Divergent secular trends in blood pressure and body mass index in children and adolescents in Hong Kong

Secular trends in blood pressure (BP) and body mass index (BMI) during childhood and adolescence are sentinels for the future population cardiovascular disease burden. We examined trends in BP z-score (ages 9–18 years from 1999 to 2014) and BMI z-score (ages 6–18 years from 1996 to 2014) in Hong Kong, China. Overall, BP z-score fell, systolic BP from 0.08 to −0.01 in girls and from 0.31 to 0.25 in boys. However, the trends were not consistent, for both sexes, systolic BP z-score was stable from 1999, decreased slightly from 2002 to 2005 and increased slightly to 2014, diastolic BP z-score decreased slightly from 1999 to 2004 and then remained stable to 2014. In contrast, BMI z-score rose from −0.15 to −0.01 in girls and from 0.14 to 0.34 in boys, mainly during 1997 to 2010. The upper tail of the systolic (except boys) and diastolic BP distribution shifted downwards, whereas the entire BMI distribution shifted upward. BP declined slightly whereas BMI rose in Hong Kong children and adolescents during the last 20 years, with systolic BP and BMI in boys above the reference. This warrants dual action in tackling rising BMI and identifying favorable determinants of BP, particularly targeting boys.

−0.004 in boys and for diastolic BP was −0.002 in both sexes (model 1). The age-standardized BMI z-score was higher by 0.016 per year in girls and 0.022 in boys. Comparing the changes in BP z-score with BMI z-score over time (from an interaction term between BMI z-score and calendar year) in a regression of BP z-score on BM z-score, the positive coefficient for interaction suggests the association of BP z-score with BMI z-score was stronger over time (model 2). Figure 1 shows mean systolic BP z-score was relatively stable from 1999 and then decreased slightly from 2002 to 2005 followed by a slight increase afterwards to 2014 in both sexes. Mean diastolic BP had a slightly decreasing tendency from 1999 to 2004 and then remained stable to 2014. Conversely, mean BMI decreased from 1996 to 1997 and then increased until 2009/2010 followed by stagnation or a very slight decreasing tendency to 2014 in both sexes. Overall, boys had higher systolic BP and BMI than girls, but similar diastolic BP to girls. Somewhat similar changes were identified by sex-specific jointpoint analyses (Appendix Table 1), although no clear changes were identified for BMI in girls. Figure 2 shows the distribution of systolic BP z-score had a marked downward shift in the upper tail for girls as did diastolic BP z-score in both sexes from 1999 to 2014, with overall a narrower distribution. The entire BMI z-score distribution shifted upward in both sexes from 1996 (or 1999) to 2014, but retained a right skewed distribution. The downward shift in the upper tail of the BP distribution and the upward shift of the entire BMI distribution was observed across all ages; except that the distribution of systolic BP in boys at age 12 years or above remained high from 1999 to 2014 (Appendix Fig. 1).

Discussion
In a recently developed Chinese setting, overall systolic and diastolic BP declined modestly whereas BMI rose generally among children and adolescents over the last two decades. As a result by 2014, girls on average had BP and BMI below or close to the reference, whilst boys had systolic BP and BMI further above the reference. In addition, the upper end of the systolic and diastolic BP distributions shifted downward (except systolic BP in boys), whereas the entire BMI distribution shifted upward from 1996 to 2014. This is the first large study with novel use of big data based on bi-annual BP and annual BMI measurements comprehensively covering childhood and adolescence, and so adds evidence concerning BP and BMI trends over nearly 20 years. To our knowledge, five previous studies examining both BP and BMI trends in children and adolescents over a similar period based on a population-representative sample from the same study have been published. The divergent patterns of BP and BMI in Hong Kong are inconsistent with patterns in rapidly developing China 9, 10, 13 , but are more similar to current trends in some other recently developed settings, such as the Seychelles 11 and South Korea 12 , and are also similar to previous trends in long-term developed settings, such as the United Kingdom 6 and the United States 7 . For instance, in recently developed South Korea, systolic BP z-score also declined but more strongly, by 0.08 per year in boys and 0.093 in girls from 1998 to 2008, with no clear changes in diastolic BP z-score. However, BMI increased from 1998 to 2005 before declining in 2007/08   21 .
The BP and BMI profile appears to be more adverse in boys than girls, whether this is a reflection of cultural practises and pressures which emphasize different physical attributes in boys than girls is unknown 22 . Nevertheless, it does suggest that any future trends in cardiovascular disease presaged by these trends are going to be more adverse in men than women, and perhaps growing boys should particularly be targets of intervention. Notably, these sex differences are congruent with the sex differences in cardiovascular disease that widen with economic development 15 , suggesting that early life factors contribute to this sexual disparity.
Our findings are consistent with a narrower diastolic BP distribution and an upward shift in BMI distribution in South Korea 12 and the Seychelles 11 , but are inconsistent with their downward shift in systolic BP in both sexes 11,12 , given the downward shift in the upper tail of systolic BP was only observed in girls but not boys in our study. It is possible that factors which cause high BP in children and adolescents, such as renal parenchymal disease 23 , or risk factors for high BP such as sleep disordered breathing 24 or early infections 25 , reduced over the period. The introduction of universal neonatal immunization for hepatitis B in Hong Kong in 1988 26 , or better access to care with the introduction of the Hospital Authority in Hong Kong in 1990 27 could perhaps play a role, but leaves the lack of fall in the upper tail of the systolic BP distribution in boys aged 12 years or older unexplained. Systolic blood pressure rises more at puberty in boys than girls, and some coincidental factor may be offsetting the downward shift in pubertal boys, such as sleep disorders due to playing computer games 28 , or environmental factors that promote specifically testosterone 29 .
Overall, our study adds evidence of some downward secular trend in BP despite rising BMI. Higher BMI is a major risk factor for BP 30 , which has been confirmed in Mendelian randomization studies of adults 31 . We also found slightly stronger associations of BP with BMI over time, suggesting the trend of slightly increasing BP after the mid-2000s could be partly related to the increasing BMI trend which could have partly counteracted a potentially more pronounced downward BP trend. Nonetheless, differing trends for BP and BMI here and elsewhere 6,7,11,12 also suggest that other environmental factors affect BP. One possibility is sodium intake, because sodium restriction in infancy might affect blood pressure in adolescence 32 . In Hong Kong, the proportion of young children with sodium intake higher than the recommended intake is less than 10% at <18 months but 31% at 48 months 33 , most (60%) young children also have vegetable intake and fruit intake lower than the recommendations 33 . However, whether these intakes have changed over the relevant time period is not known. Moreover, the role of vegetables and fruit in BP has not been clearly established in trials 34 . Breastfeeding rates have increased since the 1990s in Hong Kong 35 , although whether breastfeeding affects BP has not been unequivocally established 36 . Smoking during pregnancy is very low (<5%) in Hong Kong 37 . Alternatively, physical inactivity is associated with lower lean mass and higher fat mass 38 . Both lean mass and fat mass are similarly positively associated with BP 39 . Also, an increasing trend in BMI may be mainly due to gain in body fatness given fat mass may have increased more than lean mass in Chinese children in recent decades 40 . As such, more gain in fat mass but less lean mass might be possible, but whether this could result in slightly decreasing BP while BMI is increasing remains elusive. Physical inactivity is prevalent in children and adolesents in Hong Kong and is not known to have changed in a similar way as BP in recent decades 41,42 . A meta-analysis of short-term trials of physical activity did not reduce BP in children and adolescents 43 . Two small trials in obese children found physical activity modestly lowered BP, but one showed an increase in lean mass and no change in fat mass and weight 44 , while the other showed less abdominal fat and an increase in weight 45 . Future studies are needed to establish the causal role of physical activity and body composition (muscle mass and body fatness) in BP.
Several limitations are noted. First, attending the health assessment is voluntary. Differential selection by child health status or parental attributes or family socioeconomic position could bias the results, which is unlikely given health assessment at the Student Health Service is free and accessible to all public or private school students. However, it would only bias trends if the selection of children attending varied over time. Second, using a single BP measurement at a single visit may slightly overestimate average BP and pre-/hypertensive BP prevalence 46 , but would not affect comparisons over time. Third, random measurement error is possible, for which the large sample size compensates. Fourth, changes in devices over time could cause an artifactual change. Devices used to measure height and weight are reliable with little difference between devices. BP was measured by an automated oscillometric device, and re-checked with a sphygomanometer manually, which did not change during the period. Fifth, mid-arm circumference in children and adolescents may have increased in recent decades, implying an adult cuff size may be required for BP measurement 47 . However, BP was measured by nurses following a standard protocol with a cuff size appropriate to the age and size of the child. Sixth, we calculated BMI z-score using the 2007 WHO growth reference which used the same data as the 2000 United States Centers for Disease Control and Prevention reference but a different smoothing technique, and defined overweight and obesity using the International Obesity Task Force (IOTF) cut-offs which are based on nationally representative samples from six regions including Hong Kong 48 . As such, the absolute magnitude may not be directly comparable, but we focused on change for comparison across populations. Seventh, the study is descriptive and we can only speculate about the aetiologies of the observed divergent trends in BP and BMI. Finally, observed trends in BP might be driven by changes in potential risk factors for BP such as diet, physical activity and smoking, although most of these would also be expected to affect BP through BMI. The trends of these factors over the same period could not be analyzed due to lack of information.
In summary, as in some recently developed or long-term developed settings in earlier decades, divergent overall secular trends in BP and BMI were evident, resulting in systolic BP and BMI in boys further above the reference. Concerted efforts to tackle global increasing BMI trend is imperative; identifying the potential underlying factors that may alleviate the impact of rising BMI on BP or may lower BP independent of BMI are also important to improving future population cardiovascular health with a particular focus on boys.

Methods
Data Source. This study used routinely collected BP and BMI from the Student Health Service of the Department of Health, which provides free annual health assessments for all primary and secondary school students in Hong Kong 49 , for whom 9 years free universal public education (primary and 3-year junior secondary) has been provided since 1978 and 12 years (plus 3-year senior secondary) since 2008/09 50 . The Student Health Service was introduced in 1995/96 for primary school students, and was extended to secondary school students in 1996/97, but was suspended for secondary school students in year 2 and above in 2009/10 because of the Human Swine Influenza Vaccination Programme. The participation rate from 1995/96 to 2013/14 was 83.4% 51 . The health assessments include bi-annual assessments of BP (Primary 5 (age 10-11 years) onwards) and annual measurements of weight and height (Primary 1 (age 6-7 years) onwards). BP was measured by nurses on the right arm in a seated position after more than 10 minutes of rest with an age and size appropriate cuff size using an automated oscillometric device. Initial systolic or diastolic BP higher than the 90 th percentile for sex, age and height based on local references was re-checked by physicians with a sphygmomanometer after 15 minutes of rest and this second measurement was recorded. Height without shoes was measured by stadiometer to the nearest 0.1 centimetre and weight without shoes and outer clothing was measured by digital scales to the nearest 0.1 kilogram. BMI was calculated as weight in kilogram divided by height in meters squared. Coverage was incomplete in the early years, so we considered trends of BP since 1999 and BMI since 1996. We randomly selected one time point per participant so that there is no correlation between multiple measurements for the same participant. Given BMI was measured more often than BP and across a wider age range, more children and adolescents with BMI than BP were included.
Outcomes. Blood pressure. Systolic and diastolic BP were considered as standard deviations (z-scores) relative to age-, sex-and height-standardized blood pressure standards from the 2004 United States National High Blood Pressure Education Group 52 , to account for compositional differences by age, sex and the secular trend in height. Prehypertensive BP was also considered, defined as systolic or diastolic BP greater than or equal to the 90 th percentile but less than the 95 th percentile based on age-, sex-and height-specific standards or 120/80 mmHg for both boys and girls. Finally, hypertensive BP was defined as systolic or diastolic BP greater than or equal to 95 th percentile 52 .
Body mass index. BMI was considered as z-scores relative to the 2007 World Health Organization (WHO) growth references for 5-19 years 53 . Overweight or obesity was also considered, defined as a BMI for age and sex corresponding to an adult BMI of ≥25 kg/m 2 or ≥30 kg/m 2 using the IOTF cut-offs 54 .
For completeness, age-and height-adjusted BP and age-adjusted BMI by sex were calculated as the internally weighted average BP and BMI accounting for differences in mean age and/or height across years within the sample. Age-standardized BP and BMI by sex were also calculated using the 2000 WHO World Standard Population for direct standardization 55 .

Statistical analysis.
To examine the secular trends in BP from 1999 to 2014 and BMI from 1996 to 2014, we plotted sex-specific mean BP and BMI z-scores by calendar year, and reported the average difference per year. We used jointpoint regression analysis with modified Bayesian information criterion to identify when the slope of BP and BMI z-scores trends changed significantly 56 . To identify changes in BP or BMI distribution, we plotted their distributions in 1996, 1999 and 2014 using a kernel density function. To identify at what ages changes in BP or BMI distribution occurred, we further plotted their distributions by each age in 1996, 1999 and 2014. We summarized the overall sex-specific changes in BP and BMI z-scores across years by regressing z-scores on calendar year using linear regression. We also assessed whether the strength (slope) of the association of BP with BMI changed over time by regressing BP z-score on BMI z-score, calendar year and the interaction term between BMI z-scores and calendar year, of which a positive interaction term coefficient indicates a stronger association of BP with BMI in later periods 11  Kong West Cluster Joint Institutional Review Board. This study only used de-identified secondary data routinely collected by the Student Health Service, with informed consent obtained from a parent or guardian of the participants prior to participation to the voluntary health assessments.