عدم حضور پدر و زمان شروع قاعدگی در دختران نوجوان از یک گروه در بریتانیا: نقش واسطه افسردگی مادران و مشکلات عمده مالی

Abstract In a prospective birth cohort study of 5295 girls from the UK-based Avon Longitudinal Study of Parents and Children (ALSPAC), we examined the association between biological father absence in childhood and age at menarche whilst adjusting for antenatal indicators of socioeconomic disadvantage and maternal characteristics. We also examined whether exposure to maternal depression and financial problems during middle childhood mediate the association between father absence and age at menarche. There was stronger evidence for an association between father absence during the first 5 years of life and early timing of menarche compared with father absence between 5 and 10 years. There was evidence that maternal depression and major financial problems explained some of the association between early childhood father absence and age at menarche. Although father absence cannot be a direct target of prevention, family-based programs to address family processes influenced by maternal depression and socioeconomic disadvantage may be effective.

Introduction The pubertal transition is a critical developmental period associated with profound biological and psychosocial changes. Early timing of puberty in girls has been linked to adverse psychosocial and health outcomes (Mendle et al., 2006) including depressive symptoms (Joinson, Heron, Araya, & Lewis, 2013), poor body image and eating disorders (Striegel-Moore et al., 2001), and substance abuse (Patton et al., 2004). Age at onset of menarche is often used as an indicator of pubertal timing because it is a salient milestone that defines the female pubertal transition and signals reproductive fertility (Posner, 2006). Timing of menarche is influenced not only by genetic factors, but also by a range of environmental factors including family stressors (Graberc, Brooksgunn, & Warren, 1995). Absence of the biological father is one such stressor that has been linked to earlier timing of menarche (Ellis, 2004), particularly if it occurred during the first 5 years of life (Surbey, 1990 and Wierson et al., 1993). Life-course adversity theories view father absence as one of a range of stressors (e.g., socioeconomic disadvantage, parental conflict, negative parenting) linked to earlier onset of sexual activity and reproduction (e.g., Belsky, Steinberg, & Draper, 1991), whilst paternal investment theory (Draper and Harpending, 1982 and Draper and Harpending, 1988) emphasises the critical role of father absence during early childhood as influencing female reproductive strategies. Ellis (2004) suggested that absence/presence of the biological father and general effects of psychosocial stressors are each uniquely involved in timing of pubertal development through different neurobiological mechanisms. Previous research has provided valuable evidence with regard to the link between father absence and timing of menarche, however, there are limitations of these studies that need to be addressed in order to advance understanding. The majority are based on cross-sectional samples (Bogaert, 2008, Hoier, 2003, Jorm et al., 2004, Romans et al., 2003 and Quinlan, 2003). Of the previous longitudinal studies, some used relatively small samples (Belsky et al., 2007, Ellis and Garber, 2000, Tither and Ellis, 2008 and Wierson et al., 1993), and some relied on retrospective recall of both timing of menarche (Mendle et al., 2006; Moffitt, Caspi, Belsky, & Silva, 1992; Tither & Ellis, 2008) and age of child at which biological father departed (Graberc et al., 1995 and Mendle et al., 2006). Retrospective self-reports of age at menarche are open to recall bias, particularly in older samples where more time has elapsed since the onset of menarche. Some of the earlier studies did not differentiate between early (<5 years) and later departure of the biological father from the family (Bogaert, 2008 and Campbell and Udry, 1995; Moffitt et al., 1992; Romans et al., 2003 and Wierson et al., 1993). This could have masked important differential effects since there is evidence that early, compared with later father absence, is more strongly associated with subsequent pubertal timing (Ellis and Garber, 2000, Quinlan, 2003 and Surbey, 1990). Many previous studies did not adjust for a range of confounders that are associated with both earlier timing of menarche and father absence, including socioeconomic disadvantage (Ellis and Essex, 2007, Hulanicka et al., 2001 and James-Todd et al., 2010), maternal educational attainment (Tither & Ellis, 2008), maternal depression (Ellis & Garber, 2000), and maternal age at menarche (Wierson et al., 1993). Higher levels of maternal education have been reported to be associated with intactness of the biological family and daughters' later age at menarche (Bogaert, 2008 and Tither and Ellis, 2008). Socioeconomic disadvantage and maternal depression have been linked with family breakdown (Meadows et al., 2008 and White and Rogers, 2000) and earlier menarche (Ellis and Essex, 2007 and Ellis and Garber, 2000). There is also evidence to support a strong association between earlier age at menarche of mothers and their daughters (Campbell and Udry, 1995 and Rowe, 2002). Mothers who experience earlier menarche not only have daughters who mature at an earlier age (Ge, Natsuaki, Neiderhiser & Reiss, 2007), they also tend to engage in earlier sexual activity and childbirth (Deardorff, Gonzales, & Christopher, 2005), and have relationships that are more likely to end in dissolution (Moffitt et al., 1992 and Surbey, 1990). Although the association between father absence and earlier age at menarche has been previously reported, mechanisms underlying this association have not been extensively investigated (James, Ellis, Schlomer, & Garber, 2012). It has been previously argued that the effects of father absence on age at menarche may be explained by psychosocial and contextual stressors arising as a result of father departure, such as socioeconomic disadvantage (e.g., decrease in family income, major financial problems; Ellis, Fayden-Ketchum, Dodge, Petit, & Bates, 1999) and maternal depression (Ellis & Garber, 2000; for conceptual representation of the hypothesised mechanisms see Fig. 1). In particular, numerous studies have found that absence of the biological father from the household is associated with financial problems and decrease in economic well-being, particularly among women (Everett & Volgy, 1991; Sigle-Rushton & McLanahan, 2004). Similarly, a substantial body of research has examined the association between indices of childhood socioeconomic status (e.g., financial hardship) and pubertal timing (Arim et al., 2011 and Jean et al., 2011), with recent epidemiologic evidence linking lower household income to earlier timing of menarche (Braithwaite et al., 2009). Family dissolution and single parenthood have also been linked to poor mental health and depression in mothers (Cairney et al., 2003 and Wade and David, 2004), which have been found to predict earlier age at menarche in daughters (Ellis & Graber, 2000). Some of the effects of father absence on age at menarche may, therefore, be mediated through exposure to financial problems and maternal depression. To our knowledge, however, no study has examined possible mediating effects of these factors in the association between father absence and timing of menarche in a multiple mediation model. Hypothesised model: pathways from early father absence to earlier age at ... Fig. 1. Hypothesised model: pathways from early father absence to earlier age at menarche in daughters. Figure options The aims of the current study, based on a large UK birth cohort, are twofold. First, we examine the association between father absence and timing of menarche. We address limitations of previous research by using a prospective design, large sample size, frequent contemporaneous reports of age at menarche, repeated measures of presence or absence of the biological father from birth through to late childhood (10 years), and by adjusting for a range of indicators of socioeconomic disadvantage and maternal characteristics assessed during the antenatal period. Second, we examine whether exposure to maternal depression and major financial problems in middle childhood mediate the association between father absence and timing of menarche.

Results Distribution of confounders in father-present and father-absent families and by timing of menarche There was a strong association between socioeconomic disadvantage and father absence across the range of indicators (Table 1). Specifically, early father absence (during the first 5 years) was associated with the highest rates of manual social class, living in privately rented housing, lack of car access, major financial problems, low maternal educational attainment, and larger family size. Early father absence was also associated with the highest rate of early parenthood, maternal antenatal depression but not with early menarche in mothers. Similarly, early timing of menarche was more common whose mothers were of lower social class, lower educational attainment, who reported major financial problems, did not own their home, did not have access to a car, and had a larger family size (Table 2). Earlier age at menarche was also associated with maternal antenatal depression, but not early parenthood (although there was a trend towards earlier menarche among those in earlier category). As expected, mothers who had early menarche were more likely to have early maturing daughters. Table 1. Distribution of indicators of socioeconomic background and maternal characteristics in father-present and father-absent samples. Risk factor Sample size n Father present n (%) Father absence at age ≥5 years n (%) Father absence at age <5 years n (%) Chi2 p Social group Non-manual 4695 2030 (55.97) 170 (52.63) 310 (41.61) 51.31 <0.001 Manual 1597 (44.03) 153 (47.37) 435 (58.39) Homeownership status Owned 4986 3189 (84.41) 253 (75.30) 436 (50.00) 486.62 <0.001 Private rented 589 (15.59) 83 (24.70) 436 (50.00) Car access Yes 4984 3605 (95.42) 304 (91.02) 648 (74.31) 403.08 <0.001 No 173 (4.58) 30 (8.98) 224 (25.69) Major financial problems No 4594 3153 (89.83) 261 (84.19) 601 (77.65) 88.51 <0.001 Yes 357 (10.17) 49 (15.81) 173 (22.35) Mother's educational qualifications ≥High school 4958 2908 (77.09) 255 (76.12) 523 (61.46) 89.61 <0.001 No qualifications 864 (22.91) 80 (23.88) 328 (38.54) Family size ≥3 4958 3585 (95.52) 325 (95.87) 797 (92.03) 18.50 <0.001 3+ 168 (4.48) 14 (4.13) 69 (7.97) Early parenthood No 5148 3825 (98.71) 339 (97.41) 810 (87.57) 284.60 <0.001 Yes 50 (1.29) 9 (2.59) 115 (12.43) Maternal antenatal depression No 4586 2916 (83.39) 226 (71.75) 489 (63.18) 168.28 <0.001 Yes 581 (16.61) 89 (28.25) 285 (36.83) Mother's age at menarche Normative/late (12–15 years) 4466 2733 (81.07) 256 (81.27) 616 (78.97) 1.86 0.395 Early (8–11 years) 638 (18.93) 59 (18.73) 164 (21.03) Note: sample sizes vary because of differences in data availability for indicators of socioeconomic disadvantage and maternal characteristics. Table options Table 2. Means and standard deviations for age at menarche by family's socioeconomic background and maternal characteristics. Risk factor Age at menarche (years) Sample size n Mean S.D. t-statistic p Social group Non-manual 3600 12.70 1.16 3.32 0.001 Manual 12.57 1.15 Homeownership status Owned 3793 12.67 1.14 4.30 <0.001 Private rented 12.46 1.25 Car access Yes 3786 12.64 1.16 2.17 0.030 No 12.47 1.25 Major financial problems No 3524 12.66 1.16 −3.34 0.001 Yes 12.45 1.18 Mother's educational qualifications ≥High school 3776 12.67 1.16 3.87 <0.001 No qualifications 12.50 1.15 Family size ≥3 3782 12.64 1.16 2.34 0.019 3+ 12.42 1.20 Early parenthood No 3937 12.63 1.16 1.49 0.137 Yes 12.50 1.26 Maternal antenatal depression No 3505 12.66 1.15 3.02 0.002 Yes 12.51 1.21 Mother's age at menarche Normative/Late (12–15 years) 3411 12.74 1.13 13.34 <0.001 Early (8–11 years) 12.09 1.14 Note: sample sizes vary because of differences in data availability for indicators of socioeconomic disadvantage and maternal characteristics. Table options Main effects model: association between father absence and timing of menarche The mean age at menarche was 12 years and 6 months (M = 12.5 years, SD = 1.14) in girls from families where the biological father was present; 12 years and 6 months (M = 12.5 years, SD = 1.16) in girls whose biological father left when the study child was between 5 and 10 years; and 12 years and 4 months (M = 12.3 years, SD = 1.23) in girls whose biological father left during the first 5 years of life [F(2.3782)=10.31, p < 0.001]. The results from the unadjusted model show that girls whose father left during the first five years of life experienced earlier age at menarche compared with girls whose biological father was present (Table 3). There was a moderate attenuation in the regression coefficients for the association between early childhood father absence and early timing of menarche following adjustment for the antenatal indicators of socioeconomic disadvantage, which was followed by a subsequent increase in the coefficients after adjusting for maternal characteristics. There was evidence for a differential effect of father absence before and after the age of 5 years on timing of menarche, with little evidence for an association between father absence between 5 and 10 years and earlier age at menarche compared with father-present group. The analyses conducted on the imputed data substantially supported these findings. Table 3. Regression analysis for association between father absence and timing of menarche adjusted for antenatal indicators of family socioeconomic background and maternal characteristics in complete cased (n = 2750) and imputed samples (n = 4990). Timing of menarche (years) Father absence (reference = father present) b SE t 95% CI p Complete case: unadjusted model Father absence at age ≥5 years −0.130 0.085 −1.54 −0.296, 0.036 0.125 Father absence at age <5 years −0.274 0.067 −4.11 −0.404, −0.143 <0.001 Complete case: adjusted 1 (socioeconomic disadvantage) a Father absence at age ≥5 years −0.113 0.085 −1.33 −0.279, 0.053 0.182 Father absence at age <5 years −0.222 0.069 −3.21 −0.357, −0.086 0.001 Complete case: adjusted 2 (maternal characteristics)b Father absence at age ≥5 years −0.106 0.082 1.29 −0.267, 0.055 0.199 Father absence at age <5 years −0.237 0.068 −3.51 −0.370, −0.105 <0.001 Imputed data analysisc Father absence at age ≥5 years −0.065 0.072 −0.90 −0.207, 0.076 0.366 Father absence at age <5 years −0.191 0.056 −3.43 −0.301, −0.082 0.001 a Adjusted 1: adjusted for indices of socioeconomic background: socioeconomic status, home ownership status, car access, major financial problems, mother's educational attainment and family size. b Adjusted 2: further adjusted for maternal characteristics: early parenthood, maternal antenatal depression, mother's age at menarche onset. c Multiple imputation analysis was performed on a fully adjusted model. d Complete sample: individuals with complete data on exposure, outcome and all confounders. Table options Mediation model Finally, we tested the mediating effect of maternal depression and major financial problems on the association between early childhood father absence (first 5 years) and age at menarche (see Fig. 1). Age at menarche was utilised as an outcome variable, which was dependent on all other variables. Maternal depression and experiences of financial problems were dependent on father absence and were permitted to co-vary with each other in the model. The overall fit of the models was assessed using the comparative fit index/Tucker–Lewis index (CFI/TLI, >0.95 desirable; Bentler, 1990 and Hu and Bentler, 1999) and the root mean square error of approximation (RMSEA, <0.05 desirable; Browne and Cudeck, 1993 and Steiger, 1990). The chi-square test of overall fit is overly sensitive to model misspecification when samples size is large or the observed variables are non-normally distributed (Kline, 2005). For comparisons of nested models, the Mplus diffetst procedure was used, which computes differences in x2 of nested models ( Asparouhov & Muthen, 2006). First, we ran the indirect effects model in which the direct effect of early childhood father absence on age at menarche was constrained to zero. This model did not fit the data well (CFI = 0.95, TLI = 0.71, RMSEA = 0.06; n = 2750). The RMSEA test indicated that considerable co-variation was unexplained. Second, we tested the direct effects model in which the constraint concerning the path from early father absence to age at menarche was released. Fit for this model was very good (CFI = 1.00, TLI = 1.00, RMSEA ≤0.001; n = 2750). The x2 difference test suggested a better fit is provided by the direct effects model (x2 (1) = 9.80, p = 0.002). Following the approach for testing multiple mediation (Preacher & Hayes, 2008), we examined the total indirect effect in order to determine whether proposed mediators transmit the effects of father absence on age at menarche, as well as the specific indirect effects associated with each putative mediator. The multiple mediation analysis allows for estimation of indirect effects and 95% CIs whilst controlling for all other mediators included in the model (for graphical representation of structural equation model see Fig. 2). We found evidence for the total indirect effect of early childhood father absence on timing of menarche through exposure to maternal psychopathology and experiences of major financial problems (Table 4). As zero was not in the 95% CIs of the potential mediators examined, there was moderate evidence that both maternal depression and experiences of major financial problems explained some of the association between early childhood father absence and earlier age at menarche. However, an examination of the pair-wise contrasts of the indirect effects (Preacher & Hayes, 2008) revealed that they cannot be meaningfully distinguished in terms of magnitude and independent influence because zero is contained in the 95% CIs (Table 4). Thus, the unique abilities of each mediator to account for the effect of early childhood father absence on age at menarche could not be established on the basis of this analysis. Structural equation model: pathways from early father absence to earlier age at ... Fig. 2. Structural equation model: pathways from early father absence to earlier age at menarche in daughters. Path coefficients on the edges are unstandardised regression estimates. All variables in the model were single measured variables. Standard errors and thus 95% confidence intervals for the total and specific indirect effects were estimated using bootstrapping (1000). *p < 0.001; and **p < 0.05. Note: path coefficients in brackets are from the imputed data analysis. Figure options Table 4. Mediation of the effect of early childhood father absence (0–5 years) on age at menarche through exposure to maternal psychopathology and experiences of major financial difficulties in complete casea (n = 2750) and imputed samples (n = 4990). Model β Product of coefficients Bootstrapping SE Z BC 95% CI p Indirect effects (complete case analysis) Maternal depression −0.031 0.017 −1.823 −0.060, −0.004 0.068 Financial difficulties −0.043 0.023 −1.891 −0.086, −0.011 0.059 Total indirect effect −0.074 0.021 −3.561 −0.114, −0.043 <0.001 Direct effect −0.210 0.072 −2.958 −0.322, −0.097 0.003 Total effect −0.284 0.068 −4.169 −0.393, −0.172 <0.001 Contrast 0.013 0.034 0.365 −0.043, 0.072 0.715 Indirect effects (Imputed data analysis) Maternal depression −0.037 0.018 −2.111 −0.072, −0.002 0.035 Financial difficulties −0.027 0.017 −1.558 −0.060, 0.006 0.237 Total indirect effect −0.064 0.017 −3.741 −0.097, −0.031 <0.001 Direct effect −0.210 0.057 −3.660 −0.322, −0.098 <0.001 Total effect −0.274 0.054 −5.025 −0.380, −0.168 <0.001 Contrast 0.010 0.030 0.331 −0.049, 0.069 0.740 Note: BC: bias corrected; 1000 bootstrap samples. a Complete sample comprised individuals with complete data on exposure, outcome and all confounders. Table options Other associations within the model were in the expected direction (see Fig. 2). Specifically, early childhood father absence was associated with maternal depression (β = 0.444, 95% CI 0.325–0.563, p < 0.001) and financial problems (β = 0.471, 95% CI 0.330–0.610, p < 0.001), which were in turn associated with daughter's earlier age at menarche (maternal depression: β = −0.069, 95% CI −0.127 to −0.007, p = 0.056; financial problems: β = −0.092, 95% CI −0.166 to −0.022, p = 0.042 respectively). With only minor changes in regression coefficients, the mediation model analyses on the imputed data yielded the same pattern of results obtained from the complete case analysis. Specifically, evidence for an indirect effect of maternal depression was strengthened, whereas evidence for the indirect effect of financial difficulties was attenuated.