A longitudinal study of families created using egg donation: Family functioning at age 5.

Procedure

Phase 2 of the study was conducted between April 2018 and December 2019. All of the researchers were highly experienced in collecting data from families created by assisted reproduction. Three interviewers (JL, KS, and JG) visited families in their homes for a 2–3 hr visit when their child was 5 years old. Written informed consent was obtained from each parent. As far as possible, parents completed a similar procedure to Phase 1 ( Imrie, Jadva, Fishel, & Golombok, 2019 ). Each parent was administered an audio-recorded semistructured interview and completed a questionnaire booklet assessing psychological well-being, the quality of the relationship with their partner, social support, and the psychological adjustment of their child. Mothers and children, and fathers and children, were also filmed for 10 min completing a structured activity (Etch-a-Sketch or building blocks). To acknowledge their time, families received a small gift token and a small gift for their child. Written permission was also obtained from parents to contact the children’s teachers, who were asked to complete a questionnaire on the child’s adjustment. Written informed consent was obtained from teachers. The protocol was approved by the University of Cambridge Psychology Research Ethics Committee.

Social Support

Parents were administered the Multidimensional Scale of Perceived Social Support ( Zimet et al., 1988 ). The questionnaire has 12 items which produce three subscale scores of parents’ perceived adequacy of support from family, friends, and a significant other. For each subscale, higher scores reflect higher perceived social support. The questionnaire has good test–retest reliability and validity ( Zimet et al., 1988 ). Mean scale scores of 1 – 2.9, 3 – 5, and 5.1 – 7 are classified as low, moderate, and high support, respectively ( Zimet et al., 1988 ). Cronbach’s α were .95 (mothers) and .92 (fathers).

Couple Relationship Quality

The 36-item Golombok Rust Inventory of Marital State (GRIMS; Rust et al., 1990 ) was administered to parents to assess the quality of the couple relationship. The total score ranges from 0 to 84, with a higher scores reflecting poorer relationship quality. Relationship dissatisfaction is indicated by a score of greater than 34. The GRIMS discriminates significantly between couples who are about to separate and those who are not ( Rust et al., 1990 ). Cronbach’s α were .93 (mothers) and .93 (fathers).

Representational Measure: Parent Development Interview

Parents were administered an adaptation of the Parent Development Interview (PDI; Aber et al., 1985 ; Henderson et al., 2007 ), with mothers and fathers interviewed separately. The PDI is derived from attachment theory and is a semistructured interview examining parents’ representations of the parent–child relationship. Parents are asked to describe themselves and their child in moments of relatedness and interaction. PDIs were audio-recorded, transcribed verbatim, and coded using a coding scheme developed by Henderson et al. (2007) . The scheme yields codes assessing the parent’s representation of themself as a parent (parent affective experience codes), the parent’s representation of the child (child affective experience codes), and reflective functioning. SI and VJ were trained in the administration and coding of the PDI at the Center for Attachment Research at the New School for Social Research. SI and VJ coded the interviews, ensuring that they did not code mothers and fathers from the same family. They were largely unaware of family type, although occasionally a parent referred to their method of conception. The interviewers (JL, KS, and JG) were trained in the administration of the PDI by SI and VJ.

The parent affective experience codes were each rated on a 4-point scale, with a higher score representing a higher level of the construct: (a) degree of anger , assessing the extent to which the parent feels angry in the relationship; (b) expression of anger , assessing the degree to which anger is expressed in the relationship; (c) need for support , measuring the parent’s acknowledgement of need for support; (d) satisfaction with available support , measuring satisfaction with the support available to them; (e) guilt , assessing the extent to which guilt is present in the relationship; (f) joy/pleasure , measuring the parent’s ability to express feelings of joy in the relationship to and with the child; (g) competence , assessing how well the parent is coping with the child; (h) confidence , measuring the parent’s sense of their own competence; (i) level of child focus , assessing the extent to which the parent is focused on the needs of the child as compared to their own emotional needs; (j) disappointment/despair , measuring the degree to which the parent expresses disappointment with being a parent; (k) warmth , assessing the amount of warmth the parent feels toward the child; (l) attachment awareness and promotion , measuring the parent’s understanding of the attachment issues for their child and their ability to behave in ways that will promote the child’s attachment to them; and (m) hostility , assessing hostile feelings toward the child.

The child affective experience codes, used to assess the parent’s representation of the child, were also rated on a 4-point scale, with a higher score representing a higher level of the construct (a) child anger , measuring the extent to which the parent represents the child as experiencing/expressing anger; (b) child happiness , assessing the degree to which the parent represents the child as happy and contented as distinct from the parent–child relationship; (c) child controlling/manipulating , measuring the extent to which the parent represents the child as attempting to control the parent and their interactions; (d) child affection , assessing the degree to which the child shows and accepts physical affection in relation to the parent; (e) child rejection , measuring the degree to which the parent feels rejected by the child either emotionally or practically. The global code parental reflective functioning measures the degree to which the parent can reflect on the child and their relationship. The code captures the extent to which parents are able to “look underneath” the child’s behaviors for explanations, the extent to which they try to understand the child’s behaviors in terms of the child’s early experiences, and the extent to which they consider and evaluate their own contribution to any difficulties. One-third ( N = 39) of randomly selected mothers’ PDI transcripts were coded by a second rater to establish interrater reliability. Where discrepancies occurred, scores were discussed, and a final code agreed. Intraclass correlations ranged from .70 to .98.

Criticism

Parents were administered a semistructured interview designed to assess parenting quality ( Quinton & Rutter, 1988 ). In the interview, parents were asked to talk about detailed accounts of their child’s behavior and the parent’s response to it. Parental criticism was then rated using a standardized coding scheme based on a detailed coding manual. The degree of the parent’s criticism of the child was rated from 1 ( no criticism ) to 5 ( considerable criticism ). One-third of interviews were rated by a second coder. The intraclass correlation was .84.

Ratings of Psychiatric Disorder

A section of the interview with the mother was used to assess the presence of child psychiatric disorder using a standardized procedure ( Rutter et al., 1975 ). Mothers were asked to provide a detailed description of any emotional or behavioral problems displayed by the child, and information was gathered about the frequency, severity, precipitants, and course of behaviors over the last year. This was transcribed verbatim and rated by a child psychiatrist who was unaware of the child’s family background. A high level of agreement has been demonstrated between mothers’ assessments of their children’s emotional/behavioral difficulties and interview ratings ( Rutter et al., 1975 ). Ratings were made on a 4-point scale: 0 = none , 1 = slight , 2 = definite , 3 = marked . Type of disorder was identified as: emotional, conduct, mixed, developmental, attention deficit hyperactivity disorder, psychotic, or other.

Analysis Plan

To address the hypotheses regarding differences between family types in the mother, father, and child measures at age 5, univariate and multivariate analyses of variance were used and demographic covariates included when they differed between family types and were associated with the outcome measure. Parental age was not controlled for as it is a defining characteristic of the groups, with egg donation parents known to be older parents ( Golombok et al., 2005 ).

Given the longitudinal nature of the study, we used latent change score (LCS) models to examine intraindividual change over time in predictors ( McArdle, 2009 ) and used these scores to explore predictors of child adjustment problems. Where possible, our outcome measures reflected a composite score of maternal and paternal SDQ ratings. First, LCS models were used to examine intraindividual change over time in maternal measures ( McArdle, 2009 ). For example, the LCS model can be expressed as: Phase 2 Reflective Functioning = 1 Phase 1 R F + 1 ( Phase 2 R F − Phase 1 R F ) . 1

This first involves fixing the regression weight of the score at Time 2 as a function of Time 1 to 1. Then a latent factor score (e.g., ΔReflective Functioning) is defined by subtracting the Time 1 score from Time 2 with a factor loading fixed to 1. This interindividual change in the latent factor score can be examined error free and can subsequently be used as a predictor or outcome of interest. Thus, building on this initial model, Phase 2 child externalizing and internalizing scores were then regressed onto the LCSs, as well as the Phase 2 specific parenting measures (maternal criticism and observed mother–child interaction quality). Finally, to test whether family type moderated any of the associations, interaction terms were created. Specifically, predictor variables were first centered within Mplus and subsequently multiplied with family type, and then entered into the model. All of the models were run using Mplus Version 8 ( Muthén & Muthén, 2012 ) and model fit was evaluated using Brown’s (2015) recommended criteria: nonsignificant χ 2 , root-mean-square error of approximation (RMSEA) < .08, comparative fit index (CFI) > .90, and Tucker–Lewis index (TLI) > .90. As highlighted above, our LCS models were carried out using maternal measures as only 9.2% of mothers did not complete some of the Phase 2 questionnaire measures. Thus, we used a full information approach so that all eligible families were included in the model ( N = 122 families). This sample size gave us 80% power to detect medium size effects. Unfortunately, the proportion of missing data for fathers interview and questionnaire measures exceeded the widely established ≥ 20% threshold which would compromise any inferences (i.e., 29% missing paternal interview data and 20% questionnaire data). The sample size gave us 80% power to detect medium size effects. The data are not publicly accessible due to the potentially identifiable nature of the sample.

Fathers

Phase 2 measures of parenting stress, couple relationship quality, social support, and criticism were entered into a MANOVA, with paternal education included as a covariate. There were moderate significant differences between family types in parenting stress, with fathers in egg donation families reporting higher levels of stress compared to IVF fathers, F (1, 77) = 5.75, p = .019, Cohen’s d = 0.55. There were also moderate significant differences between family types in criticism, with egg donation fathers expressing higher criticism of their children than fathers in IVF families, F (1,77) = 4.80, p = .032, Cohen’s d = 0.50. There were no significant differences between family types in couple relationship quality or perceived social support.