This study was based on questionnaire data from the Norwegian Mother and Child Cohort (MoBa) study [25, 26], an ongoing longitudinal study of the general health and well-being of mothers and their children, starting during pregnancy. A subgroup of the mothers in the MoBa study participated in the MIDIA study (Environmental Triggers of Type 1 Diabetes) [6, 7] as well. In MIDIA, the newborn child was tested for HLA-conferred genetic susceptibility to type 1 diabetes. Questionnaire data from the MoBa study were retrieved and used for the analysis of mental health measures for mothers whose children were genotyped in the MIDIA study. MoBa questionnaire data from the 30th week of pregnancy ("the 30th week questionnaire") and 6 months post-partum (3-3.5 months after disclosure of test results, "the 6 month questionnaire") were used for the analysis.
The MoBa study was conducted by the Norwegian Institute of Public Health . In brief, MoBa is a cohort consisting of more than 100 000 pregnancies recruited between 1999 and 2008. The majority of all pregnant women in Norway were invited to participate, and the participation rate was approximately 44%. Participants were recruited for the study by postal invitation in connection with being invited to a routine ultrasound examination offered to all pregnant women in Norway at 17-18 weeks of gestation http://www.fhi.no/morogbarn. Informed consent was obtained from each participant before inclusion in the study, which was approved by the Regional Committee for Medical Research and the Norwegian Data Inspectorate. The current study is based on version 3 of the quality-assured data files released for research in 2007. Information from the Medical Birth Registry of Norway (MBRN) is also included in the dataset .
The MIDIA study is a longitudinal prospective study in which newborns from the general Norwegian population carrying the high-risk HLA-DRB1*0401-DQA1*03-DQB1*0302/DRB1*03-DQA1*05-DQB1*02 genotype were identified and followed for the development of islet autoantibodies and the onset of type 1 diabetes [6, 7]. Approximately 50% of the children tested in MIDIA have mothers who also participated in MoBa. Recruitment for the MIDIA study, which started in 2001, was described in detail by Stene et al. . All health care nurses across Norway who took part in the recruitment, were trained to give correct information about the project. Informed consent was obtained from each participant before inclusion in the study, which was approved by the Regional Committee for Medical Research and the Norwegian Data Inspectorate.
The mothers included in this study delivered between 2001 and 2006, with the majority of deliveries in 2005 (37%). The study initially included 9762 mothers who had their child genotyped in MIDIA and, at the same time, participated in the MoBa study. Of these mothers, 7801 returned both the baseline (30th week of gestation) and follow-up questionnaires (6 months post-partum). We excluded 411 pregnancies of women with more than one child. In most cases the first pregnancy or the first twin was included. In six cases the older sibling was excluded and the youngest included because the younger sibling was a high-risk child. A total of 7390 eligible participants were left for analysis.
Procedure for communicating genotype results in the MIDIA study
Information to parents of children with the high-risk genotype
The high-risk genotype information was conveyed to the parents first by an initial phone call from one of four health care workers on the MIDIA project. The information provided during the phone call followed a semi-structural scheme to ensure that important topics were correctly communicated, including how to understand the test results. They were also told that the genes were inherited from both the mother and the father. The parents were informed that they would receive a letter a few days later with much of the same information in written form. In the letter, the risk was explained with the wording: "Children with the 'diabetes risk genes' have a 20% lifetime risk of developing diabetes (one of five) and a 7% risk of developing diabetes by the age of 15 years (1 of 17 children)." All parents were told that they were welcome to call the project if they had more questions or were worried about aspects related to the increased genetic risk. After 1-2 weeks, the parents received a second phone call with the purpose of clarifying unanswered questions and providing additional information if needed. In addition, all questionnaires received in the MIDIA study were scrutinized by one of the nurses to detect negative reactions due to the risk notification and to become aware of mothers or fathers who were thinking about their child's high genetic risk for type 1 diabetes daily. If such signs were found, the parents were contacted by phone. The term "diabetes risk genes" was used in all communication with the parents, whereas the perhaps more frightening term "high-risk genes" was avoided.
Information to parents of children without the high-risk genotype
The parents of children without the high-risk genotype were informed by letter. The letter stated that their child did not have the diabetes risk genes, but it also informed the parents that children without these genes can develop type 1 diabetes. These children were not offered further follow-up.
Follow-up of children with high genetic risk in the MIDIA study
It is of relevance for the present study, how the children were followed-up between the high risk information was provided and until 6 months of age. Follow-up after 6 months is described in Stene et al. . When the high-risk-children in the MIDIA study were 3 months old, their parents received a package with a questionnaire. The package also included equipment for a blood sample to be taken at the health care clinic, and equipment for stool samples to be collected from the diaper when the child was 3, 4 and 5 months old . The parents were told that they would be notified if their child's blood samples showed elevated levels of islet autoantibodies. A similar package was received when the child was 6 months old. The 6 month follow-up package was sent to the parents approximately 2 weeks prior to the child turning 6 months old.
In the MoBa study, the questions concerning mental health were part of the questionnaire completed during the 17th and 30th week of pregnancy and when the child was 6 months, 18 months, and 3 years old. The present study used data from the 30th week of pregnancy and 6 month questionnaires (Additional file 1 and Additional file 2). The 6 month questionnaire from the MoBa study was sent to the parents the same week as the child became 6 months old.
Maternal symptoms of anxiety and depression: SCL-8
Maternal symptoms of anxiety and depression were assessed using a short version of the SCL-25 . The short version, SCL-8, consists of eight items: four items measuring depression and four items measuring anxiety. The correlation between SCL-8 and SCL-25 is 0.94 . Shorter versions of SCL-25 have been found to perform almost as well as the full version when measuring anxiety and depression [30, 31]. SCL-8 has been demonstrated to have an internal consistency of α = 0.88 .
A global SCL-8 score was calculated as a sum of the eight item scores. As the distribution of the total SCL-8 scores was positively skewed, a natural log transformation [ln(sum score)] was applied to the scores in order to approximate a normal distribution. The skewness was reduced from 2.6 to 1.2 and kurtosis from 10.3 to 1.2 for the SCL-8 variable at 6 months. The baseline SCL-8 variable was also transformed using the natural logarithm (ln). Cronbach's alpha (internal consistency) in the present sample was 0.83 for SCL-8.
The satisfaction with life scale
The five-item Satisfaction With Life Scale (SWLS)  was developed to measure the cognitive component of subjective well-being. In the original validation study , the SWLS demonstrated an internal consistency of α = 0.87. The instrument was calculated as a sum of the five individual item scores. Cronbach's alpha in the present data set was 0.88 for SWLS.
Rosenberg self-esteem scale
The short-form of the Rosenberg Self-Esteem Scale (RSES)  used in the MoBa study includes four items. These four items had a 0.95 correlation with the original ten item scale and an internal consistency of α = 0.80 . The instrument was calculated as a sum of the four individual item scores (inverting the scores from the two middle questions). Cronbach's alpha in the present sample was 0.78 for RSES.
Maternal worry about the child
Maternal worry about the child was one of the items in an 11-item checklist of life events experienced during the last year, given in the 6 month questionnaire. The question was phrased 'Have you been seriously worried that there is something wrong with your child?' Responses were coded as 'yes' or 'no'.
Maternal diabetes status
A dichotomous variable was constructed to indicate the presence of maternal type 1 diabetes. The variable was based on health questions from both the MBRN  and the MoBa questionnaires. The mothers' own answers from the MoBa study were used to supplement and, if contradictory, correct the MBRN records. The agreement between questionnaire data and MBRN records was assessed as κ = 0.81. The MoBa questionnaire concerning father's health, including if he had type 1 diabetes or not, was not available for this study.
Sociodemographic characteristics of the participants
Sociodemographic characteristics from the MoBa questionnaires that were used to describe the sample were maternal income and maternal education. Maternal age at delivery and marital status were abstracted from the records of the MBRN.
Child's genetic risk
A child's genetic risk was based directly on the results from the genotyping in the MIDIA study. The value of the variable was set to 1 if the mother had a newborn with high genetic risk for type 1 diabetes, and 0 if the mothers had a newborn without high genetic risk.
Handling of missing data in SCL-8, RSES, and SWLS
The MVA (EM) imputation procedure in SPSS 14.0 was used to impute missing values for each construct item using the remaining construct items as predictors. Data from respondents with more than four values of SCL8 missing from the eight items in total were excluded from the analyses. Data from respondents with more than two values of SWLS missing from the five items in total were excluded from the analyses. Data from respondents with more than one value of RSES missing from the four items were excluded from the analyses.
The imputation of the SCL-8 items increased the sample with valid data from 6994 subjects (complete data on all eight SCL-8 items) to 7349 subjects at the 30th week of pregnancy and from 6726 to 7361 subjects when the infant was 6 months old, whereas the number of subjects with valid data on both occasions increased from 6487 to 7321. Corresponding values for SWLS were 7277 subjects before and 7346 subjects after imputation (30th week of pregnancy), 7237 before and 7308 after (when the infant was 6 months old), and 7134 before and 7266 after (valid data on both occasions). Valid RSES data increased from 7301 subjects to 7339 subjects (30th week of pregnancy), from 7298 to 7345 (when the infant was 6 months old), and from 7215 to 7296 subjects (valid data on both occasions).
Regression analysis included answers from mothers with valid data at both baseline (the 30th week questionnaire) and after the disclosure of genetic risk (6 months questionnaire). Linear regression analyses were conducted separately for the dependent variables:
Symptoms of anxiety and depression (SCL-8)
Subjective well-being (SWLS).
In each of the three analyses, the dependent variable was the post-disclosure score for the maternal mental health variable. The independent variables were: child's genetic risk, maternal type 1 diabetes, and the baseline score for the maternal mental health variable. Exposure to risk information was the principal explanatory variable. By including the mental health baseline score as an independent variable, the effect of the independent variables can be interpreted as the effect in terms of a change in mental health from baseline to post-disclosure. Variables that are often chosen as covariates in multivariate analyses, such as demographic factors, are likely to be distributed by chance between the genetic high-risk and non-high-risk groups. As expected, preliminary bivariate tests showed no relationship between genetic risk and the demographic variables and, accordingly, demographic variables were not included in the analysis. However, maternal type 1 diabetes co-varies with the child's genetic risk and was entered as a covariate.
Logistic regression analysis was conducted with maternal worry about the child as a dependent variable. No baseline measurement was available for this dichotomous variable. Child's genetic risk and maternal type 1 diabetes were entered as predictors.
All analyses were conducted using SPSS, version 14.02.