Study population
Our study is conducted within the MoBa, which is a prospective population-based pregnancy cohort study conducted by the Norwegian Institute of Public Health [12]. Pregnant women from all over Norway were recruited from 1999 to 2008 at 17–18 weeks of pregnancy and 40.6% of invited women consented to participate. There are 114,500 children, 95,200 mothers and 75,200 fathers recruited in the cohort. Data used in this study are based on version 8 of the quality-assured data files, released for research in February 2014. The establishment and data collection in MoBa has obtained a licence from the Norwegian Data Inspectorate and approval from The Regional Committee for Medical Research Ethics. This study was approved by the Regional Committee for Medical Research Ethics in South-Eastern Norway
There were 96,875 singleton, live born pregnancies with no malformations and chromosomal anomalies. After excluding women with missing information in cell phone use in the first trimester (n = 9843), as well as in parity, maternal age, maternal education, year of delivery and child gender (n = 1804) and child’s language, communication, and motor skills at 3 years of age (n = 39,839), the eligible study population was 45,389 mother-child pairs. For the neurodevelopmental outcomes at 5 years our study population was 17,310 mother child pairs, with no additional missing information on the communication and motor skills.
Maternal cell phone use during pregnancy
The use of cell phones during early pregnancy was assessed by a questionnaire administered at 17th weeks of gestation. Pregnant women were asked to report their frequency of talking on the cell phone by choosing 1 of the 4 fixed frequency answers: “seldom/never”, “few times a week”, “daily” and “more than an hour daily”. In our analysis, women were categorized into 4 groups of cell phone use in early pregnancy according to their answer in this question as: “no use”, “low use”, “medium use”, and “high use”. Similar information on maternal cell phone use was collected at 30th week of pregnancy (n = 44,339).
Child language, communication and motor skills at 3 and 5 years
Early language development of the children at 3 years was assessed by the Dale and Bishop Grammar rating, in which the mother was asked to rate her child’s typical sentence structure by choosing one of the six response categories [13, 14]. The list of options is an ordinal grammar rating with the highest rate indicating the most complex use of language. We assessed the risk of having lower sentence complexity, by grouping any ratings bellow six (≤5) and using the highest rating as the reference group. We used this cut-off to capture potentially late language development, basing the rationale on the publication by Dale P.S. et al., where 11% of the typical children and 46% of the early language delay children scored ≤5 [13]. As we do not expect children with severe language delay in our study population, we used this cut-off for capturing children with inflated language scores, i.e. children who score lower than the typical developing groups. Children categorized as “not yet talking” (lowest rating) were excluded from our analysis (n = 103).
Communication skills at 3 and 5 years, were assessed by the “Ages and Stages” questionnaire (ASQ) [15]. We defined children with low communication skills, as those with score < 40 on the ASQ total score (0 to 60) at 3 years and those with score ≤ 30 on the ASQ total score (0 to 60) at 5 years. At 3 years, 524 (1.2%) children, and at 5 years, 95 (0.6%) children were ranked as having low communication skills, and 42 children were ranked with low communication skills in both time points.
Motor skills at 3 years were also assessed by the “Ages and Stages” questionnaire (ASQ) [15], and motor skills at 5 years by the “Child Development Inventory” questionnaire (CDI) [16]. We defined low motor skills at 3 and 5 years as having a score in the lowest tertile calculated for the included study sample. Due to skewed distributions of the scores the lowest tertile did not include the exact 33% of the children, but 32% of the 3-year olds and 23% of the 5-year olds, which were categorized as having low motor skills. The score cut-off reflecting the lowest tertile was 30 (score range 0–40) for 3 year olds and 10 (score range 0–12) for 5 year old children. Approximately 2069 children were ranked as having low motor skills in both time points assessed. More detailed information on the validity and scoring of the used instruments are presented in supplementary material (Additional file 1: Tables S1, S2a, b, S3a, b).
Other characteristics
Several maternal socio-demographic, lifestyle and pregnancy related characteristics were examined as potential confounders of the associations under study, including: maternal age (years), maternal education (≤12 years/13–16 years/≥17 years), parental income (both parents low income/either parent high income/both parents high income), parity (primiparous/multiparous), maternal occupation (public sector or military/private sector or self-employed/other), computer screen use during pregnancy (yes/no), marital status (living with partner/other), smoking prior to and during pregnancy (no/occasionally/daily), alcohol consumption prior to and during pregnancy (never or <1 time per month/1–3 times per month/≥1 time per week), use of folic acid supplements during pregnancy (yes/no), pre-pregnancy body mass index (BMI; <18.5, 18.5–24.9, 25–29.9, ≥30 kg/m2), type of delivery (c-section/normal) and the length of gestation (in weeks).
During the first years of life the interaction of the child with the mother/caregiver can affect the child’s psychosocial and cognitive development [17]. The amount of talk in the child’s environment, including talkative mothers, can promote vocabulary output and syntactic skills, trough high language input [18]. We hypothesized that women with an extrovert personality would talk more and report higher cell phone use than those with lower score, which can promote child’s communication skills. Maternal extrovert personality was assessed by the International Personality Item Pool (IPIP) Big-Five factor markers via a questionnaire administered at the 5 –years follow-up [19]. The scoring of the included 10 items (5 positive and 5 negative) resulted to a continuous score from 10 to 50, and women were categorized as “low” (score < −1SD), “average” (−1SD < score < +1SD) and “high” (score > +1SD) on extraversion, as suggested by Goldberg et al. [19].
Additionally, we used the year of delivery to assess time trends of cell phone use. Even though the recruitment of MoBa was finished in 2008, meaning that the cell phone questions were answered up to 2008, there are women delivering at 2009. Paternal use of cell phones for the 6 months before the pregnancy was assessed by a questionnaire administered around 15 weeks of pregnancy, but only 20,424 (45%) of the fathers provided information on mobile phone use, due to delayed administration of the fathers’ questionnaire. In addition, two different questionnaire versions were administered with different fixed answers of the questions assessing mobile phone use. We have included the question with the answers similar to those the mothers had to answer for comparability reasons, using the same labeling of the categories as for the mothers.
Based on a-priori assumptions, breastfeeding duration until 18 months (no breastfeeding/1–6 months/7–13 months/>13 months), child gender (boy/girl), maternal depression and/or anxiety before and/or during pregnancy (yes/no) were also assessed.
Characteristics that were univariately related with both the exposure and the outcome at 3 years, were included in our adjusted models as confounders.
Statistical analysis
We described the distribution and assessed the differences of maternal socio-demographic, lifestyle and pregnancy related characteristics by no use or any use of cell phone in early pregnancy. Further, in a bar graph we described the distribution of cell phone use in early pregnancy, by year of delivery.
The crude and adjusted associations between maternal cell phone use in early pregnancy and language, communication and motor skills of the children were assessed by crude and multiple logistic regression models. Two different classifications of the exposure variables were used: i) a bivariate of no use vs. any use, and ii) a 4-level variable of no use, low use, medium use and high use of cell phone in early pregnancy. In addition, two adjusted models were formed. First, one model with variables that were identified as confounders in univariate analyses and second, another model with the maternal extrovert personality score added. We performed complete case analysis of 45,389 mother-child pairs with the neurodevelopmental outcomes at 3 years and of 17,310 mother-child pairs with the neurodevelopmental outcomes at 5 years.
Further adjustment for maternal self-reported anxiety and/or depression during pregnancy did not modified our results, hence it was not included in the final models.
Sensitivity analyses
The association with language skills at 3 years was examined as the risk of having lower sentence complexity and, in sensitivity analysis, as the risk of having any of the four specific categories of language skills. We conducted all the analysis after excluding non-users, with low cell phone users constituting the reference group. In addition, we performed a stratified analysis by child’s gender to examine any gender-specific susceptibility.
Several studies have shown an exponential increase of mobile phone users from the mid to late 1990s’ in the Nordic countries as well [20,21,22]. Around 2003, mobile operators in Europe deployed Universal Mobile Telecommunications System networks (3G), which were upgraded in 2006, leading to a 4400 fold increase in data transmission rates [23]. In addition, from the mid- to late 2000s’ smartphones became more popular, including the introduction of the iPhone. Hence, we have stratified our analyses by these periods to investigate possible time trends in cell phone use. In addition, this categorisation provided groups with similar numbers of mother-child pairs (37%, 32%, 31% of the study population in each period group respectively). Further sensitivity analyses included stratified analysis by year of birth (1999–2004, 2005–2006, 2007–2009) to study the potential effect of changes in cell phone use. Since the number of non-users decreased substantially during the period 2005–2009, the exposure variable used was a 3-categories variable with low cell phone users as the reference group, after excluding non-users.
In addition, the association between paternal mobile phone use during pregnancy and child neurodevelopment was investigated as a sensitivity analysis. Approximately 16% of the women included in our study have participated in the MoBa study with more than one pregnancy. Hence, as a sensitivity analysis, we performed logistic regression analyses by taking into account the clusters of siblings within their mothers. Our study population changed from the 3 year to the 5 year follow-up due to loss of participants in addition to delayed administration of the questionnaire (approximately 3 years after the cohort had reached 5 years). We investigated whether the exposure variable and the confounders included in the models had a different distribution in each sub-sample.
All analyses were performed using STATA 12.1 (Stata Corporation, College Station, Texas).