Epidemiology of Non-communicable Diseases and Risk factors in South African Adolescents and Youth Living with HIV: Implications for Integrated Prevention


 Background

Adolescents and youth living with HIV (AYLHIV) face an elevated NCD risk resulting from HIV, psychosocial challenges and the complications of antiretroviral therapy (ART).
Methods

We conducted a cross-sectional study in six primary care facilities to investigate the prevalence of common NCDs and risk factors among AYLHIV in Cape Town, South Africa between March 2019 and January 2020. We collected information on pre-existent and previously unidentified NCDs and risk factors and collected sociodemographic information of adolescents and youth enrolled for primary HIV care. Characteristics between sexes and age groups were compared using parametric or non-parametric statistical tests.
Results

Three out of four participants were female, and the median age was 20.5 years (IQR 18.9–22.9). More than a quarter were not in education, employment or training (NEET) and 44% were multidimensionally poor. Our results show an existent burden of hypertension (5%) and central obesity (37%) as well as high levels of depression (43%) and psychological distress symptoms (44%). AYLHIV further self-reported high levels of household food insecurity (70%), low fruit and vegetable consumption, high fast-food and sugar-sweetened beverage intake, low nutritional knowledge and insufficient physical activity. Beyond the NCD risk attributable to HIV and ART, these multiple risk factors coupled with early initiation of high-risk behaviours like smoking, alcohol and drug use, further increase risk.
Conclusions

Our findings highlight the importance of integrated prevention with NCD risk screening as part of HIV care for AYLHIV and the need for early intervention on social, environmental and economic determinants of NCDs targeting adolescents and youth.

Sample size was determined using prevalence estimates from SANHANES-2012 for the 15-24 years age group [21]. The con dence level was set at 95%, with a 5% degree of precision, and an obesity prevalence of 5.6%. The minimum required sample size was 82 and the nal sample size determined to accommodate a non-response rate of 5% was 86 participants.
Prior to data collection, we conducted stakeholder engagement, liaising with facility managers and adolescent healthcare providers. Each facility has different schedules and models of HIV care for AYLHIV necessitating customised recruitment plans. We previously described the method used to estimate the number of AYLHIV accessing care at each facility (author). Facilities in the Eastern and Khayelitsha substructure had the biggest number of adolescents and youth receiving care. The AYLHIV population size was used as a guide for the proportion required for recruitment at each facility. Given the challenges of recruiting adolescents, in order to reach the required sample size, a convenience sampling approach was used recruiting all AYLHIV who showed an interest in participating. Participant recruitment commenced in March 2019 after ethical clearance (HREC ref no: 520/2017) and approval from Provincial and Local Government Departments of Health. Recruitment and study procedures were then conducted after gaining written informed consent and assent (for participants under 18 years) at six facilities until January 2020 (Fig. 1). A total of 176 adolescents and youth were recruited and invited to participate, of which 92 attended the follow up appointment and were successfully interviewed, yielding a response rate of 52%. Informed consent and assent forms were developed in English and translated into two local languages; isiXhosa and Afrikaans. At the rst encounter, participants were given information sheets and a detailed overview of the study procedures and invited to participate during routine clinical visits. Prior to enrolment, those aged less than 18 years received the informed consent and assent forms in order to gain parental or caregiver approval and a return date was scheduled on which the interview would take place (usually the date of their next clinic appointment). Teach-back questions on the nature of the study were incorporated into the assent form to ascertain adequate understanding. Those of legal age (18-24 years) were immediately enrolled after signing the consent forms and a convenient, future time was scheduled for their interview if they were unable to participate on the same day.
On the scheduled appointment date, participants rst underwent a 30-minute physical examination before completing a partially assisted self-administered questionnaire using electronic forms on a handheld Android tablet. The questionnaire, which was administered in English, took approximately 90 minutes to complete and included questions on socio-demographics, self-reported health status, family history, household characteristics, mental health, self-reported physical activity and dietary practices, use of alcohol, tobacco and illicit substances, and nutritional knowledge.

Physical Examination
Physical examinations were carried out according to the study protocol Standard Operating Procedures.
Height and weight were measured using a sliding balance weight-and-height measuring scale with participants barefoot and wearing light clothing. Height was measured to the nearest 0.5 cm and weight to the nearest 0.1 kg. Waist and hip circumference were measured using stretch-resistant measuring tape according to the WHO STEPS Protocol [79]. Readings were taken to the nearest 0.1 cm. Two measurements were taken and recorded on the electronic form of which an average was computed during analysis. For weight, height and waist-and hip circumference, if the two readings differed by more than 100 g, 2 cm and 0.1 cm respectively, a third measurement was taken, and the two closest measurements were recorded.
Sitting blood pressure (BP) and pulse were measured using a Rossmax automatic blood pressure monitor (Rossmax (Shanghai) Incorporation Ltd). Two readings were taken at least two minutes apart and the average was computed. The procedures were repeated in instances where there was a signi cant difference between readings (more than two units) and the closest two readings were recorded.
Participants were asked whether they had a family history of diabetes or experienced any of the following symptoms over the past three months: frequent urination, increased thirst, unexplained weight loss of more than 1.5 kg in the last month, unexplained fatigue and blurry vision. Random blood glucose was measured in those with a family history or reported symptoms using a point-of-care (POC) glucometer with reactive test-strips (Glucocheck Evolve ® Homemed Pty).
Respiratory symptom screening was conducted determining any di culty breathing or shortness of breath, prolonged cough for more than two weeks with sputum, chest tightness, noisy breathing (wheezing or whistling in the chest) or a history of asthma. Respiratory volume was measured in those reporting respiratory symptoms without known asthma or chronic respiratory conditions using a handheld peak ow meter with disposable mouth pieces. Several repeated attempts were made in accordance with standard guidelines [80]. If the variation between attempts was greater than 20 litres/minute [80], a further two attempts were made. The two largest values were recorded and the average computed.

Data collection tools and de nitions of composite measures Structural and household risk factors
Adolescence is often categorized into three primary developmental stages: early adolescence (10-14 years), middle adolescence (15-17 years), and late adolescence / young adulthood (18-24 years) [81]. For the purposes of this study, we categorized participants into four age groups in line with these stages and further sub-divided the oldest age group as follows: [15][16][17][18][19], 20-21 and 22-24 years.
Questions on socioeconomic status were derived from the 2011 South African Census Questionnaire's subset of variables used to measure multidimensional poverty [82]. The Youth Multidimensional Poverty Index (YMPI) is a multidimensional individual-level measure of poverty comprised of 11 weighted indicators in ve dimensions: educational attainment, general health and functioning, living environment, household assets and employment, using the method by Alkire and Santos [83]. Each of the indicators is associated with a deprivation cut-off that de nes whether a young person is deprived in that area [84]. The cut-offs for the educational attainment dimension are designed to coincide with key stages in the South African schooling system, allowing for a degree of delay in an individual's schooling career [84]. A deprivation score was calculated for each dimension and an overall composite score was derived from the weighted indicators. An individual is identi ed as being multidimensionally poor -MPI poor-if they are deprived in a third or more of the weighted indicators, with a composite score ≥ 33.3% [83]. The extent of poverty is measured by the percentage of deprivations experienced [85].
Food insecurity was measured using the Household Food Insecurity Access Scale (HFIAS) score. The HFIAS score is a continuous measure of the degree of food insecurity experienced in a household in the past month [86]. The participant is expected to answer for all members of the household and not just themselves. The higher the score, the more food insecurity the household experienced. Participants were categorized as living in food secure, mildly-, moderately-, or severely food insecure households depending on the severity and frequency of food insecurity experiences according to the HFIAS protocol [87].

Behaviour and Knowledge
Physical activity was assessed using the International Physical Activity Questionnaire (IPAQ) short form using the last seven days self-administered format [88] and graphic images of different kinds of vigorous and moderate-intensity forms of physical activity as an aid. The IPAQ has been validated in youth and adults in South Africa [89]. We used the Ainsworth et al. scoring algorithms to derive an average metabolic equivalent of task (MET) intensity level score for each type of physical activity: vigorous, moderate, walking and cycling [90]. Physical activity levels were further categorized according to the IPAQ scoring protocol into low, moderate and high [91]. Insu cient physical activity was de ned as a score below 600 MET minutes/week according to the World Health Organization (WHO) recommendations [92]. The presence or absence of sedentary behaviour was dichotomized as spending three or more hours per day watching television, playing computer games, talking with friends or other sitting activities according to the Global School-based Student Health Survey criteria [93].
Dietary intake was assessed using a 23-item food frequency questionnaire (FFQ) adapted from the Health Behaviour in School-aged Children Survey [94]. Participants reported their 'usual' consumption frequency of 23 different food groups, with response categories ranging from 'never' to 'more than once a day'. The FFQ has been found to have moderate reliability and acceptable validity for assessing the consumption of most food groups among adolescents [95]. Dietary intake was summarized into weekly consumption frequencies: 1, never; 2, less than once a week; 3, once a week; 4, 2-4 days a week; 5, 5-6 days a week; 6, once a day, every day; and 7, every day, more than once. We estimated the proportion who ate fresh fruits and vegetables daily (once or more than once a day in the previous week) and frequently (on 4 or more days in the previous week) and the proportion of respondents who reported daily consumption of sugar-sweetened beverages (SSB), deep-fried foods, fast foods, salty snacks, and processed meats. Skipping breakfast was de ned as eating breakfast on 0-2 days/week; semi-skipping, 3-4 days/week and not skipping, 5-7 days/week. Tobacco, alcohol use and substance use were assessed using questions from the 2011 South African Youth Risk Behaviour Survey [38]. Heavy episodic or binge drinking was de ned as drinking ve or more drinks in succession on one or more days in the preceding month. Risky drinking was de ned as binge drinking or underage drinking (any alcohol consumption below the legal age of 18 years).
Nutritional knowledge was assessed using a revised form of the General Nutrition Knowledge Questionnaire (GNKQ-R) [96]. The GNKQ-R has demonstrated internal consistency and is a valid and reliable measure of nutrition knowledge among young people [96]. A nutritional knowledge score was generated by totalling correct answers in four nutrition domains. The maximum possible score was 88; 18 for questions on "dietary recommendations", 36 for questions on food groups, 13 for "healthy food choices" and 21 for "associations between diet and disease". No norms exist to determine an adequate nutrition knowledge score [97]. Hence we computed the average nutrition knowledge score and compared the mean scores for each of the nutrition domains by gender and age groups comparable to a previous study of South African adults [98]. The percentage of respondents who answered the questions correctly in each domain was calculated.
Comorbidities (pre-existing diagnoses or presence of symptoms) Respiratory disease was de ned as self-reported pre-existing diagnosis of asthma, tuberculosis, bronchitis, or other lung disease. Experiencing any of the following symptoms in the preceding three months was characterized as presence of respiratory symptoms: prolonged cough with sputum for more than two weeks, chest tightness, shortness of breath, di culty breathing; or having an abnormal peak ow reading. Similarly, diabetes was de ned either as i) a pre-existing self-reported diagnosis; ii) a random blood glucose reading of > 7 mmol/l and having a family history of diabetes; or iii) a random blood glucose reading of > 7 mmol/l and experiencing any of the following diabetes-related symptoms over the past three months: frequent urination, increased thirst, unexplained weight loss of more than 1.5 kg in the last month, unexplained fatigue, blurry vision.
Depression and psychological distress were de ned using symptom screening questions from the 10item Centre for Epidemiological Studies Short Depression Scale (CESD-10) [99,100] and the Kessler Psychological Distress Scale (K10) [101]. Both tools have been validated in South African HIV-positive populations [102,103] and in adolescents and young adults [104]. The CESD-10 scale assesses depressive symptoms in the past week, while the K10 scale assesses symptoms of distress during the previous 30 days. Depression was de ned as a binary indicator using a cut-off score of 10 or more on the CESD-10 scale [99] and the likelihood of psychological distress was categorized according to the K10 score: K10 < 20, mentally well; K10 20-24, likely to have mild psychological distress; K10 25-29, likely to have moderate psychological distress; K10 30-50, likely to have severe psychological distress [105].

Measured clinical signs
Overweight and obesity were categorized using standard BMI categories and cut-offs for central obesity. The criterion used to identify overweight and obesity in children and adolescents using growth charts corresponds to the criteria used for adults thus we opted for the latter for ease of comparability [106]. The abdominal obesity cut-off point for a high waist circumference (WC) was ≥ 102 cm in males and ≥ 88 cm in females and a waist-hip ratio (WHR) > 0.85 for females and > 0.90 for males [79]. Waist-to-hip ratio was calculated by dividing the waist circumference by the hip circumference in centimeters rounded to two decimal places. A threshold value of 0.5 for the waist-to-height ratio (WHtR) was used as a measure of central obesity, calculated by dividing the waist circumference by the height in cm [107]. Blood pressure (BP) was categorized according to the South African Hypertension practice guidelines [108]: Normal (systolic BP, SBP < 130 mmHg and diastolic BP, DBP < 85 mmHg); elevated blood pressure (SBP 130-139 mmHg or DBP 85-89 mmHg) and hypertension (SBP ≥ 140 mmHg or DBP ≥ 90 mmHg) or a self-reported pre-existing diagnosis of hypertension.

Data Management and Statistical Analysis
Data from the physical examination and self-administered interviews were captured into an electronic form on an Android device using a unique anonymous study ID for each participant. At the end of each day, data was synced to a password-protected server only accessible to the study team. Data was then imported into Stata version 14.0 (StataCorp, College Station, Texas, USA) for cleaning and statistical analyses.
Graphical data exploration and Shapiro-Wilk's tests were used to test for normality of variables. All data were analysed and reported by sex and age group. Demographic and socioeconomic variables are described using summary statistics (frequencies, percentages, median and interquartile range (IQR)). Characteristics between sexes and age groups were compared using Pearson's χ2 and Fisher's exact tests for categorical measures. The Wilcoxon-Mann-Whitney test was used to compare medians between groups and the Kruskal-Wallis test was used for comparing continuous measures in more than two groups. Point prevalence estimates for comorbidities and symptoms were described and their exact binomial 95% con dence intervals were computed. All tests of signi cance were two-tailed and performed at the 5% signi cance level (p < 0.05).

Ethical Considerations
This study was approved by the University of Cape Town Faculty of Health Sciences Human Research Ethics Committee (HREC ref no: 520/2017). Approval was also granted by the Western Cape and City of Cape Town Departments of Health to access facilities for participant enrolment. Parents or legal guardians gave permission for their children to participate in the study, and participants provided informed assent (or consent if 18 years or older) for participation in the study. Participants received reimbursement for transport costs.

Results
Socio-demographic characteristics, structural and household risk factors A total of 92 participants completed study procedures, 76% of whom were female. The median age was 20.5 years (IQR 18.9-22.9) and the majority of our respondents (60%) were in the young adulthood stage between 20-24 years old (See Table 1). Over half (58%) were enrolled in school/college/university or other tertiary education or training (median age 20.3 (IQR 18.8-22.7) years). The majority lived with a biological parent or a relative, and 2% reported living with non-family members in foster care or children's homes. More than a quarter of female respondents reported ever being pregnant (28%) compared to 5% of male participants reporting ever impregnating someone. Of those ever pregnant, the majority (70%) reported having one child.

Multidimensional Poverty
Overall, 44% of participants can be considered multi-dimensionally poor as they were deprived in a third or more of the ve dimensions of MPI indicators as shown in Table 1. More than two-thirds (68%) were deprived in the general health and functioning dimension on account of experiencing some di culty with either hearing, sight, movement, concentration or with self-care. Fifty-nine percent were living-environment deprived with more than one-third living in informal housing, a quarter living in asset-deprived households, 28% living in households that do not use electricity, gas or solar power for heating and 17% living in households without piped water available on site. Almost a quarter (22%) of our respondents were deprived in the educational attainment dimension, comprising 3% of those aged 17-20 years who had completed less than nine years of schooling, and 19% of those aged 21-24 years who did not have a high school degree (or the equivalent of 12 years of schooling). Overall, 39% were economically deprived: 27% were neither in education, employment or training (NEET) (median age 20.8 (IQR 19.8-23.3 years)), while 15% were living in households with no employed adults of working age and 3% were deprived in both economic indicators.

Food insecurity
Seventy percent of participants were living in food insecure households with 38% considered as severely food insecure: either having to cut back on meal size, number of meals or going a whole day and night without eating due to a lack of resources or receiving relief food in the last 30 days.

Behaviour and Knowledge
Physical activity Overall, a third of respondents had insu cient levels of weekly physical activity, 41% had moderate levels and 27% had high levels of physical activity [91]. A greater proportion of males had high levels of physical activity compared to females (44% versus 22%). The total median MET-minutes of physical activity a week (total of all activities including active transport) was higher for males (2504.25 minutes) compared to females (1173 minutes), but this was not statistically signi cant ( Table 2). The youngest age group reported the highest rates of vigorous-intensity physical activity (Table 3). Over two-thirds of all participants reported using active transport in the preceding week (mostly walking to school/work for at least ten minutes continuously). Almost half of respondents (49%) spent more than three hours sedentary during a typical day, with no difference in sedentary behaviour by gender.

Dietary intake
Overall, less than a third of respondents ate fruits and wholegrains daily (  Table 3).
Meals eaten outside the home Two-thirds of respondents ate at least one meal that was prepared outside the home in the previous week. Those who ate food prepared outside the home, ate a median of two take-away or sit-down meals in the past week (IQR 2-4 meals). Males had a larger variability in meals consumed that were prepared outside the home compared to females; median 2 meals (IQR 1-10 meals) versus (IQR 2-3 meals) respectively.

Breakfast
Over half of participants (58%) reported eating breakfast on at least ve days in the previous week, 21% ate breakfast on 3-4 days in the week and 21% reported skipping breakfast or eating breakfast on less than two days in the previous week. Breakfast skipping did not differ signi cantly by gender or age.

Nutrition Knowledge
Overall, the mean general nutrition knowledge score achieved by adolescents was 33 out of a total of 88 points (37.5% There were no signi cant differences in nutrition knowledge amongst adolescents by sex and age group (See Fig. 2).

Tobacco use and exposure
Signi cantly more males (58%) than females (30%) reported ever smoking cigarettes (  . 3). Overall, 21% reported using tobacco products other than cigarettes (including water or hookah pipes and vaping) during the past month, with 6% reporting using these products on six or more days in the past month. Those in the youngest age group reported the most use of tobacco products other than cigarettes in the past month; 45% compared to less than 30% in the other age groups (Table 5). Almost half the participants (45%) reported being exposed to secondary smoke from people who smoked in their presence in the past week. More than half (15/26) of current smokers also drank alcohol in the preceding month.

Alcohol use
Over two-thirds of participants reported ever drinking alcohol (68%). Overall, the median age of rst alcohol use was 16.5 years (IQR 15-18 years) with no signi cant difference by gender. However, age of rst alcohol use was signi cantly earlier in the youngest age group with 82% of 15-17-year olds reporting ever drinking alcohol and a median age of rst alcohol use of 14 (IQR 12-14) years (See distribution in Table 5). Overall, 41% had drunk alcohol in the past 30 days and 24% reported binge drinking in the past month. Those who reported binge drinking were signi cantly younger (median 18.8 (IQR 16.9-21.5) years) than those who did not binge drink (median 21.1 (IQR 19.5-23.4) years) (p = 0.0152). Current drinking and binge drinking did not differ by gender. The youngest age group (15-17 years) reported the highest rates of risky drinking: 45% had drunk alcohol in the past 30 days (underage drinking) and 55% reported binge drinking compared to less than 30% in older age groups as shown in Table 5.

Substance use
More than a quarter (27%) reported ever using cannabis. The median age at rst cannabis use was 16 (IQR 14.5-18.5) years and did not differ signi cantly by gender. The frequency of cannabis use in the past year ranged from never (78%) to almost every day (6%) (See Table 4). Participants aged 18-19 and 20-21 years reported the most frequent use of cannabis with 13% reporting daily or weekly cannabis use each in the past year. Six participants (7%), all over 18 years old, reported ever taking "hard" drugs (cocaine, methamphetamine or mandrax); three of whom reported taking hard drugs 20 or more times in their lifetime. Nine participants (10%) reported ever using inhalants such as glue, aerosols, paint thinners, petrol or benzene "to get high" and 15% reported ever using over-the-counter or prescription drugs at least once or twice "to get high". Overall, more males (53%) reported ever using any illegal drugs or substances "to get high" compared to females (31%).

Respiratory symptoms
Eleven participants (12%; 95% CI: 6.3-21%) reported experiencing one or more of the following respiratory symptoms over the past three months: shortness of breath (3%), chest tightness (3%), prolonged cough with sputum for more than two weeks (7%), or di culty breathing (3%). Of those who reported di culty breathing, 2/3 reported it to be worse at night, with wheezing or whistling in the chest or occurring during exercise. Peak ow measurement was done for ve participants who reported respiratory symptoms but had no known asthma diagnosis. All the measurements were within the normal -green-peak ow zone.

Diabetes
Only one participant (who also reported diabetes symptoms) reported a previous diabetes diagnosis. A quarter of participants (95% CI: 16-35%) reported experiencing one or more of the following diabetesrelated symptoms over the past three months: frequent urination (4%), increased thirst (10%), unexplained weight loss of more than 1.5 kg in the last month (7%), unexplained fatigue (15%) and blurry vision (8%).
Younger age groups (15-17 years and 18-19 years) reported more diabetes-related symptoms compared to older age groups (See Table 7). Overall, 27% reported a family history of diabetes (95% CI: 18-37%). Of those with reported symptoms or a reported family history of diabetes who had their blood glucose measured (n = 31), none had a measured random blood glucose of more than 7 mmol/L.

Mental health screening
According to the CESD-10 depression scale, 43% of participants were classi ed as experiencing signi cant depression in the past week (95% CI: 32-54%). CESD-10 scores differed signi cantly by gender with female participants more likely to report depressive symptoms compared to males as shown in Table 6 and Fig. 4  reported a previous diagnosis of depression or anxiety. There were no signi cant differences in depression scores by age. Almost half the participants reported some level of psychological distress over the past month (45%; 95% CI: 34-56%) with 8% indicating symptoms of severe psychological distress on the K10 scale (See Fig. 5). Only 11% of those with mild, moderate or severe psychological distress reported previously being diagnosed with depression or anxiety. The prevalence of psychological distress symptoms was comparable across age groups (See Table 7).

Abdominal obesity
There was a markedly signi cant difference in the waist-hip ratio (WHR) of males and females. A total of 27% of females and no males had central obesity as shown in Table 6

Measured Blood Pressure
Overall, 75% of participants had a normal blood pressure, 20% had an elevated blood pressure and 5% had hypertension. Figure 6 shows that the prevalence of elevated blood pressure and hypertension was higher in males compared to females (41% and 20% respectively, p = 0.0367). Systolic and diastolic blood pressure showed an increasing trend with age (
We highlight several key ndings. Almost half of our participants faced multiple deprivations of poverty, signi cant symptoms of depression and psychological distress, and multiple risk factors for NCDs. More than a third were overweight or obese, a third had insu cient levels of weekly physical activity and the majority did not meet dietary guidelines for fruit and vegetable intake. There was low nutritional knowledge, particularly on healthy food choices and diet-disease relationships. Tobacco use and exposure and harmful use of alcohol were highly prevalent with male participants engaging in more substance use than females. An alarming pattern of early initiation of high-risk behaviours including underage and binge drinking, smoking and experimentation with cannabis emerged in the youngest age group. A detailed interpretation of these ndings, comparisons to the general population and previous ndings in PLHIV as well as implications for integrated prevention are discussed below.

76% of participants were female
More than two-thirds of our respondents were female which is consistent with national laboratory data for adolescents in HIV care [136]. This re ects the gendered nature of the HIV epidemic in South Africa, where almost a quarter of all new HIV infections occur in young women aged 15-24 years [137], and much higher rates of health care-seeking among young women compared to young men [42,138].
44% of AYLHIV were multidimensionally poor This is higher than national estimates which indicate that 33,4% of young people aged 15-24 years are MPI poor [139]. Nationally 34% of young people aged 15-24 years were not in education, employment, or training (NEET) in 2019 [140]. An "idle" youth population, not only impacts on the social cohesion and safety of a community, but is also linked to uptake of risk behaviours [141]. Although the proportion of NEETs in this study was lower than the national average with 73% of respondents either in education or engaged in income-generating activities, they experienced other deprivations which may interact to exacerbate vulnerability to NCDs and poor mental health. The highest deprivation was in general health and functioning dimension; with 68% experiencing di culties with general health and functioning, 35% lived in informal housing, and 38% reported experiencing severe food insecurity.
Previous studies in South Africa have demonstrated that HIV/AIDS-affected and infected youth face multiple deprivations of poverty [60,119], including poor educational outcomes [142], informal housing, lack of basic necessities like warm clothing, toiletries and school fees [143] and food insecurity [144].
Socioeconomic barriers such as poverty and food insecurity are commonly cited challenges for adolescents receiving HIV treatment and care in SSA [145]. This is concerning because socioeconomic factors like food insecurity impact adherence to ART and retention in HIV care [146], which has implications for viral suppression and chronic disease pathways [64,147,148]. These factors also have an impact on mental health and may increase susceptibility to alcohol and substance use [149][150][151][152].
Addressing this challenge requires a multi-sectoral approach for NCD prevention with appropriate social protection systems to meet the needs of the most vulnerable [143].

36% overweight or obese and 37% had abdominal obesity
More than a third of our respondents were overweight or obese (36%), with signi cantly more overweight and centrally obese females compared to males. Although our rates of overweight and obesity are slightly lower than prevalence rates for youth in the Western Cape (31.5% overweight and 11.3% obese) [38], the rates for ALYHIV appear similar to obesity trends in the general population. A previous study in adult patients attending primary health care HIV-clinics in South Africa found that more than half of female patients were overweight or obese compared to 16% of male patients [153]. Obesity in PLHIV is well documented in high-income countries and is emerging as a major challenge in Africa [50] with numerous studies showing increased rates of obesity in PLHIV [153][154][155][156][157][158]. But few studies in Africa have reported on overweight and obesity levels in AYLHIV other than in the context of ART-associated dyslipidaemia [129,159,160].
One study conducted among South African university students living with HIV (the majority aged 20-25 years) found that 21% were overweight and 30% were obese [161]. Findings from the United States Adolescent Trials Network showed that more than 40% of behaviourally HIV-infected young women (14-24 years) were overweight or obese [162] and approximately 36% of perinatally HIV-infected adolescents were overweight or obese [163]. Our results are consistent with these ndings and con rm results from a folder review conducted in this same population that reported similar levels of overweight and obesity [66]. Overweight and obesity in AYLHIV may accelerate their lifetime risk of cardiovascular diseases, additional to the effects of HIV-infection and exposure to ART [162]. In our study, obesity co-occurred with hypertension -35% of those with elevated blood pressure or hypertension were also overweight or obese.
By screening for obesity, other related conditions which tend to cluster with obesity can also be detected.
Closer monitoring of overweight and obesity pro les in young PLHIV in SSA is needed as they transition into adulthood, especially with prolonged exposure to ART regimens which are linked to obesity, altered glucose metabolism and dyslipidaemia [164].
In addition to BMI, we assessed central obesity using waist and hip circumference indicators. We found that 26% of our respondents with normal BMI had high WHR or WHtR, meeting criteria for central obesity.
Another South African study in adults attending three HIV-clinics reported a high prevalence of central obesity, primarily in women -45% (4% in men) [161]. In contrast, in a study conducted in Brazil, only 2.5% of children and adolescents on ART were overweight or obese based on subscapular skinfold thickness and less than 1% had a high WC [165]. In that study, the authors highlight that they expected a higher prevalence of overweight and obesity, but this was likely in uenced by parallel gains in height and weight observed in adolescents [165]. In a comparison of anthropometric measures for predicting cardiometabolic risk in HIV patients in Cameroon, Dimala et al found that markers of adiposity like WC, WHR and WHtR are better than BMI at identifying HIV/AIDS patients with increased cardiometabolic risk [166].
The WHO and the International Diabetes Federation (IDF) recommend monitoring changes in waist circumference in addition to measuring BMI, as this can provide an estimate of increased abdominal fat even without a change in BMI, particularly in HIV-positive populations on ARV medication and in female patients who have a higher prevalence of obesity [79,167]. Waist circumference and WHtR have been shown to be better predictors of cardiovascular disease risk factors in children and adolescents than BMI [168]. Waist-to-height ratio has the additional merit of not being dependent on age, sex or ethnicity, as the standard cut-off value of 0.5 is indicative of an increased cardiometabolic risk universally [107]. These ndings underline the importance of anthropometry beyond BMI, especially in females. Anthropometric measurements and calculations are non-invasive, low-cost and easy-to-use interventions that can be used in primary care to identify AYLHIV who are at increased cardiometabolic risk.

High levels of depression and psychological distress
Almost half our participants reported symptoms of psychological distress, compared to less than a quarter of young people nationally [21]. Only 11% of those with mild, moderate or severe psychological distress reported previously being diagnosed with a mental health condition. Mental health conditions are prevalent in AYLHIV in both high-income and resource-limited settings [169]. Our results are generally consistent with prevalence rates of depression among children and adolescents living with HIV from other African countries which ranged from 18.9% in Malawi [67], 25% in Rwanda [170], 27% in Tanzania [171] and 51.2% prevalence of psychological distress in Uganda [172]. These results are di cult to pool together due to non-uniformity in the methods used for mental health screening [169].
Signi cantly more female participants reported depressive symptoms compared to males (51% compared to 19%) which is in line with global statistics on depression [173]. After the age of 15, girls and women are twice more likely to be depressed compared to boys and men [174]. Our ndings show that only 17% of those identi ed as having signi cant depression via CESD-10 reported being previously diagnosed with anxiety or depression. In a previous retrospective review in the same population, mental health conditions were documented in less than 5% of folders reviewed [66]. These ndings highlight that there is a missed opportunity for identifying youth with mental health problems before suboptimal ART adherence or other adverse HIV, NCD and mental health outcomes occur.
A recent study conducted amongst AYLHIV aged 9-19 years attending a primary care clinic in Johannesburg found that 8% screened positive for symptoms of depression which is much lower than our ndings. However, 60% of those study participants were young adolescents aged 9-12 years and almost all were perinatally-infected (92%) [116]. Older adolescents in that study (aged [16][17][18][19] were more likely to screen positive for depression compared to younger adolescents, which is in line with our ndings of higher depression scores in older adolescents and young adults. It has been documented that perinatally-infected adolescents may present with less psychological problems compared to behaviourally-infected adolescents [119,175]. But similar rates of mental health conditions have been reported in HIV-exposed but uninfected youth entering adolescence [176]. The use of inappropriate comparison groups in studies showing higher rates of mental health disorders in AYLHIV, makes it di cult to determine whether these impairments are in fact due to HIV-infection or other social confounding factors. In a recent study conducted in Soweto, South Africa, the authors found similar rates of mental illness in perinatally-infected and uninfected adolescents, suggesting that other contributing social factors prevalent in the community may override the effect of HIV, especially in the era of highly active antiretroviral treatment [177].
Mental health screening is crucial in HIV care due to multiple psychological vulnerabilities associated with living with HIV and high rates of suicide in PLHIV [178][179][180]. Recommendations calling for integration of mental health services into HIV care have been made for adults [181]. It is important that adolescents are not overlooked in this respect. A study conducted in Johannesburg found that a simple way of identifying youth struggling with mental health problems at primary care level is by asking them about their future aspirations. Those who do not feel like they have control of their future or do not have a dream for the future are more likely to have symptoms of depression, anxiety or PTSD requiring further support [114].
Low nutritional knowledge, especially on healthy food choices Participants scored less than 40% on general nutrition knowledge questions and particularly had poor knowledge of healthy food choices and associations between diet and diseases. Only 10% recognised that eating thick-cut instead of thin or crinkle cut chips could help reduce the amount of fat in someone's diet and 22% correctly identi ed that eating less trans-fats can prevent heart disease. To our knowledge, this is the rst study to assess nutrition knowledge in AYLHIV in Africa. A South African study in schoolgoing adolescents aged 15-18 years found that 77.5% scored below average on diet and nutrition knowledge questions [182]. The poor knowledge on nutrition-related NCDs amongst AYLHIV in our study is concerning in a country undergoing nutritional transition [183]. Nutrition knowledge is strongly correlated with dietary intake and is needed for better dietary habits [97]. Although adolescents may lack autonomy in navigating their food environment, this life stage is characterised by increasing independence and as such they need to be informed about the importance of diet and how it can affect their current health status or future adult health [182]. Some practices like healthy cooking methods requiring steaming, roasting or baking can be adopted in their homes with cooperation of parents and caregivers [184]. With adequate knowledge, older adolescents in our study who ate more fast-food and more meals prepared outside the home, could be encouraged to make healthier food choices.
High fast-food and SSB consumption, and low physical activity Almost three-quarters of AYLHIV did not eat fruit daily and almost half did not eat vegetables daily, falling below recommended dietary guidelines of eating at least ve portions of fruit and vegetables daily in order to reduce the risk of NCDs [185]. Female participants ate less fruit and ate more fast-food, deepfried foods and foods with added sugar daily compared to males. Younger adolescents ate less fruits, vegetables and wholegrains compared to older age groups.
More respondents, particularly females, consumed deep-fried foods and fast-foods daily compared to provincial estimates for youth in the Western Cape, while the proportion who consumed SSB daily was lower (29% compared to 42%) [20]. Our results are comparable to local estimates that one in ve schoolgoing youth skip breakfast [38]. A recent study on fast food and carbonated soft drink consumption among adolescents aged 12-15 years in 44 LMICs (not including South Africa) found that 44% of adolescents consumed a carbonated soft drink at least once per day in the past month [186]. A metaanalysis on SSB intake found that individuals who consumed 1-2 servings per day, had a 26% greater risk of developing type 2 diabetes, and a 20% greater risk of metabolic syndrome compared to those who did not consume SSB or had less than one serving/month [187]. In order to promote healthy diet, a multisectoral approach that promotes a healthy food environment is required. The South African government has made strides in promoting healthier food environments by implementing mandatory legislation for salt reduction in processed foods in June 2016 [188] and a tax on sugar-sweetened beverages in April 2018 [189]. These measures were introduced in efforts to reduce the prevalence of hypertension, obesity, NCDs and excess salt and sugar consumption [190]. More efforts are needed to translate these measures into action at a community and household level, especially amongst young girls who have higher prevalence of obesity which is likely to persist till adulthood without intervention.
Our results support ndings of gender differences in physical activity levels among South African adolescents similar to global reports [191,192]. Despite more than two-thirds of our respondents using active transport; either walking to and from school or work, fewer than one-third had su cient levels of physical activity necessary to promote health and prevent chronic diseases. Almost half spent three or more hours per day of their leisure time sedentary. This is similar to physical inactivity levels reported in urban-based South African students [191]. Sedentary behaviour was higher than general population estimates which showed that 30% of youth watched TV or played computer games for over three hours per day [38] and higher than estimates from other LMICs which found that 27.0% of adolescents engaged in three or more hours of sedentary behaviour per day. Our results are consistent with those from a study in Brazil which found that 71% of ALHIV were sedentary with a higher proportion among girls [193]. Another study conducted in Botswana, found that youth living with HIV had signi cantly lower levels of daily PA compared to uninfected controls [129]. Similarly, a study in Brazil in 10-15-year-old perinatallyinfected adolescents and age-sex matched controls also found that participants living with HIV had lower physical activity scores compared to healthy peers [194]. Additional research is needed in this setting to explore the relationship between PA and HIV in adolescents. A study with an age-and sex-matched uninfected control group from the same community would help to elucidate whether this relationship exists in South Africa.
Early initiation of high-risk behaviours (smoking, alcohol, drugs) The median age at smoking initiation of 16 years found in our study was similar to the national average of 15.8 years for youth aged 15-24 years. However, our ndings that 48% of male and 25% of female AYLHIV smoked daily or occasionally was much higher than national estimates for young people aged 15-24 years reported in the 2016 Demographic and Health Survey (29% of males and 5% of females [20]). Furthermore, almost half of those in the youngest age group reported at least occasional cigarette smoking in the past month and used more alternative tobacco products like water-pipes compared to older age groups. Young adolescents in South Africa increasingly use water-pipes (known as hookah pipes) which are often available without restriction [195,196]. Our results likely re ect trends in the Western Cape -the province with the highest prevalence of tobacco smoking in South Africa-where a quarter of school-going youth are current smokers [38] and the mean age of smoking initiation is signi cantly lower than the national average at 14.5 years [21]. There are only a few studies, most from high-income settings, that address the prevalence of smoking among AYLHIV [197][198][199]. These studies report higher rates of smoking among AYLHIV compared to the general population, particularly among those who were behaviourally-infected [200]. Our ndings corroborate this and add to the limited literature from low and middle-income settings.
Smoking increases the vulnerability of PLHIV for adverse lung health and multiplies their risk of developing cardiovascular diseases compared to HIV-negative smokers [201][202][203]. Smoking often cooccurs with other health risk behaviours like alcohol consumption [204]. More than half (58%) of current smokers in our study had also drunk alcohol in the preceding month. Our results also indicate higher rates of current alcohol consumption compared to national estimates for South African youth (41% versus 33%) [38]. A few studies in SSA have reported higher occurrence of alcohol consumption in AYLHIV compared to HIV-negative adolescents [135]. Auvert

et al reported that 29% of AYLHIV in a South
African mining town drank alcohol at least once a month although this study was conducted almost 20 years ago [134]. Another study conducted in Zimbabwe reported that 5.6% of adolescent females living with HIV drank alcohol in past month, compared to 4.5% of HIV-negative females [133]. Several studies conducted in young people from other regions have reported increased alcohol and drug use especially among male AYLHIV [205]. Our results suggest that this may be the case, but further research is needed in South Africa and SSA to elucidate whether HIV infection is associated with increased alcohol consumption in young people living in settings like South Africa where heavy drinking is endemic [22].
Our respondents reported higher rates of binge drinking in the past month compared to estimates for the general population [20] (21% of females and 37% of males in our study, compared to 5% and 21% nationally). Alarmingly, half of the underage respondents (< 18 years) were current drinkers and 55% of them engaged in recent binge drinking. Males reported signi cantly more use of illegal drugs or substances to-get-high compared to females. Previous studies on risk behaviour among AYLHIV in SSA report a high prevalence of alcohol and substance use behaviour, especially among males in late adolescence [135,206]. In a study conducted amongst young people aged 15-26 years in the rural Eastern Cape province of South Africa, 4% of female AYLHIV reported problem alcohol drinking and 5% reported ever using drugs [131], whilst 31% of male AYLHIV reported problem drinking and 54% reported ever using drugs [132]. Alcohol and drug use did not differ from HIV-uninfected young people residing in the same setting [131].
The Western Cape has signi cant rates of stimulant use such as methamphetamine and cocaine compared to the rest of the country [207]. Approximately 5% of learners in the Western Cape have used methamphetamine within their lifetime [208]. Our ndings may indicate a general underlying substance use problem in young people in this setting, not necessarily related to HIV, but which predisposes AYLHIV to more vulnerability. High rates of substance use and implications for brain development, particularly amongst males and younger adolescents, is concerning and warrants targeted intervention. Prevention and early intervention strategies aimed at harm reduction are needed that incorporate environmental factors beyond individual behaviour.
Other NCD comorbidity Our blood pressure ndings are consistent with ndings from studies in general adolescents in urban South Africa which have reported hypertension prevalence rates ranging from 8-16% [209] and elevated BP prevalence of 35% [210]. Chatterton-Kirchmeier et al reported a signi cantly higher prevalence of elevated blood pressure in a cohort of HIV-infected, predominantly African-American adolescents and young adults in the US compared to healthy children [211]. Globally, studies involving HIV-infected adults have demonstrated higher hypertension prevalence than the general population [212] and hypertension has been found to be associated with ART [213,214]. One in ve young adults (18-35 years) attending an HIV clinic in the same setting in Khayelitsha had comorbid hypertension [65]. There is no clear link in the literature between HIV infection and elevated BP in paediatric populations. Nevertheless, routine monitoring of blood pressure in HIV care, even in younger populations, is warranted in settings like South African with a high background prevalence of hypertension, in order to avert future disease. This is especially relevant in light of evidence that blood pressure trajectories in childhood and adolescence predict future elevated BP and cardiovascular risk in adulthood [209,210,215]. None of the 25% with self-reported diabetes symptoms in our study had an abnormal measured random blood glucose. In a cohort study of South African youth living with perinatally-acquired HIV, the authors found a high prevalence of insulin resistance but it did not differ from that in uninfected age-matched adolescents [128]. A systematic review and meta-analysis of African studies, recently reported that there was no statistically signi cant association between HIV infection or ART exposure and type 2 diabetes (T2DM) prevalence in adults [216]. This is in contrast with study ndings from European and North American settings that have shown a higher prevalence of T2DM in HIV-infected adults particularly those on ART [217][218][219]. However, the cumulative incidence of T2DM in patients with HIV across Africa was higher than international incidence data for HIV-infected individuals. The International Diabetes Federation estimates that 60% of people with diabetes in Africa are undiagnosed [220] suggesting that T2DM might be a major, underdiagnosed public health problem in African populations in general due to the presence of traditional risk factors.

Strengths and Limitations
Our study adds to the limited evidence base on NCD prevalence and risk factors in AYLHIV in sub-Saharan Africa. To our knowledge, only four other studies in sub-Saharan Africa have investigated modi able NCD risk factors besides alcohol and substance use in AYLHIV [127,129,161,221]. While our study provides novel ndings for the sub-Saharan African context, the ndings should be interpreted within the following limitations. Firstly, participants were recruited using convenience sampling from healthcare facilities in a peri-urban setting. Second, there was an almost 50% non-response rate. The low response rate and lack of random sampling may limit the generalizability of our ndings, however sampling from six different facilities across all substructures in the City of Cape Town, proportional to the total numbers of AYLHIV within each substructure) mitigated unmeasured facility-speci c effects.
Limited participation and low response rates are a major challenge in adolescent research due to the complexities of enrolment and consent procedures [222,223]. Although we recruited younger adolescents, requiring parental consent may have led to participant bias as the majority enrolled were older adolescents and young adults (aged ≥ 18 years) who could provide independent consent to participate. We therefore conducted an age-strati ed analysis of participant characteristics.
Although we were not statistically powered to detect differences by gender or age groups, some clear differences emerged particularly gender differences in physical activity, obesity levels and mental health which have been previously documented. Striking differences in high-risk behaviours were also identi ed in the younger age group. Due to the cross-sectional design, we were unable to establish temporality and whether NCD risk factors and risky behaviours preceded an HIV diagnosis amongst behaviourallyinfected youth. Our study did not include a control group of HIV-uninfected adolescents, neither did we differentiate between perinatally and behaviourally-infected adolescents who may have very different risk pro les [57,119]. Nevertheless, since young people come from the same communities and access the same HIV services, irrespective of mode of transmission, interventions targeting risk factors generally may be more effective.
We used subjective recall methods of measuring physical activity which may be prone to over-reporting [224,225], recall bias and cultural misinterpretation [226,227]. However, self-report methods like the IPAQ have acceptable validity and are most widely used to measure physical activity in PLHIV [51]. The use of point-of-care random blood glucose testing may have underestimated diabetes risk, however, POC methods are better suited for community screening of diabetes, have high speci city (90%) and provide reliable and immediate results [228,229]. Similarly, the mental health tools used are screening tools-a diagnosis of depression or anxiety was not con rmed using these tools. But they are appropriate for case-nding in primary care and have been validated in HIV-positive populations in South Africa [103,230]. Despite using a self-administered questionnaire, there remains a possibility that social desirability or other reporting biases may affect reports of mental health and substance use, with potential for underreporting. To minimise this risk, we used tablet computers for data collection which have been found to reduce reporting bias on sensitive questions [231,232]. Moreover, the reported rates were still signi cant and if underreported, warrant further attention. Despite these limitations, this study is an important contribution to the limited literature in sub-Saharan Africa.

Conclusions
Page 50 /73 This paper contributes to a key gap in the literature on NCD risk in AYLHIV in SSA. The ndings highlight the existence of cardiometabolic risk factors (obesity, abdominal obesity, hypertension, physical inactivity, unhealthy diet), smoking, excessive drinking, and mental health problems in this vulnerable population, highlighting the need for NCD screening and integrated primary and secondary prevention.
NCDs and their ensuing burden of disability and premature mortality are costly to health systems and to wider societal development. Beyond primary care, the complex and interlinked social, economic and environmental factors that in uence these behaviours highlight the importance of intersectoral action for disease prevention. It is therefore necessary to go beyond the healthcare sector to address the root causes and multiple deprivations that increase the risk of NCDs and ill-health, and to support equitable access to the necessary physical and social infrastructure required to make the healthy choice the easy choice. Upstream strategies that incorporate the basic living conditions and environments in which young people live are necessary to ensure well-being and interrupt disease pathways. More studies are needed to assess risk factors at a broader socio-ecological level and explore inter-relationships between NCD/HIV comorbidity and the environment in order to identify effective and sustainable risk-reduction interventions.