]. Interestingly, even though tobacco use was not -directly or indirectly-associated to depression or any of the mediating variables in this study, it had a signifi cant total effect on depression, which suggests that tobacco use has an effect on depression but it is diffi cult to detect in these data.
The role of the Internet in the lives of young adults is clear.
According to the Pew Center, among 18-to 29-year-olds, 39% are connected online almost constantly and 49% go online multiple times per day [165]. Almost all of the students in our sample had an account in Facebook or other social networking site. This ubiquitous use has consequences for the day-to-day lives of young adults. About one fi fth of the students had an YIAT score suggestive of problematic or pathological Internet use [145]. Scientifi c evidence regarding the impact of internet use on the mental health of teens and young adults has been mixed [166]. It has been argued that the impact of online communication varies depending on the quality of the offl ine social network: that is, online communication with close friends and family enhances psychological well-being, while internet communication with weak social ties might not [167].
Research looking into the quality of friendships and social support as predictors of internet problematic use supports this explanation [168][169][170]. We found a signifi cant negative correlation between Internet addiction and perceived social support that suggests that this might also be the case among our students; in addition, together these two factors impacted depression. Lack of social support was a key factor predicting depression in our population, which is consistent with research evidence showing that social support improves well-being [36,37,111,112]. Consistent with Wang et al., [111] we found that social support regulated the effect of stress on depression.
Finally, about one fourth of our students (25.5%) reported engaging in recommended levels of physical activity (> 1500 MET min/week). However, we found no effect of PA on any of our outcome variables (i.e., sleep, pain / discomfort, stress, depression). There is some evidence suggesting that it is leisure-time physical activity [171], or even light PA (as pleasant activation) [172], rather than the energy expenditure of physical exercise that impact well-being. This bring us back to the importance of social support in modulating depression.
Pickett et al [171]. Found that higher levels of leisure-time PA, but not non-leisure time PA were signifi cantly associated with lower depression. Unfortunately, our data do not allow us to differentiate between leisure and non-leisure PA. Thus, this question remained unanswered.

Conclusions
This is the fi rst attempt at identifying potential factors involved in the relation between ADz and depression using multivariate data modeling to evaluate the causal pathways linking the two conditions, while accounting for multiple potential contributing factors (i.e., sleep quality, pain/ discomfort, BMI, stress, substance use, Internet addiction, physical activity, and SSSQ-6) among emerging adults. Our results are consistent with previous research, suggesting that allergic disease is only one of multiple factors modulating depression. Although the mechanisms underlying these relationships remain elusive, there is evidence showing that infl ammatory processes can be the underlying physiological mechanism linking many of these conditions including Adz [172,174], obesity, stress, pain / discomfort, poor sleep quality, and depression [175]. Research exploring the underlying factors linking infl ammatory processes to behavioral outcomes suggests that infl ammatory processes and social behavior regulate one another. Proinfl ammatory cytokines aid in coordinating the body's response to illness, injury, or infection. Sickness behavior is part of this coordinated response of the organism to conserve energy and support recovery. Sickness behavior includes a constellation of symptoms such as fatigue, increased pain, anhedonia, loss of appetite, and social withdrawal or loss of interest in social activities. Multiple pathways have been suggested to explain how peripheral immune activation induces sickness behavior. These include increases in endotoxin and proinfl ammatory cytokine IL-6 production that can lead to increases in neural sensitivity, particularly to social stressors [176]. Thus, socially isolated individuals tend to have increased levels of infl ammation that further exacerbates infl ammatory responses to other social stressors [177]. Sickness behavior in patients with chronic infl ammatory disorders (e.g., allergic disease, pain, and obesity) may lead to hypersensitivity to social stressors (e.g., isolation and discrimination) increasing their risk for depression [176]. These factors may tart an interdependent uncontrolled cycle where chronic infl ammatory conditions increased depression exacerbation of chronic conditions, and so on. Following this rationale, research could be conducted to determine if anti-infl ammatory medications can impact symptoms of depression, for example. In a recent clinical trial, investigators showed a signifi cant benefi t of low dose buprenorphine on suicidal ideation; however, the benefi t on depression was smaller and not statistically signifi cant.
Nevertheless, this is a promising fi nding for the treatment of depression that warrants further research [178].

Limitations
Although mediation analysis assumes both causal and temporal relations, the proposed mediation pathways require further testing with prospective studies on a clinical population with ADz. Only prospective studies can establish the true nature of the association between the network of comorbidities included in this study, and answer with certainty whether the interactions among these factors increase the risk for depression among patients with ADz. Additionally, it is possible that age-related paths and trajectories may modify these relationships.
Our study was based on self-reported data and is, hence, susceptible to bias. The characteristics of the study participants, college students, limit the generalizability of our fi ndings. Finally, we did not collect detailed information on current symptoms of ADz, so, we cannot rule out the possibility that our fi ndings might be due to other extraneous variables or chance.