Puberty Timing and Sex-Specific Trajectories of Systolic Blood Pressure: a Prospective Cohort Study

Background: Sex differences in systolic blood pressure (SBP) emerge during adolescence but the role of puberty is not well understood. We examined sex-specific changes in SBP preceding and following puberty and examined the impact of puberty timing on SBP trajectories in females and males. Methods: Trajectories of SBP before and after puberty and by timing of puberty in females and males in a contemporary birth cohort study were analyzed. Repeated measures of height from age 5 to 20 years were used to identify puberty timing (age at peak height velocity). SBP was measured on ten occasions from 3 to 24 years (N participants, 4062; repeated SBP measures, 29 172). Analyses were performed using linear spline multilevel models based on time before and after puberty and were adjusted for parental factors and early childhood factors. Results: Mean age at peak height velocity was 11.7 years (SD, 0.8) for females and 13.6 years (SD, 0.9) for males. Males had faster rates of increase in SBP before puberty leading to 10.19 mm Hg (95% CI, 6.80–13.57) higher mean SBP at puberty which remained similar at 24 years (mean difference, 11.43 mm Hg [95% CI, 7.22–15.63]). Puberty timing was associated with small transient differences in SBP trajectories postpuberty in both sexes and small differences at 24 years in females only. Conclusions: A large proportion of the higher SBP observed in males compared with females in early adulthood is accrued before puberty. Interventions targeting puberty timing are unlikely to influence SBP in early adulthood.


Contents
Methods S1 Details of deriving age at peak height velocity Methods S2 Details on measurement of blood pressure Methods S3 Details of model selection Methods S4 Additional and sensitivity analyses Table S1 Details of height measures available for deriving aPHV Table S2 Model details for SBP trajectories modelled using pubertal age, by sex and sexspecific quartiles of pubertal age Table S3 Results from likelihood ratio test examining linearity of association between age at peak height velocity and SBP at each age by sex Table S4 Characteristics at birth of the mothers of children included in models compared with those excluded due to missing exposure, outcome or co-variate data Table S5 Adjusted mean SBP in females and males and mean difference in SBP at age 3 years, puberty and age 24 years from pubertal age multilevel models Table S6 Unadjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models including all participants with data on aPHV and at least one measure of SBP from 3 to 24 years Table S7 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from weighted pubertal age multilevel models Table S8 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models restricted to participants with at least one measure before and one measure after puberty Table S9 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models restricted to participants with more than five measures of SBP Table S10 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from chronological age multilevel models Table S11 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models -including adjustment for fat mass at age 9 years Table S12 Adjusted mean trajectory in females and mean difference in trajectory of SBP in males, from sex-combined chronological age multilevel models -including adjustment for fat mass at age 9 years Table S13 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models -including adjustment for lean mass at age 9 years Table S14 Adjusted mean trajectory in females and mean difference in trajectory of SBP in males, from sex-combined chronological age multilevel models -including adjustment for lean mass at age 9 years Figure S1 Mean growth curve (black line) and velocity (blue dashed line) plots estimated by SITAR for females and males. Figure S2 Flowchart of study sample Figure S3 Mean adjusted trajectories of SBP in females and males before and after puberty from multilevel models based on pubertal age of 13 Methods S1 Details of deriving age at peak height velocity

Height measures included in the analysis
Height data from questionnaires and health records were excluded. Only data measured at clinic assessments carried out after age 5 years were included in the analysis. A Child in Focus (CIF) clinic measured height using a Leicester Height Measure on a 10% sub-sample of participants measured at age five. From 7 years onwards, standing height was measured to the last complete millimetre using the Harpenden Stadiometer at clinics carried out at ages 7,9,10,11,13,14,15 and 18 years. Data were further restricted to only include individuals with at least one measurement of height from 5 to <10 years, 10 to < 15 years and 15 to 20 years. The final dataset for analysis included 20,849 height measurements for 2,688 boys and 24,216 measurements for 3,019 girls. The number of height measures for females and males is shown in eTable 1. Further details of how age at peak height velocity (aPHV) was derived are described elsewhere (1-3).

Analysis deriving aPHV
Available height measures were analysed for females and males separately using Superimposition by Translation and Rotation (SITAR) growth curve analysis with five degrees of freedom (2,3). This method is a validated method of deriving aPHV and is described elsewhere in detail (1,2). aPHV was defined as the age when the first derivative of the mean curve, plotted as height versus age, was maximal. After fitting the initial model, the data were checked and points with velocity exceeding four standard deviations (SDs) and standardized residuals exceeding three in absolute value were removed. The model explained 98.5% of variance in males and 98.7% in females. Mean growth curve and velocity plots for females and males are shown in Figure S1.

Methods S2 Details on measurement of blood pressure
A Dinamap 9300 Vital Signs Monitor A was used to measure blood pressure at the 3, 4 and 5 year clinics. A Dinamap 9301 Vital Signs Monitor (Morton Medical, London) was used at the 7, 9, and 11-year clinics; an Omron MI-5 was used at the 10-year clinic; and an Omron IntelliSense M6 (Omron Healthcare, Kyoto, Japan) was used at the 15-and 18-year clinics. An Omron M6 upper arm blood pressure monitor was used at the 24-year clinic.

Methods S3 Details of model selection
Systolic Blood Pressure (SBP) was measured on ten occasions between 3 and 24 years. Values of SBP four SDs greater than or less than the mean were excluded from the analysis. We included all participants with at least one measure of SBP in each multilevel model, under a missing at random (MAR), to minimise selection bias. The observations of participants who reported being pregnant at the 18-year and 24-year clinics were excluded from the multilevel models at that time point only. We modelled sex-specific change over time in SBP according to pubertal age to better understand the association of age at peak height velocity with change in SBP during childhood and adolescence. In both models, linear splines were used to examine change in SBP (4). aPHV was normally distributed in both sexes. Linearity of associations of aPHV and change in SBP was examined by comparing the model fit of regressions of SBP on aPHV, with aPHV treated as a continuous exposure and as a categorical exposure (fourths of aPHV). Linearity was formally tested using a likelihood ratio test (see eTable 3). Prior to analysis, aPHV was centred on the sex-specific mean of aPHV for females and males.

Models based on pubertal age
Models examining change in SBP according to pubertal age were modelled de novo for this paper. We examined observed data at each age by sex to examine whether the shape of change over time was similar or different between quartiles of pubertal age. Shape of change over time across quartiles of pubertal age differed for females and males. Therefore, based on the observed data, we examined the fit of two models with four periods of change (one pre-and three post-puberty) and another model with five periods of change (two pre-and three post-puberty) separately in females and males. The model with the best fit in both females and males across each quartile of pubertal age was a four-spline model allowing for four periods of change; one pre-pubertal period (from three years before puberty to puberty) and three post-pubertal periods (from puberty to three years after puberty, three to five years after puberty and from five years after puberty to the end of follow-up) in females and two pre-pubertal periods (from three years to three years before puberty, from three years before puberty to puberty) and two post-pubertal periods (from puberty to three years after puberty, and from three years after puberty to the end of follow-up) in males.
The models for took the form of SBPij = β0 + u0j + (β1+ u1j )sij1 + (β2+ u2j )sij2 + (β3 + u3j )sij3 + (β4 + u4j)sij4 + eij where for person j at measurement occasion i; β0 represents the fixed effect coefficient for the average intercept, β1 represents fixed effect coefficients for the first average linear slope, β2 represents fixed effect coefficients for the second average linear slope, β3 represents the fixed effect coefficients for the third average linear slope, β4 represents the fixed effect coefficients for the final average linear slope, u0j to u3j indicate person-specific random effects for the intercept and slopes respectively, and eij represents the occasion-specific residuals or measurement error which was allowed to vary with age. The covariance between the final two spline periods was set to zero to improve model convergence.

Measurement of confounders adjusted included in main analyses
Birthweight was extracted from medical records. Gestational age at birth was estimated from clinical records. A questionnaire at 32 weeks gestation asked mothers to report their educational attainment, which was categorized as below O-Level (Ordinary Level; exams taken in different subjects usually at age 15-16 at the completion of legally required school attendance, equivalent to today's UK General Certificate of Secondary Education), O-Level only, A-Level (Advanced-Level; exams taken in different subjects usually at age 18), or university degree or above. Parity was defined as the number of previous pregnancies that had resulted in a live-or still-born infant collected at 18 weeks gestation. Smoking in the first trimester of pregnancy was self-reported by mothers at 18 weeks gestation; responses to smoking any tobacco (cigarettes, cigars, pipes, or other) were grouped as follows: no smoking, <10 per day, 10-19 per day or greater than 19 per day. Maternal age was reported in the mother's antenatal questionnaires. Maternal height and weight were self-reported from the questionnaire administered at 12 weeks gestation; these were used to calculate maternal BMI. Household social class was measured as the highest of the mother's or her partner's occupational social class using data on job title and details of occupation collected about the mother and her partner from the mother's questionnaire at 32 weeks gestation. Social class was derived using the standard occupational classification (SOC) codes developed by the United Kingdom Office of Population Census and Surveys and classified as I professional, II managerial and technical, IIINM non-manual, IIIM manual, and IV&V part skilled occupations and unskilled occupations. Marital status was obtained from antenatal questionnaires and classified as never married, widowed, divorced, separated, first marriage, marriage two or three. At 32 weeks gestation mothers were also asked about their partners' educational attainment, which was categorized as below O-Level (Ordinary Level; exams taken in different subjects usually at age 15-16y at the completion of legally required school attendance, equivalent to today's UK General Certificate of Secondary Education), O-Level only, A-Level (Advanced-Level; exams taken in different subjects usually at age 18), or university degree or above. Breastfeeding information used here was collected via questionnaires administered at 4 weeks, 6 months and 15 months.

Body Mass Index
From 1 to 5 years, measures were available from routine child health clinics for most children and extracted from health visitor records, which form part of standard child care in the UK. Data were also available from research clinic measurements on a random 10% subsample of the cohort. All cohort members were invited to research clinics from age 7 onwards. Across all ages parent-reported measures were available.
At the clinics, crown-heel length for children aged four to 25 months was measured using a Harpenden Neonatometer and from 25 months onwards standing height was measured using a Leicester Height Measure; weight was measured using Fereday 100kg combined scale (fourmonth clinic), Soenhle scale or Seca scale model 724 (eight-month clinic), Seca 724 or Seca 835 (12-month clinic), Seca 835 (18 months onwards). From age 7 years, all children were invited to annual clinics, at which standing height was measured to the last complete mm using the Harpenden Stadiometer and weight was measured to the nearest 0.1kg using the Tanita Body Fat Analyser and Weighing Scales (Model TBF 305). BMI was then calculated using these measures as weight (in kilograms) divided by height (in metres squared).
BMI has been modelled previously using fractional polynomials and is described elsewhere (4). Briefly, BMI was log transformed due to skewness of the data and fractional polynomials were used where age was raised to various combinations of powers (each of the following single powers, plus each combination of two powers: 0.5, 1, 2, 3, -0.5, -1, -2, natural log), from which we selected the best fitting curve (the one with the lowest likelihood value). The resulting curve contained three age terms including log age, log age* age and log age *age^2. To account for the likely reduced accuracy of parent-reported measurements, a binary indicator of measurement source (research clinic or health records versus parent-report) was included as a fixed effect. The variance of measurement occasion-level residuals (the differences between observed and predicted measurements) was allowed to vary with age for log BMI. The model took the form of: log BMIij = (β0+u0j+e0ij) + (β1+u1j)(ln(age)ij) + (β2+u2j)(age*ln(age)ij) + (β3+u3j)(age2*ln(age)ij) + (β8+e1ij)(measurement_sourceij) + eij(age_monthsij) where for person j at measurement occasion i; β's represent fixed effect coefficients, u0j to u3j indicate person-specific random effects for the intercept and linear, quadratic and cubic age terms respectively, and e1 represents the occasion-specific residuals or measurement error which was allowed to vary with age and according to measurement source.
Individual-specific residuals were derived from fitting these multilevel models of weight and height from one up to nine years of age, dropping all measurements before 12 months and after 108 months, fitted using the statistical software package MLWiN version 3.04.

eMethods 4 Additional and sensitivity analyses
We examined the characteristics of participants included in our analysis compared with participants excluded from our analysis due to missing exposure, outcome or confounder data, using socio-demographic characteristics measured at or close to birth to better understand generalisability and the potential for selection bias. We performed weighted sensitivity analyses using inverse probability weighting to address potential selection bias. The participant level weights were estimated using logistic regression using all sociodemographic characteristics listed above as potential confounders with the addition of gender and were subsequently incorporated into the multi-level models as level two weights which adjust for the unequal probability of selection of the participants (5). We additionally performed unadjusted analyses on the sample of participants that had data on aPHV and at least one measure of SBP from 3 to 24 years; this analysis included an additional 1,640 participants excluded from our main analysis due to missing confounder data and provided insight into potential selection due to missing confounder data. We also performed sensitivity analyses restricting the sample to participants with at least one SBP measure before and one after aPHV and to those with a minimum of five SBP measures in total during follow-up.
We explored the association between aPHV and chronological age-based trajectories of SBP using sex-stratified linear spline multilevel models with five periods of linear change (age 3-7 years, 7-12 years, 12-16 years, 16-18 years and 18-24 years) and compared results to findings from the pubertal age-based models used in our main analysis.
SBP was previously modelled according to chronological age using linear spline multi-level models from 7 to 18 years. The model for SBP is described elsewhere in detail (4). The knots for the model were placed at 12 and 16 resulting in three periods of change: from 7-12, 12-16 and 16-18. To model SBP from 3 to 24 years of age, we added two additional linear splines so that there were five periods of change: from 3-7, 7-12, 12-16, 16-20, 20-24. The models took the form of: SBPij = β0 + u0j + (β1+ u1j )sij1 + (β2+ u2j )sij2 + (β3 + u3j )sij3 + (β4 + u4j )sij4 + (β5 + u5j )sij5 + eij where for person j at measurement occasion i; β0 represents the fixed effect coefficient for the average intercept, β1 to β5 represent fixed effect coefficients for the average linear slopes of each linear spline, sij represents the specific spline period, u0j to u5j indicate person-specific random effects for the intercept and slopes respectively, and eij represents the occasion-specific residuals or measurement error which was allowed to vary with age.
For this analysis, we examined whether this model was appropriate for modelling change over time within quartiles of pubertal age to ensure that model fit was good across the entire distribution of pubertal age. The association between aPHV and chronological age-based trajectories was then examined separately for females and males by including an interaction between centred sex-specific aPHV and the intercept and each spline period, providing an estimate of the difference in the average trajectory of SBP from age 3 years to 24 years, per year later aPHV. Confounders were included as interactions with both the intercept and linear slopes.
Finally, to account for the potential confounding effect of pre-pubertal adiposity and body composition on the association between puberty timing and SBP, we performed all analyses adjusted for fat mass and lean mass at age 9 years, respectively. Whole body less head, and central fat and lean mass were derived from whole body dual energy X-ray absorptiometry (DXA) scans assessed at age 9 using a Lunar prodigy narrow fan beam densitometer. To further account for pre-pubertal adiposity and body composition in sex differences in SBP, we modelled data for females and males together using the chronological age model where we adjusted for fat mass and lean mass at age 9 years. This sex-combined model included sex and an interaction between sex and the intercept and each spline period.  P value from likelihood ratio test comparing fit of models regressing thirds of pubertal age treated as a continuous exposure on SBP at each age to models regressing thirds of pubertal age treated as categorical exposure on SBP at each age. P>0.05 indicates the more parsimonious model (pubertal age treated as continuous exposure) is a better fit, suggesting linearity of associations of age at peak height velocity and SBP. Note that although the p value for males at age 11 indicated some departure from linearity, associations were still broadly linear. Thus, given the linearity of all other associations in females and males, age at peak height velocity was examined as a continuous exposure in our analyses.

Table S6 Unadjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models including all participants with data on aPHV and at least one measure of SBP from 3 to 24 years
*Mean trajectory is centred on the sex-specific median of age at peak height velocity for each sex (age ~11.7 for females and age ~13.6 for males). †Estimated using regression coefficients for the intercept and rates of change per year spent in each spline period. SBP, Systolic Blood Pressure; CI, confidence interval; mmHg/y, millimetres of mercury per year.  Table S7 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from weighted pubertal age multilevel models *Mean trajectory is centred on the sex-specific median of age at peak height velocity for each sex (age ~11.7 for females and age ~13.6 for males). †Estimated using regression coefficients for the intercept and rates of change per year spent in each spline period. Adjusted for birth weight, gestational age, maternal education, parity, maternal smoking during pregnancy, maternal age, maternal pre-pregnancy BMI, household social class, marital status, partner education, breastfeeding, BMI residuals. SBP, Systolic Blood Pressure; CI, confidence interval; mmHg/y, millimetres of mercury per year.  Table S8 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models restricted to participants with at least one measure before and one measure after puberty *Mean trajectory is centred on the sex-specific median of age at peak height velocity for each sex (age ~11.7 for females and age ~13.6 for males). †Estimated using regression coefficients for the intercept and rates of change per year spent in each spline period. Adjusted for birth weight, gestational age, maternal education, parity, maternal smoking during pregnancy, maternal age, maternal pre-pregnancy BMI, household social class, marital status, partner education, breastfeeding, BMI residuals. SBP, Systolic Blood Pressure; CI, confidence interval; mmHg/y, millimetres of mercury per year.   Table S12 Adjusted mean trajectory in females and mean difference in trajectory of SBP in males, from sex-combined chronological age multilevel models -including adjustment for fat mass at age 9 years *Mean trajectory is centred on the sex-specific median of age at peak height velocity for each sex (age ~11.7 for females and age ~13.6 for males). †Estimated using regression coefficients for the intercept and rates of change per year spent in each spline period.
Adjusted for birth weight, gestational age, maternal education, parity, maternal smoking during pregnancy, maternal age, maternal pre-pregnancy BMI, household social class, marital status, partner education, breastfeeding, fat mass at age nine years. SBP, Systolic Blood Pressure; CI, confidence interval; mmHg/y, millimetres of mercury per year.

Trajectory Mean trajectory (95% CI) of SBP* in females
Mean difference (95% CI) in SBP in males Table S13 Adjusted mean trajectory and mean difference in trajectory of SBP per year later age at peak height velocity, from pubertal age multilevel models -including adjustment for lean mass at age 9 years *Mean trajectory is centred on the sex-specific median of age at peak height velocity for each sex (age ~11.7 for females and age ~13.6 for males). †Estimated using regression coefficients for the intercept and rates of change per year spent in each spline period. Adjusted for birth weight, gestational age, maternal education, parity, maternal smoking during pregnancy, maternal age, maternal pre-pregnancy BMI, household social class, marital status, partner education, breastfeeding and lean mass at age nine years. SBP, Systolic Blood Pressure; CI, confidence interval; mmHg/y, millimetres of mercury per year.

Figure S1 Mean growth curve (black line) and velocity (blue dashed-line) plots estimated by SITAR for females and males.
Vertical dotted line represents age at peak height velocity.

Figure S3 Mean adjusted trajectories of SBP in females and males before and after puberty from multilevel models based on pubertal age of 13
Models are adjusted for birth weight, gestational age, maternal education, parity, maternal smoking during pregnancy, maternal age, maternal prepregnancy BMI, household social class, marital status, partner education, breastfeeding, BMI residuals. Exact ages are 12.8 years for females and 12.4 years for males. SBP, Systolic Blood Pressure; mmHg/y, millimetres of mercury per year