Introduction
Early-life infections are associated with an increased risk of serious mental health disorders in adulthood. There is an extensive literature linking prenatal maternal and childhood infections with schizophrenia and related psychotic disorders, which are associated with both central nervous system (CNS) and non-CNS infections (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011; Brown & Derkits, Reference Brown and Derkits2010; Dalman et al., Reference Dalman, Allebeck, Gunnell, Harrison, Kristensson, Lewis and Karlsson2008; Khandaker et al., Reference Khandaker, Dalman, Kappelmann, Stochl, Dal, Kosidou and Karlsson2018; Khandaker, Zimbron, Dalman, Lewis, & Jones, Reference Khandaker, Zimbron, Dalman, Lewis and Jones2012; Khandaker, Zimbron, Lewis, & Jones, Reference Khandaker, Zimbron, Lewis and Jones2013). Early-life exposure to infection may result in activation of an acute inflammatory response potentially affecting neurodevelopment (Boulanger & Shatz, Reference Boulanger and Shatz2004). Acute inflammation causes reductions in neuroplasticity, an important process for brain functioning, which may lead to neuropsychiatric outcomes later on (De Picker, Morrens, Chance, & Boche, Reference De Picker, Morrens, Chance and Boche2017). Compared with psychosis, longitudinal studies of early-life infection and depression are relatively rare, though inflammation is increasingly thought to play a role in the pathogenesis of depression (Dantzer, O'Connor, Freund, Johnson, & Kelley, Reference Dantzer, O'Connor, Freund, Johnson and Kelley2008; Khandaker, Dantzer, & Jones, Reference Khandaker, Dantzer and Jones2017; Miller, Maletic, & Raison, Reference Miller, Maletic and Raison2009). Most of the existing longitudinal studies of early-life infection have focused on psychosis and depression in adults (Bechter et al., Reference Bechter, Reiber, Herzog, Fuchs, Tumani and Maxeiner2010; Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011; Eisenberger et al., Reference Eisenberger, Berkman, Inagaki, Rameson, Mashal and Irwin2010; Harrison et al., Reference Harrison, Brydon, Walker, Gray, Steptoe and Critchley2009; Laske et al., Reference Laske, Zank, Klein, Stransky, Batra, Buchkremer and Schott2008; Meyer, Schwarz, & Müller, Reference Meyer, Schwarz and Müller2011). Longitudinal studies of depressive and psychotic experiences (PEs) during childhood/adolescence are scarce (Goodwin, Reference Goodwin2011). Childhood/adolescent depressive symptoms are associated with adult depression (Copeland, Shanahan, Costello, & Angold, Reference Copeland, Shanahan, Costello and Angold2009; Dunn & Goodyer, Reference Dunn and Goodyer2006; Fombonne, Wostear, Cooper, Harrington, & Rutter, Reference Fombonne, Wostear, Cooper, Harrington and Rutter2001; Lewinsohn, Rohde, Seeley, Klein, & Gotlib, Reference Lewinsohn, Rohde, Seeley, Klein and Gotlib2000). PEs during childhood/early-adolescence may be part of typical development and are transient for most individuals. However, population-based longitudinal studies suggest association between early-life PEs with the risk of psychotic disorders subsequently in adulthood (Poulton et al., Reference Poulton, Caspi, Moffitt, Cannon, Murray and Harrington2000; Zammit et al., Reference Zammit, Kounali, Cannon, David, Gunnell, Heron and Lewis2013), and with risk factors for schizophrenia including impaired neurodevelopment (Cannon et al., Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002; Kelleher & Cannon, Reference Kelleher and Cannon2011). PEs are also associated with adolescent psychiatric multi-morbidity (Kelleher et al., Reference Kelleher, Keeley, Corcoran, Lynch, Fitzpatrick, Devlin and Cannon2012), and with other psychiatric disorders in adulthood (Fisher et al., Reference Fisher, Caspi, Poulton, Meier, Houts, Harrington and Moffitt2013). Therefore, studying adolescent PEs and depressive symptoms may offer insights into the development of adult psychotic and mood disorders.
Some previous longitudinal studies have considered the issue of timing, i.e. whether early-life infections are associated with psychotic outcomes closer to the time of exposure or subsequently after several years (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011; Nielsen, Benros, & Mortensen, Reference Nielsen, Benros and Mortensen2014), but such studies of depression are scarce. Moreover, previous studies have typically examined effects of severe infections (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011, Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013), but studies of common childhood infections and subsequent depressive/psychotic outcomes, or studies of the number of childhood infection and subsequent depression are scarce. Prospective cohort studies with repeated measures of depressive symptoms and PEs over a long period are required to address these issues, but such studies are relatively rare.
Infections are common during childhood, but some children are disproportionately prone to high infection burden possibly due to genetic and environmental factors. While childhood infection severity, duration, and related hospitalization have been linked with depression and psychosis in adulthood (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011, Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013; Nielsen et al., Reference Nielsen, Benros and Mortensen2014), it is unclear whether the degree of childhood infection burden is associated with psychotic/depressive outcomes in childhood/adolescence. A high number of childhood infections may be a risk factor for psychiatric disorders, as a dose-response association between increasing number of childhood infection and adult mood or psychotic disorders has been reported (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011, Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013). Early-life infections could also be a marker for shared genetic and environmental risk factors for infection and major psychiatric disorders (Nielsen, Laursen, & Mortensen, Reference Nielsen, Laursen and Mortensen2013). It is possible that inflammatory immune response during a critical developmental window is detrimental for brain development/function (Cordeiro, Tsimis, & Burd, Reference Cordeiro, Tsimis and Burd2015). A high burden of common childhood infections may also reflect underlying familial factors predisposing to infection, such as socioeconomic status, living conditions, and genetic factors (Nielsen et al., Reference Nielsen, Laursen and Mortensen2013). Therefore, studies investigating the effects of infection burden are required.
Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) (Boyd et al., Reference Boyd, Golding, Macleod, Lawlor, Fraser, Henderson and Davey Smith2013), a general population-representative prospective British birth cohort, we have investigated the longitudinal associations of childhood infections from age 1.5 to 7.5 years old with depressive symptoms measured six times from age 10 to 19 years and with PEs at age 12 and 18 years. We examined not only the number of childhood infections as exposure, but also the effect of infection burden, grouped as low, medium, high or very high. We hypothesized that a higher overall number of infections and very high infection burden would be associated with higher risks for depressive symptoms and PEs subsequently up to early-adulthood.
Methods
Description of cohort and sample
ALSPAC is a general population-based birth cohort in the former Avon County in the South West region of England. Initially, 14 541 pregnant women resident in the study catchment areas and with expected delivery dates between 1 April 1991 and 31 December 1992 were recruited into the cohort. Detailed information about the ALSPAC cohort can be found on the study website (http://www.bristol.ac.uk/alspac), and the sample characteristics and methodology have been previously described (Boyd et al., Reference Boyd, Golding, Macleod, Lawlor, Fraser, Henderson and Davey Smith2013; Fraser et al., Reference Fraser, Macdonald-Wallis, Tilling, Boyd, Golding, Davey Smith and Lawlor2013). For information on all available ALSPAC data, a fully searchable data dictionary is also available (http://www.bris.ac.uk/alspac/researchers/our-data).
The risk set for this study comprised 11 786 individuals with data on infections during childhood. Out of the risk set, the number of individuals with depressive symptoms data decreased from 6685 at age 10 to 3101 at age 19 (online Supplementary Fig. 1). From the risk set, PEs data were available for 6176 individuals at age 12 and for 4253 individuals at age 18 (online Supplementary Fig. 2). These samples formed the basis for the analyses presented.
Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. Written informed consent was provided by all participants. No financial compensation was given.
Exposure measures
Childhood infections from age 1.5 to age 7.5 years
Approximately once per year when the child was aged between 1.5 and 7.5 years old, caregivers completed seven postal questionnaires about common childhood infections experienced by their child. These included German measles, measles, chickenpox, mumps, meningitis, cold sores, whooping cough, urinary infections, eye infections, ear infections, chest infections, tonsillitis/laryngitis, scarlet fever, influenza, cold and ‘other’.
Overall, 11 786 children had fully/partly completed infection questionnaires from the seven time-points. In the questionnaire about childhood infections, caregivers could tick the ‘Yes’ or ‘No’ boxes for infections that their child has had. Some people simply ticked the ‘Yes’ box to indicate that their child has had certain infections, but did not tick the ‘No’ box for infections their child did not have. In those instances, we coded the un-ticked boxes as ‘No’ that year rather than dropping the subject from analysis for missing data. However, if a parent had all boxes unticked, they were coded as missing, as we did not have any indication whether their child had any infection that year.
Childhood infection count was used as a standardized continuous variable. Infection was also used as a categorical variable representing the degree of infection burden: low (50th percentile and below; 0–4 infections), medium (51–75th percentile; 5–6 infections), high (76–90th percentile; 7–9 infections), and very high burden (above 90th percentile; 10–22 infections). These categories were based on a prior study and were chosen to capture the positive skew of childhood infection distribution (Mackinnon, Zammit, Lewis, Jones, & Khandaker, Reference Mackinnon, Zammit, Lewis, Jones and Khandaker2018).
Outcome measures
Depressive symptoms at age 10, 13, 14, 17, 18, and 19 years
Depressive symptoms were self-reported by the child/young person using the Short Mood and Feelings Questionnaire (SMFQ) (Sharp, Goodyer, & Croudace, Reference Sharp, Goodyer and Croudace2006). Depressive symptoms were measured at age 10 (mean age in years = 10.6; standard deviation (s.d.) = 0.3), 13 (mean = 12.8; s.d. = 0.2), 14 (mean = 13.8; s.d. = 0.2), 17 (mean = 16.7; s.d. = 0.2), 18 (mean = 17.8; s.d. = 0.4), and 19 years (mean = 18.6; s.d. = 0.5). The SMFQ is a widely used, age-appropriate, and validated tool comprising 13 items that cover core symptoms of depression and anxiety experienced in the past 2 weeks. Each item is scored zero (not true), one (sometimes true) or two (true) giving a total score of 0–26. Depressive symptoms were used as a continuous variable.
PEs at age 12 and 18 years
PEs were identified using the face-to-face, semi-structured Psychosis-Like Symptom Interview (PLIKSi) (Horwood et al., Reference Horwood, Salvi, Thomas, Duffy, Gunnell, Hollis and Harrison2008; Zammit et al., Reference Zammit, Kounali, Cannon, David, Gunnell, Heron and Lewis2013) conducted by trained psychology graduates in assessment clinics. PEs were coded according to the definitions and rating rules for the Schedules for Clinical Assessment in Neuropsychiatry, Version 2.0. (World Health Organisation, 1994). PLIKSi at age 12 has ‘fair’ inter-rater (kappa = 0.75) and test-retest (kappa = 0.48) reliability (Horwood et al., Reference Horwood, Salvi, Thomas, Duffy, Gunnell, Hollis and Harrison2008). PLIKSi at age 18 has ‘good’ inter-rater (kappa = 0.83) and test-retest (kappa = 0.76) reliability (Zammit et al., Reference Zammit, Kounali, Cannon, David, Gunnell, Heron and Lewis2013).
PLIKSi covers the three main domains of positive PEs: hallucinations (visual and auditory), delusions (spied on, persecuted, thoughts read, reference, control, grandiosity, and other), and thought interference (insertion, withdrawal, and broadcasting). After cross-questioning, interviewers rated PEs as ‘not present’, ‘suspected’, or ‘definitely present’. Uncertain responses were always ‘rated down’, and symptoms were rated as definite only if a clear example could be provided. For suspected or definite PEs, interviewers also recorded the frequency, affect, effects on social/educational/occupational function, help-seeking, and attributions including fever, hypnopompic/hypnogogic state, or illicit drugs.
PEs measured at age 12 refer to experiences from the previous 6 months while PEs measured at age 18 refer to experiences since age 12. In line with previous papers originally describing PEs in the ALSPAC birth cohort, we used three increasingly strict definitions for this outcome, i.e. any PEs (suspected or definite PEs); definite PEs; and definite PEs without attribution. The number of participants meeting each of the outcome definitions are different (online Supplementary Fig. 2) and the risk of each outcome was analysed separately. The comparison group for each outcome included all the individuals who did not meet that specific definition for the outcome (Khandaker, Zammit, Lewis, & Jones, Reference Khandaker, Zammit, Lewis and Jones2014).
Confounders
Based on previous studies, we included sex, ethnicity, birth weight, maternal social status, and parental history of severe depression or schizophrenia as these are associated with exposure and/or outcome and so could be confounders (Abel et al., Reference Abel, Wicks, Susser, Dalman, Pedersen, Mortensen and Webb2010; Benros et al., Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013; Khandaker et al., Reference Khandaker, Zimbron, Dalman, Lewis and Jones2012; Khandaker, Stochl, Zammit, Lewis, & Jones, Reference Khandaker, Stochl, Zammit, Lewis and Jones2014). Sex (binary) and birth weight (grams, continuous) were assessed at birth. Ethnicity was recorded as White, Black African, Black Caribbean, Black Other, Bangladeshi, Chinese, Indian, Pakistani, and Other. We re-coded this variable as White and Other due to low counts for non-White ethnic groups. Maternal social status was originally documented using Office of National Statistics categories (Office for National Statistics, 2016) and re-coded as non-manual (I, II and IIIa) and manual (IIIb, IV and V). The armed forces were excluded as the sample size for this category was too small (n = 4). Mothers and their partners completed separate questionnaires at 12 weeks gestation in which they self-reported having severe depression or schizophrenia, which were used as binary variables. See Supplementary Material for how these binary variables were defined.
Statistical analysis
All analyses were carried out using R version 3.6.1. Regression analyses were performed before and after adjusting for potential confounders. All main analyses were based on the dataset after imputation of missing values for confounders.
Imputation of missing confounder variables
Regression analyses testing associations between exposure and outcomes were conducted after imputation of missing data for confounders (ethnicity, maternal social status, birth weight, parental schizophrenia, and parental severe depression) to increase the sample size. Sex had no missing data. The percentage of missing values across the relevant confounder variables varied between 1.5% and 35.0% (online Supplementary Table 1). We used the TestMCARNormality function to test whether data were missing completely at random (MCAR) (Jamshidian, Jalal, & Jansen, Reference Jamshidian, Jalal and Jansen2014). The hypothesis of MCAR was rejected at the 0.05 significance level. Using the missing_compare function, each variable returned significance (p < 0.05) with at least one other variable, suggesting that the data met the missing at random (MAR) assumption.
We used multiple imputations using fully conditional Markov chain Monte Carlo method for the above confounder variables plus auxiliary variables that were indicators of missingness. Exposure and outcome variables were also included. The selected auxiliary variables included: housing/living conditions, parental education/employment status, financial difficulties, life events, and maternal characteristics (age, body mass index, marital status, anxiety/post-natal depression, and smoking status).
We used the R package mice version 3.0 to create and analyse the multiply imputed datasets (van Buuren & Groothuis-Oudshoorn, Reference van Buuren and Groothuis-Oudshoorn2011). Since missing data were present in 19.9% of subjects, we used 20 imputations as recommended (White, Royston, & Wood, Reference White, Royston and Wood2011). The parameters of interest were estimated in each dataset separately and combined using Rubin's rules.
Associations of childhood infection with depressive symptoms and PEs
Linear regression was used to examine the association between the number of childhood infections/infection burden and depressive symptoms. Low infection burden (0–4 infections) was used as the reference group where infection burden was used as the exposure. Regression models were adjusted for sex, birth weight, maternal social status, ethnicity, and parental history of severe depression. Beta estimates and standard error (s.e.) from regression models are presented.
Logistic regression was used to calculate odds ratios (OR) and 95% confidence intervals (CI) for PEs at age 12 or age 18 associated with the number of childhood infections/infection burden. The low burden of infection (0–4 infections) was used as the reference group for the latter. Regression models were adjusted for sex, birth weight, maternal social status, ethnicity, and parental history of schizophrenia.
For all analyses, we used the maximum available sample for each outcome measure to increase statistical power. Holm-Bonferroni p value correction was performed to correct for multiple testing (online Supplementary Table 2) (Holm, Reference Holm1979).
Sensitivity analysis exploring impact of missing data
To explore the potential impact of missing data we repeated our main analyses based on the complete case-set, defined as participants with no missing data for exposure, outcome or confounder measures. We then carried out a number of comparisons between different samples. (1) We compared the risk set (i.e. participants with data on childhood infection) with the complete case-set for depressive symptoms (i.e. data on childhood infection, all confounders, and depressive symptoms at all follow-ups are available). We did the same comparison for PEs. (2) We compared the analytic sample for depressive symptoms at age 19 years (i.e. data on both childhood infection and depressive symptoms at age 19 available) with the missing sample (i.e. childhood infection data present but depressive symptoms data at age 19 missing). We performed the same comparison for PEs at age 18 separately.
Results
Baseline characteristics of the sample
The risk set (N = 11 786) was predominantly of white ethnicity (97.7%) (Table 1). The average number of childhood infections in the total sample was 4.6 (s.d. = 3.2), and the median was 4 (inter-quartile range = 2–6). The number of children with very high infection burden (⩾90th percentile; 10–22 infections) was 947 (8.0%).
There was a general increase in depressive symptoms between age 10 and 19 years; mean SMFQ score at age 10 was 4.0 (s.d. = 3.5), and was 6.8 (s.d. = 5.9) at age 19 (Table 1). The percentage of participants with definite PEs was similar at age 12 (5.7%) and at age 18 (4.8%).
Association between number of childhood infections and depressive symptoms
Number of childhood infections from age 1.5 to age 7.5 years was associated with subsequent depressive symptoms at age 10 (N = 6685; beta = 0.14, s.e. = 0.04; p = <0.01), 13 (N = 6087; beta = 0.24; s.e. = 0.05; p = <0.001), and 14 years (N = 5504; beta = 0.23, s.e. = 0.06; p = <0.001) (Table 2). Evidence for these associations remained after adjusting for potential confounders. Childhood infections were not associated with depressive symptoms at age 17, 18, or 19 years.
a Adjusted for sex, birth weight, maternal social status, ethnicity, and parental history of severe depression.
After correction for multiple testing, evidence remained for an association between number of childhood infections and depressive symptoms at age 10 (p = 0.02), 13 (p = <0.001) and 14 years (p = <0.01) (online Supplementary Table 2).
Association between number of childhood infections and PEs
Number of childhood infections was associated with subsequent PEs at age 12 years (N = 6176), including suspected/definite PEs (OR = 1.17, 95% CI = 1.09–1.26), definite PEs (OR = 1.16, 95% CI = 1.04–1.29), and definite PEs without attribution (OR = 1.17, 95% CI = 1.04–1.31) (Table 3). Evidence for these associations remained after adjusting for potential confounders. The number of childhood infections was not associated with PEs at age 18 years.
a Adjusted for sex, birth weight, maternal social status, ethnicity, and parental history of schizophrenia.
After correction for multiple testing, evidence remained for an association between number of childhood infections and suspected/definite PEs at age 12 (p = <0.001), but not with definite PEs (p = 0.12) or definite PEs without attribution at age 12 years (p = 0.12) (online Supplementary Table 2).
Association between childhood infection burden and depressive symptoms
Compared with low infection burden, very high infection burden was associated with depressive symptoms at age 10 (beta = 0.38; s.e. = 0.16; p = 0.01), 13 (beta = 0.76; s.e. = 0.17; p = <0.001), 14 (beta = 0.66; s.e. = 0.21; p = <0.01), and 17 years (beta = 0.68; s.e. = 0.30; p = 0.02) (Table 4). Evidence for these associations remained after adjusting for potential confounders. Very high infection burden was not associated with depressive symptoms at age 18 (beta = 0.33; s.e. = 0.29; p = 0.26) or 19 years (beta = 0.41; s.e. = 0.38; p = 0.27).
a Adjusted for sex, birth weight, maternal social status, ethnicity, and parental history of severe depression.
After correction for multiple testing, evidence remained for the association between very high infection burden and depressive symptoms at age 13 (p = <0.001) and 14 years (p = 0.02) (online Supplementary Table 2). Please see Table 4 for risk associated with other categories of infection burden.
Association between childhood infection burden and PEs
Compared with low infection burden, very high infection burden was associated with suspected/definite PEs at age 12 years (OR = 1.59, 95% CI = 1.24–2.04). Evidence for this association remained after adjusting for potential confounders. Very high infection burden was not associated with PEs at age 18 (Table 5).
a Adjusted for sex, birth weight, maternal social status, ethnicity and parental history of schizophrenia.
After correction for multiple testing, evidence remained for an association between very high infection burden and suspected/definite PEs at age 12 years (p = <0.01) (online Supplementary Table 2). Please see Table 5 for risk associated with other categories of infection burden.
Results based on complete case-set analyses and comparisons between analytic and missing samples
Results from the complete case-set for depressive symptoms (N = 1133) and that for PEs (N = 2495) were very similar to the main results reported above (online Supplementary Tables 3–6). See online Supplementary Tables 7 and 8 for comparison of the risk set (N = 11 786) with the complete case-set for depressive symptoms or PEs, respectively.
Comparison between the analytic sample for depressive symptoms (N = 3101) and the missing sample (N = 8685) showed that those included in analyses were more likely to be female (p = <0.001), white (p = 0.05), have a higher number of childhood infections (p = <0.001), and have lower depressive symptoms at age 10 and 18 years (p = <0.001) (online Supplementary Table 9). The analytic sample for depressive symptoms was less likely to have a mother of manual occupational status (p < 0.001) or have a parent with a history of severe depression (p < 0.001).
The analytic sample for PEs (N = 4253), compared with the missing sample (N = 7533), was more likely to be female (p = <0.001), not have a mother of manual occupational status (p = <0.001), and have a higher number of childhood infections (p = <0.001) (online Supplementary Table 10).
Discussion
Our findings suggest that common early-childhood infections, particularly a very high infection burden, are associated with (i) the risk of depressive symptoms subsequently up to the age of 17, and (ii) the risk of suspected/definite PEs subsequently at age 12. The associations of infection with depressive symptoms at age 13 and 14 years and with suspected/definite PEs at age 12 were robust and persisted after correction for multiple testing. Childhood infections were not associated with risk of depressive symptoms or PEs at age 18/19 years.
Strong links exist between child/adolescent and adult depression (Copeland et al., Reference Copeland, Shanahan, Costello and Angold2009; Dunn & Goodyer, Reference Dunn and Goodyer2006; Fombonne et al., Reference Fombonne, Wostear, Cooper, Harrington and Rutter2001; Lewinsohn et al., Reference Lewinsohn, Rohde, Seeley, Klein and Gotlib2000). Adolescents with recurrent depressive episodes are at particularly high risk of subsequent recurrent depression as adults (Lewinsohn et al., Reference Lewinsohn, Rohde, Seeley, Klein and Gotlib2000). Longitudinal studies suggest that recurrence of depression may be similar in clinic and community samples (Dunn & Goodyer, Reference Dunn and Goodyer2006). PEs in childhood or adolescence may also provide a valid method for studying the development of adult psychotic disorders (Kelleher & Cannon, Reference Kelleher and Cannon2011; Murray & Jones, Reference Murray and Jones2012). PEs in childhood may be transient but population-based studies suggest that PEs in the general population and those observed in psychotic disorders may exist on a continuum (van Os, Linscott, Myin-Germeys, Delespaul, & Krabbendam, Reference van Os, Linscott, Myin-Germeys, Delespaul and Krabbendam2009). Prospective birth cohort studies have reported that PEs in childhood are associated with an increased risk of psychotic disorders in adulthood (Zammit et al., Reference Zammit, Kounali, Cannon, David, Gunnell, Heron and Lewis2013). Common underlying mechanisms for PEs in healthy individuals and in schizophrenia have also been reported (Howes et al., Reference Howes, Shotbolt, Bloomfield, Daalman, Demjaha, Diederen and Sommer2013). Early-life infections could increase psychosis risk by affecting neurodevelopment, consistent with the neurodevelopmental hypothesis of schizophrenia (Murray & Lewis, Reference Murray and Lewis1987; Weinberger, Reference Weinberger1987). In many cases, PEs in our sample are likely to be part of normal development while for other participants these symptoms may be more pathological (De Loore et al., Reference De Loore, Gunther, Drukker, Feron, Sabbe, Deboutte and Myin-Germeys2011; Rubio, Sanjuán, Flórez-Salamanca, & Cuesta, Reference Rubio, Sanjuán, Flórez-Salamanca and Cuesta2012). Childhood PEs are also familial and heritable and associated with multiple risk factors for schizophrenia (Khandaker, Zammit, et al., Reference Khandaker, Zammit, Lewis and Jones2014; Polanczyk et al., Reference Polanczyk, Moffitt, Arseneault, Cannon, Ambler, Keefe and Caspi2010; Thomas et al., Reference Thomas, Harrison, Zammit, Lewis, Horwood, Heron and Gunnell2009; Zammit et al., Reference Zammit, Odd, Horwood, Thompson, Thomas, Menezes and Harrison2009). PEs and depressive symptoms in children and adolescents appear to be markers of mental distress and represent useful indicators for subsequent mental health problems.
In our sample, associations observed between childhood infections and PEs/depressive symptoms in childhood and adolescence did not persist for these outcomes in early-adulthood at age 18/19 years. There could be a number of explanations for this, including that childhood infections do not have a lasting effect on the risk for adult mental health disorders. However, findings from other long-term follow-up studies argue against this. For instance, a meta-analysis of childhood infections reported that children exposed to viral infections of the CNS were at higher risk of adult psychosis (Khandaker et al., Reference Khandaker, Zimbron, Dalman, Lewis and Jones2012). Longitudinal studies of childhood infection and subsequent non-affective psychosis found similar results (Dalman et al., Reference Dalman, Allebeck, Gunnell, Harrison, Kristensson, Lewis and Karlsson2008; Khandaker et al., Reference Khandaker, Dalman, Kappelmann, Stochl, Dal, Kosidou and Karlsson2018). Furthermore, population-based longitudinal studies suggest that childhood infections and autoimmune disease increase risk for adult schizophrenia and mood disorders in a dose-response fashion (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011, Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013). Benros and colleagues reported that a history of hospitalization for infection increases the risk of mood disorders and schizophrenia by 62% and 60%, respectively (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011, Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013). One explanation for the null findings could be attrition. At age 19 years, 86% of the sample with data on childhood infections were missing from follow-up for depressive symptoms and at age 18 years, 68% of the sample were missing from follow-up for PEs. Another explanation could be the choice of exposure and outcome used. For instance, previous studies used serious infections requiring hospitalization as exposure (Benros et al., Reference Benros, Nielsen, Nordentoft, Eaton, Dalton and Mortensen2011, Reference Benros, Waltoft, Nordentoft, Østergaard, Eaton, Krogh and Mortensen2013), and diagnosis of schizophrenia or depression as an outcome. We have used common childhood infections as exposure and depressive symptoms/PEs as outcomes. Nevertheless, we are still able to show that relatively common childhood infections are associated with risk for depressive symptoms and PEs in adolescence. The sample size for outcomes in early-adulthood was relatively small and there was attrition over time, a common issue for prospective studies. In future, studies with larger sample sizes for cases are required.
It is possible that childhood infections contribute indirectly via inflammatory mechanisms to the risk of mental health disorders. Inflammatory responses to infection, such as elevated cytokine levels and fever, may represent the mechanism through which risk for mental disorders is increased (Flinkkilä, Keski-Rahkonen, Marttunen, & Raevuori, Reference Flinkkilä, Keski-Rahkonen, Marttunen and Raevuori2016). ‘Sickness behaviour’ is common in infection and is present in some cases of depression. Sickness behaviour is triggered by proinflammatory cytokines in response to infectious agents and includes symptoms such as fatigue, anhedonia, concentration difficulties, social withdrawal, and appetite changes (Dantzer, Reference Dantzer2009; Dantzer et al., Reference Dantzer, O'Connor, Freund, Johnson and Kelley2008). Along with sickness behaviour, proinflammatory cytokines may induce depression in physically ill individuals with no history of mental health disorders. Depression may be therefore a maladaptive form of cytokine-induced sickness in some individuals (Dantzer, Reference Dantzer2009; Dantzer et al., Reference Dantzer, O'Connor, Freund, Johnson and Kelley2008). In addition, infections in early-life have been associated with adult schizophrenia via inflammation (Al-Haddad et al., Reference Al-Haddad, Jacobsson, Chabra, Modzelewska, Olson, Bernier and Sengpiel2019; Brown et al., 2000, Reference Brown, Cohen, Harkavy-Friedman, Babulas, Malaspina, Gorman and Susser2001, Reference Brown, Schaefer, Quesenberry, Liu, Babulas and Susser2005; Brown, Begg, et al., Reference Brown, Begg, Gravenstein, Schaefer, Wyatt, Bresnahan and Susser2004; Brown, Deicken, et al., Reference Brown, Deicken, Vinogradov, Kremen, Poole, Penner and Schaefer2009; Brown, Hooton, et al., Reference Brown, Hooton, Schaefer, Zhang, Petkova, Babulas and Susser2004; Brown, Vinogradov, et al., Reference Brown, Vinogradov, Kremen, Poole, Deicken, Penner and Schaefer2009; Buka, Cannon, Torrey, & Yolken, Reference Buka, Cannon, Torrey and Yolken2008; Buka, Tsuang, Torrey, Klebanoff, Bernstein, et al., Reference Buka, Tsuang, Torrey, Klebanoff, Bernstein and Yolken2001; Buka, Tsuang, Torrey, Klebanoff, Wagner, et al., Reference Buka, Tsuang, Torrey, Klebanoff, Wagner and Yolken2001; Khandaker et al., Reference Khandaker, Dalman, Kappelmann, Stochl, Dal, Kosidou and Karlsson2018; Reference Khandaker, Zimbron, Lewis and Jones2013; Mortensen et al., Reference Mortensen, Nørgaard-Pedersen, Waltoft, Sørensen, Hougaard, Torrey and Yolken2007, Reference Mortensen, Pedersen, Hougaard, Nørgaard-Petersen, Mors, Børglum and Yolken2010; Murphy et al., Reference Murphy, Fineberg, Maxwell, Alloy, Zimmermann, Krigbaum and Ellman2017; Zammit et al., Reference Zammit, Allebeck, David, Dalman, Hemmingsson, Lundberg and Lewis2004). Infection-related inflammation may influence neurodevelopment resulting in heightened risk for schizophrenia (Benros, Mortensen, & Eaton, Reference Benros, Mortensen and Eaton2012). Inflammation represents a compelling link between infection and mental health disorders and requires further investigation.
The interaction of multiple genetic and environmental factors is likely to contribute to risk for psychosis or depression. Genetic susceptibility plays an important role in the risk of mental health disorders, but there is evidence to suggest that a significant proportion of cases may be preventable through modification of the environment (Benros, Eaton, & Mortensen, Reference Benros, Eaton and Mortensen2014; Uher, Reference Uher2014). We attempted to explore the effects of environmental factors by controlling for maternal social status and birth weight. Evidence for the association between infection and PEs/depressive symptoms attenuated but did not disappear, suggesting that these environmental factors partly explain these associations. Some genetic and environmental risk factors for psychosis and depression may be shared (Uher, Reference Uher2014). Childhood maltreatment, social disadvantage, and minority status are independently associated with both psychosis and depression (Uher, Reference Uher2014). By recognizing infection and other environmental factors that are associated with depression and psychosis, we may be able to identify at-risk individuals early and prevent the onset of symptoms.
The strengths of this study include longitudinal design and repeated measures of both depressive symptoms and PEs. A limitation is the method for collection of childhood infection data. The questionnaire containing questions on infections were completed by primary caregiver throughout childhood with a short period of recall (past 12 months), thus minimizing, though not completely eliminating, the risk of erroneous recall. The questions on infections also require parents to have a good understanding of infectious diseases and may be vulnerable to misreporting. For example, 8.2% of parents reported that their child had no infections during childhood; a somewhat unlikely scenario. Another limitation is attrition; the number of completed mental health assessments decreases over time. This may be indicative of selective attrition of mental health cases since mental health problems have been associated with non-response and attrition (Dupuis, Strippoli, Gholam-Rezaee, Preisig, & Vandeleur, Reference Dupuis, Strippoli, Gholam-Rezaee, Preisig and Vandeleur2019). Attrition and subsequent smaller sample size could result in underestimation of the true effect of infection on mental health outcomes. Another possible explanation for the lack of association with PEs/depressive symptoms in early-adulthood could be that the relative contribution of infection to mental health becomes negligible over time due to neuroplasticity and brain development. Finally, statistical power was an issue in some of our PE analyses, resulting in wide CI.
Childhood infections are inevitable, but the adverse outcomes of such physical health problems may go beyond physical health later in life. Here we present evidence that childhood infections, particularly a very high infection burden, can negatively impact mental health well into adolescence. Future work is needed (1) to replicate the observed associations between common childhood infections and mental health outcomes during adolescence; (2) to examine whether common childhood infections have an effect on depressive symptoms and PEs in adulthood; and (3) to elucidate potential mechanisms for these associations.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291720004080.
Acknowledgements
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. The UK Medical Research Council and Wellcome (Grant ref: 102215/2/13/2) and the University of Bristol provide core support for ALSPAC. The publication is the work of the authors who will serve as guarantors for the contents of this paper. A comprehensive list of grants funding is available on the ALSPAC website (http://www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf). This report of research supported in part by the National Institute for Health Research (NIHR) Applied Research Collaboration (ARC) East of England expresses the views of the authors and not necessarily those of the NHS, the NIHR, the Department of Health and Social Care or other funding bodies. ABC is supported by the National Institute of Health Research CLAHRC RCF (MRR73-1795-00000), NIHR ARC East of England, the MQ: Transforming Mental Health (Data Science Award; grant code: MQDS17/40) and Wolfson College. GMK acknowledges funding support from the Wellcome Trust (Intermediate Clinical Fellowship; grant code: 201486/Z/16/Z), MQ as above, and the Medical Research Council (MICA: Mental Health Data Pathfinder; grant code: MC_PC_17213). PBJ acknowledges funding from MQ and the MRC, as above, NIHR PGfAR 0616-20003, and from the NIHR ARC East of England.
Conflict of interest
None.