Transgender fathering: Children’s psychological and family outcomes

Medical advances in assisted reproductive technology have created new ways for transgender persons to become parents outside the context of adoption. The limited empirical data does not support the idea that trans-parenthood negatively impacts children’s development. However, the question has led to lively societal debates making the need for evidence-based studies urgent. We aimed to compare cognitive development, mental health, gender identity, quality of life and family dynamics using standardized instruments and experimental protocols in 32 children who were conceived by donor sperm insemination (DSI) in French couples with a cisgender woman and a transgender man, the transition occurring before conception. We constituted two control groups matched for age, gender and family status. We found no significant difference between groups regarding cognitive development, mental health, and gender identity, meaning that neither the transgender fatherhood nor the use of DSI had any impact on these characteristics. The results of the descriptive analysis showed positive psycho-emotional development. Additionally, when we asked raters to differentiate the family drawings of the group of children of trans-fathers from those who were naturally conceived, no rater was able to differentiate the groups above chance levels, meaning that what children expressed through family drawing did not indicate cues related to trans-fatherhood. However, when we assessed mothers and fathers with the Five-Minute Speech Sample, we found that the emotions expressed by transgender fathers were higher than those of cisgender fathers who conceived by sex or by DSI. We conclude that the first empirical data regarding child development in the context of trans-parenthood are reassuring. We believe that this research will also improve transgender couple care and that of their children in a society where access to care remains difficult in this population. However, further research is needed with adolescents and young adults.

Consent for publication: Consent for publication was included in the Informed written consent and was 8 0 obtained from parents and children.

1 2
Availability of data and material: The whole data and material are available from the corresponding 8 3 author.
8 4 8 5 Code availability: As requested by French regulation, all data were processed anonymously and 8 6 confidentially. Data were identified only by a code number and correspondence between this code and the 8 7 participant's name/surname could only be established through a private list kept separately in another office. We 8 8 used the Pitié-Salpêtrière child psychiatry computerized database for the processing of these data (CNIL 8 9 declaration n° 1303778). The data of the present study were analysed using the statistical programme R, version 9 0 3.3.1 (R Foundation for Statistical Computing) [32].

Introduction 9 5
For years, individuals and couples with fertility issues have been able to conceive thanks to biomedical 9 6 technology advances [1]. These medical advances in assisted reproductive technology (ART) have created new 9 7 ways for transgender persons to become parents outside the context of adoption [2]. Becoming a parent is a 9 8 major life experience for human beings, and many trans people want to become parents [2]. However, 9 9 hormone/surgical treatments that can benefit transgender people are potentially sterilizing. Currently, it remains 1 0 0 technically impossible to develop the capacity to procreate after surgical transformation, but it is possible to help 1 0 1 trans people use their own gametes with assisted reproductive technology (ART). Trans parenthood is therefore 1 0 2 possible through adoption but also through other paths, including using trans person's own gametes. Therefore, available in many countries, as they are framed by laws and bioethics regulations, which can vary from one 1 1 3 country to another.

0 0
In order to test our third hypothesis, we used: (1)   answer any questions before the end of the 5 minutes". The critical dimension is based on the initial statement, 3 1 2 the expressed relationship and the blame or the dissatisfaction supported, whereas the emotional over-3 1 3 involvement includes emotional displays, statements of attitude (e.g., extreme loving), self-sacrificing and 3 1 4 overprotection or a lack objectivity, an excess of detail about the past, and more than five positive remarks 3 1 5 regarding the child [27]. For each child we recorded separately one session with the mother and one session with 3 1 6 the father. There is no specific age range for this instrument. A total score is also calculated.

1 9
Experimental procedure 3 2 0 To explore more subtle differences between children born from Trans-DSI (meaning a transgender 3 2 1 father) and NC children born from natural conception from cis-gender parents who had sexual intercourse and to 3 2 2 test our fourth hypothesis, we used a method that was previously developed to explore how traumatic experience 3 2 3 could be guessed without explicit information through participants' responses from an experimental task using a 3 2 4 permutation test [28]. Here, the task we proposed to children was drawing a family. We hypothesized that 3 2 5 children born from Trans-DSI would not be more engaged during family drawing to use atypical representations 3 2 6 (e.g., of men/fathers and of sexual indices) to be detectable by external raters. The task was inspired by the 3 2 7 Corman's Family Drawing Test [29], which assesses the child's perception of family relationship. Drawing is a 3 2 8 mediation offering the possibility of working from the projective and symbolic value of their contents. Corman's 3 2 9 Family Drawing Test examines the graphic level which considers the quality of the production (the line, its size, CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 26, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 its strength, the pace of drawing and the space of the sheet used to make it), the formal level derived from the 3 3 1 original studies of the "drawing of the good man" (the degree of development of the child through the 3 3 2 representation of the body of the characters' drawn, the link between the different characters, as well as the 3 3 3 different elements drawing), and the level of contents indicating a projective value of the drawing (unusual or 3 3 4 anxious representations, specific psychological problems of each child, valorisation/devaluation of certain 3 3 5 characters) [29].

6
This is on the ability of Family Drawing to promote child's projection associated with what the family 3 3 7 represents that we proposed our experimental procedure. The experimental procedure responds to an assumption 3 3 8 commonly found in French society and among childcare professionals: having a transgender parent could 3 3 9 influence the development of the child, his/her identity construction but also his/her representations especially 3 4 0 with regard to the family [16]. That is why we included a group of raters who were experienced child 3 4 1 psychoanalysts expert in the interpretation of children's drawings (see below). The previous study using the same 3 4 2 method [28] showed that psychoanalyst evaluators were the only ones to recognize better and above chance 3 4 3 adults who experienced a childhood trauma.

4 4
Raters with diverse experiences would eventually be in position to guess children's group by viewing 3 4 5 the drawings. To explore which experiences in raters may be helpful, the family drawings were analysed by 20 3 4 6 raters (4 child and family psychoanalysts (FAMPSY), 4 adult psychiatrists (ADUPSY), 4 biologists working in 3 4 7 ART (BIOL), 4 endocrinologists working with transgender individuals (ENDOC) and 4 students (STUD)). They

4 8
were randomly shown the drawings and asked to blindly classify them according to whether the child had a 3 4 9 transgender father using a 4-level Likert scale: I am certain that the drawing was done by a child from the Trans-3 5 0 DSI group, I think that the drawing was probably done by a child from the Trans-DSI group, I think that the 3 5 1 drawing was probably done by a child from the NC group, and I am certain that the drawing was done by a 3 5 2 child from the NC group. Differences between children's family drawings were evaluated with a generalization 3 5 3 of the "lady tasting tea" procedure to link qualitative and quantitative approaches in psychiatric research [30]. Power calculation 3 5 6 Given that we used 2 different methods, we had two power calculations to determine. For clinical 3 5 7 assessment, as we hypothesized that the psycho-affective development of children born by DSI whose father is a 3 5 8 transgender man will not significantly differ from that of children born by conventional DSI or from that of 3 5 9 children conceived by sexual intercourse of both cisgender parents, we needed to ensure that the number of 3 6 0 individuals included was high enough to ensure that if we had no differences between groups that the statistical 3 6 1 power was sufficient. The minimum size of the sample was calculated to be able to show with an alpha error 3 6 2 probability of 5% and a statistical power of 80% a significant difference between two groups on the CBCL, one  French population for each scale. We calculated for each scale the sample size needed to highlight a difference 3 6 6 between the normal range and the clinical range defined for the scale with an alpha error probability of 5% and a 3 6 7 statistical power of 80% given the reference mean scores and standard deviations for the French population. We . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

7 1
For the experimental procedure exploring whether raters blind to children status could classify 3 7 2 children's family drawings above chance levels, we used a permutation test based on a modified version of 3 7 3 Fisher's lady tasting tea procedure [28,30,33]. This statistical procedure was chosen to limit type I error. The 3 7 4 number of cases, controls and raters required to detect differences with a power greater than 80% for a p<.05 was 3 7 5 calculated by Falissard et al. [30]. For a sensitivity and specificity of correctly categorizing each subject, both 3 7 6 equal to 80%, 23 cases, 23 controls and 4 raters are enough to detect significant differences using the procedure  Data processing and statistics 3 8 0 As requested by French regulation, all data were processed anonymously and confidentially. Data were  The first analysis compared each variable across the three study groups: naturally conceived children 3 8 8 (NC group), children conceived by conventional donor sperm insemination (Cis-DSI group) and children 3 8 9 conceived by DSI from a transgender father (Trans-DSI group). Based on the qualitative exploratory study [15], 3 9 0 we hypothesized no difference between the groups. For each quantitative variable, we explored data distribution 3 9 1 and normality using visual exploration. When normality was not reached, we used the Kruskal-Wallis 3 9 2 nonparametric test. When normality was reached, ANOVA was used for 3-group comparisons, followed by 3 9 3 Student's t-test for 2-group comparisons. For qualitative variables, we used the chi-squared or Fisher exact test 3 9 4 according to the number of values. No correction for multiple testing has been done since our main hypotheses 3 9 5 were in favor of the null hypothesis.

9 6
The second analysis explored whether raters blind to children's status could classify children's family 3 9 7 drawings above chance levels. This analysis was limited to 2 groups. We used a permutation test based on a 3 9 8 modified version of Fisher's lady tasting tea procedure [28,30,33]. It is noticeable that since the raters know that 3 9 9 half of the records belong to "cases" and the other half to "controls", the ratings cannot be considered as

0 7
Therefore, the association between judges' ratings and the actual distribution of subjects into cases and 4 0 8 controls was tested in the following way. First, a score was computed for each group of raters: FAMPSY, 4 0 9 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. ; https://doi.org/10.1101/2020.06.25.20140020 doi: medRxiv preprint 1 0 when the raters correctly answered yes or no, +1 when they correctly answered probably yes or probably no, -1 4 1 1 when they incorrectly answered probably yes or probably no, and -2 when they incorrectly answered yes or no.

1 2
Thus, for each rater, the score could vary from +92 for all correct guesses (with a maximal certainty) to -92 for 4 1 3 no correct guesses. For each group of raters, the score could range from +368 for all perfects to -368 for 4 1 4 maximum failure.

1 5
To determine whether a group classified cases and controls better than could have been expected by 4 1 6 chance, a permutation test was performed as described above using R software version 3.3.1 (R Foundation for 4 1 7 Statistical Computing) [32]. The p-value was finally equal to twice the number of permutations for which the 4 1 8 score was above the score obtained for the original data set in the experiment. Given that we used a modified

2 4
Developmental characteristics of the participants 4 2 5 As shown in table 1, we found no significant differences between the 3 groups regarding general 4 2 6 intelligence, gender identity, or overall mental status as assessed with the CBCL. Only the Somatic complaints 4 2 7 Father's T-Scores showed a significant difference between the three groups (p = 0.043). Two-to-two 4 2 8 comparisons showed that the Trans-DSI group T-scores were significantly higher than those of the Cis-DSI 4 2 9 group (Wilcoxon p = 0.013). There was no significant difference between the Trans-DSI and the NC group or 4 3 0 between the Cis-DSI group and the NC group. However, the means of the T-scores in each of the 3 groups were 4 3 1 neither in the pathological zone nor in the limit zone. Regarding CBCL scores on an individual level, very few 4 3 2 had CBCL subscores reaching pathological scores. Father's CBCL Total score as Mother's CBCL Total score 4 3 3 was in the clinical range for one child in the Cis-DSI group. Father's CBCL Total score was in the clinical range 4 3 4 for one child in the NC group and for three children in the Trans-DSI group, but Mother's CBCL Total score was 4 3 5 in the normal range for all these children. In addition, no child or parent reported bullying or harassment during 4 3 6 the semi-structured interview. The trans identity of the parent was known within his family and in most cases 4 3 7 within his in-laws (15 out of 17 families). On the other hand, neither the friendly environment nor the 4 3 8 professional or school environment for children was informed of trans identity.

3 9
As the groups were quite young, especially in the Cis-DSI group, we had not enough data for the 4 4 0 Kidscreen to perform statistical analysis. Table S1 summarizes the quality of life characteristics of the 4 4 1 participants. Kidscreen was performed by 13 children and adolescents in the NC group, 10 in the Trans-DSI in the NC group. Then, we calculated the intra-class correlations between mother and father CBCL scores in 4 4 8 each group (table 2). We found that fathers globally responded as mothers for their child in the 3 groups looking 4 4 9 at the internalizing, externalizing and total scores. Differences were observed between fathers' and mothers' 4 5 0 responses to certain domains (withdrawn/depressed in the Cis-DSI group and attention problems and 4 5 1 aggressive/behaviour scales in the Trans-DSI group and in the NC group).

5 2
Family dynamics 4 5 3 Inventory of Parents and Peers Attachment was performed by 11 children and adolescents in the NC 4 5 4 group, 8 in the Trans-DSI group and only 2 in the Cis-DSI group. As the groups were quite young, especially in 4 5 5 the Cis-DSI group, we had not enough data to perform statistical analysis. Table S1 summarizes the attachment 4 5 6 characteristics of the participants. It appears that these preliminary data indicate rather secure attachment.

5 7
Concerning parental expressed emotions toward the child, table 3 summarizes the FMSS for mothers 4 5 8 and fathers of each group. In contrast to children's characteristics, all parental expressed emotions showed 4 5 9 significant differences across groups. Two-group comparisons (tables S2, S3, S4) showed that Expressed 4 6 0 Emotions, Criticism and Emotional Over-Involvement were significantly higher in Trans-DSI fathers than in 4 6 1 both NC and Cis-DSI fathers. There was no significant difference between the Cis-DSI and NC groups in 4 6 2 fathers' Expressed Emotion, Criticism and Emotional Over-Involvement. For mothers, we found that they 4 6 3 responded differently across groups but given the distributions in percentages, which were neither homogeneous 4 6 4 nor linear, we only conducted two-group comparisons: Expressed Emotions were higher in NC mothers than in 4 6 5 Cis-DSI mothers, Criticism was higher in Trans-DSI mothers than in NC mothers, and Emotional Over-

6 6
Involvement was higher in NC mothers than in both Cis-DSI and Trans-DSI mothers. Finally, we performed 4 6 7 intra-class correlations between parent pairs (table 2). We found that fathers and mothers responded differently  Are expert raters able to guess whether children have a transgender father when observing family 4 7 2 drawings? 4 7 3 In the experimental procedure, we asked four groups of expert raters to guess whether children had a  CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. ; https://doi.org/10. 1101/2020 To ease discussion of the current results, we propose to explore our 4 hypotheses. Our results validate 4 8 5 our first hypothesis namely that psycho-affective development of children born by DSI whose father is a 4 8 6 transgender man do not significantly differ from that of children born by conventional DSI or from that of 4 8 7 children conceived by sexual intercourse of both cisgender parents. We did not show any difference in cognitive 4 8 8 development, gender identity, or mental health problems when comparing the three groups. Moreover, results 4 8 9 regarding attachment and quality of life (descriptive analysis table S1) are overall reassuring. The cognitive 4 9 0 development of all children was good, with an average IQ of 113.9 (ranges: 93-150).

9 1
In terms of psychopathology, only three children whose father was transgender had a Father's CBCL

9 2
Total score in the clinical Z-score range (total sore = 64; 65; 68) and one had a score in the limit zone. For all the parents' transition, as they found that 34% of the children had mental health disorders in a study population of 55 4 9 6 children, even if these numbers did not significantly exceed the rates found in the general population. In

9 8
Our findings are consistent with not attributing disorders found in previous studies to the fact that these

1 6
Quality of life was only assessed in a small number of children due to the young age in our cohort, but 5 1 7 these descriptive results show no negative signal. In addition no child has been bullied or harassed in school or in 5 1 8 social relationships. The risk of harassment and stigmatization for children of transgender parents is a potential 5 1 9 risk [12,35]. Several studies have shown that peer harassment and teasing are infrequent i.e Freedman et al [13] 5 2 0 reported 33% of difficulties with peers in general, but no harassment or victimization; whereas Veldorale-Griffin 5 2 1 [36] found 33% of bullying at school. Qualitative studies found that the children of transgender parents reported 5 2 2 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. ; https://doi.org/10.1101/2020.06.25.20140020 doi: medRxiv preprint feeling protective towards their parent when they noticed discrimination or social rejection of their transgender 5 2 3 parent [37][38][39]. Also, many transgender parents were developing preventative strategies for themselves and their 5 2 4 children not to be stigmatized [40]. Our results are in lines with previous studies exploring children of 5 2 5 transgender parents. We believe that the low rate of harassment and stigmatization we found is related to the fact 5 2 6 that children in our sample were born after the transition of their parent. It is likely that the absence of disclosure 5 2 7 in the social environment outside the family circle contributed to prevent stigmatization and harassment.

2 8
We also note that all the children were cisgender identified. The interest in observing children's gender 5 2 9 identity was to answer the question of whether having a transgendered person as a parent had an impact on the 5 3 0 development of the child's gender identity. One assumption could have been that having a transgender parent

3 5
Regarding our second hypothesis (father and mother reports regarding their child do not differ) was 5 3 6 addressed by comparing mothers and fathers CBCL scores of the same child. We found that fathers globally 5 3 7 responded as mothers for their child in the 3 groups although differences were observed between fathers' and 5 3 8 mothers' responses to certain domains (withdrawn/depressed in the Cis-DSI group and attention problems and 5 3 9 aggressive/behaviour scales in the Trans-DSI group and in the TD group). It seems that neither gender identities, 5 4 0 nor the trans factor nor the mode of conception, and therefore whether there is a genetic link between the father 5 4 1 and the child, have an impact on the parent's view of the possible symptoms of their child.

4 2
Our third hypothesis was that family dynamics from children born by DSI whose father is a transgender to the young age of our sample. But the descriptive results show no negative signal in this area (table S1).

4 6
Regarding Expressed Emotions investigated with the FMSS, discussion should be cautious. Key aspects for 5 4 7 interpreting Expressed Emotions in families are the need (i) to distinguish fathers and mothers, (ii) to distinguish 5 4 8 the type of condition as it is well known that chronic and severe conditions may impact family functioning [27, 5 4 9 44-46], and (iii) to compare the same condition with the same instruments [47]. In our case, the current study is 5 5 0 the first to explore parenting in the context of DSI. As transgender fathers differed from both the two other 5 5 1 control groups, it seems that not the ART with donor insemination i.e. no genetic link between father and child, 5 5 2 but the trans factor influenced the scores.

3
In contrast with our hypothesis, parental expressed emotion towards the child was not similar across 5 5 4 fathers and across mothers. Indeed, we found that fathers and mothers responded differently in all three groups 5 5 5 except for the critical dimension in the Trans-DSI group, where fathers' and mothers' responses were correlated.

6
It seems, therefore, that the parent's gender identity has an impact on his/her expressed emotions towards the 5 5 7 child.

8
We believe that the facts that (i) transgender fathers differed from both NC group fathers and Cis-DSI is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. ; https://doi.org/10.1101/2020.06.25.20140020 doi: medRxiv preprint in terms of criticism might have some meanings for trans-parenting. Indeed, by combining the objective power calculation. (iii) It should also be noted that we did not reach the theoretical size of the sample required 6 0 0 for a power of 0.8 for the total score of the CBCL. This weakens our results although the theoretical number was 6 0 1 reached for each sub score of the instrument. (iv) Our study group and consequently our matched control groups 6 0 2 showed an over-proportion of boys (75%). This may be a viewed as a limitation. Nevertheless, despite matched 6 0 3 comparisons between groups, the CBCL considers the gender of the child in the Z-score statistics that we used.

0 4
Also, epidemiological studies show a 15% incidence of mental health in childhood and adolescence, slightly 6 0 5 higher and earlier in boys [49]. This would rather tend to reinforce the validity of our results. Second, the study 6 0 6 is transversal in nature and not prospective. Therefore, many children were young. This did not allow the 6 0 7 evaluation of certain parameters, such as quality of life and self-reported attachment instruments. Third, despite 6 0 8 our efforts to match participants for age, the NC and Trans-DSI groups were not perfectly matched for age to the 6 0 9 Cis-DSI group. Fourth, the response rate was 56% in our study group (17 families out of 30). Further research is 6 1 0 needed, especially with adolescents and young adults, as we cannot exclude that adolescence would eventually 6 1 1 impact children's development.

1 2
In conclusion, we explored cognitive development, mental health, gender identity, and family 6 1 3 dynamics in 32 children who were conceived by donor sperm insemination (DSI) in couples with a cisgender 6 1 4 woman and a transgender man who had his transition before conception. We compared children's psychological 6 1 5 and family outcomes in these children and in two matched control groups. Our study showed that the psycho-6 1 6 emotional development of children whose fathers are transgender is good and that there is no difference between 6 1 7 these children and those of control groups. Similarly, no rater was able to differentiate the family drawings of 6 1 8 children of trans-fathers from those of children in the NC control group. We also showed that the emotions 6 1 9 expressed by transgender fathers who conceived by DSI were higher than those of cisgender fathers who 6 2 0 conceived by sex or by DSI. The generalization of our results should consider the limitations listed above but 6 2 1 also the context of the sample that show middle/good SES and excellent family stability, two factors that 6 2 2 contribute to children's mental health [50,51]. We believe that this research will also improve transgender The authors thank all children and parents who participated in the study.    . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020. ; https://doi.org/10.1101/2020.06.25.20140020 doi: medRxiv preprint reproductive technologies for transgender people: a comparative cross-sectional study. Front Psychiatry. 2020;    . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 26, 2020.