Adolescent physical activity trajectories and later non-communicable disease risk factors: A systematic review of longitudinal studies Authors: Poovitha Shruthi Paramashiva, Suresh Sukumar, Dilip Shettigar, K Vaishali, Baskaran Chandrasekaran, Hari Prakash Palaniswamy, Winniecia Dkhar, Abhimanyu Pradhan, Nitika C Panakkal, Leena R David, Shivashankar Kaniyoor Nagari, Prateek Srivastav, Tulasiram Bommasamudram, Revati Amin, KS Naga Padmaja, Sneha Ravichandran, Sathya Sabina Muthu, Koustubh Kamath, Neil Abraham Barnes Published: 2026 ABSTRACT AbstractObjectivesTo systematically examine the associations between adolescent physical activity (PA) trajectories and the development of non-communicable disease (NCD) risk factors, and to identify key mediating pathways, including fat mass and inflammation.Study designSystematic review of longitudinal studies.MethodsFive electronic databases (PubMed, Embase, Scopus, Ovid Medline, and Web of Science) were searched for longitudinal studies assessing PA trajectories in adolescents (10–19 years) and subsequent NCD risk factors. Eligible studies included at least two time points of PA measurement and reported cardiovascular, metabolic, anthropometric, or inflammatory outcomes. Risk of bias was assessed using the ROBINS-E tool. Due to methodological heterogeneity, findings were synthesized narratively.ResultsTen longitudinal studies comprising 20,133 adolescent participants were included, with follow-up periods ranging from 4 to 25 years. Most studies identified declining PA trajectories across adolescence. Lower PA was consistently associated with adverse anthropometric outcomes, including higher BMI and increased obesity risk (OR = 0.50–0.67 for higher PA). Cardiovascular outcomes, particularly blood pressure, showed inverse associations with PA, while sedentary time was positively associated with risk. Metabolic outcomes indicated increased insulin resistance and impaired glucose regulation among low or declining PA groups (e.g., OR = 1.20 for hyperinsulinemia). Inflammatory markers, including C-reactive protein, were elevated with greater sedentary exposure. Fat mass emerged as a central mediator, accounting for 30–77% of the associations between PA and inflammatory outcomes. Even light-intensity PA demonstrated protective effects. Mid-adolescence (∼15 years) was identified as a critical period for trajectory-related risk.ConclusionsAdolescent PA trajectories are important determinants of later cardiometabolic health. Declining activity patterns contribute to NCD risk through complex pathways involving adiposity and inflammation. Interventions targeting sustained or increased PA, particularly during mid-adolescence, and reducing sedentary behaviour may mitigate long-term disease risk. Introduction Non-communicable diseases (NCDs) such as cardiovascular diseases, diabetes, cancer, and chronic respiratory diseases represent the leading cause of mortality and morbidity globally. Insufficient physical activity (PA) is a key risk factor for NCDs [1]. Studies demonstrate that PA provides substantial health benefits for school-aged children and youth, including improved cardiovascular and metabolic health profiles, enhanced bone health, and better cognitive outcomes [2]. Adolescence represents a crucial window for establishing long-term health behaviours and risk factor profiles [3]. They often develop modifiable risk behaviours during this period, such as smoking, unhealthy eating habits, and low PA levels, creating compounded risk for future NCD development [4,5]. A distinguishing feature of adolescent PA research is the use of trajectory modelling to capture heterogeneous, dynamic patterns of change over time, rather than relying on single cross-sectional assessments. PA has been associated with physical and mental health across the life course. Yet, few studies have used group-based trajectory modelling to examine the effects of longitudinal patterns of PA during childhood and adolescence on adult health outcomes [6]. Longitudinal research has demonstrated that PA and fitness levels during adolescence can serve as significant predictors of adult physical fitness and health status [7]. The majority of adolescents do not meet current PA guidelines. Urgent scaling up of known effective policies and programs is needed to increase adolescent activity [8]. Research examining PA trajectories has identified distinct patterns of change throughout adolescence. The development of leisure-time vigorous physical activity (LVPA) from adolescence to adulthood is heterogeneous, suggesting the need for targeted health promotion initiatives [9]. The relationship between PA during adolescence and later NCD risk factors has been established through various longitudinal studies. Regular PA is associated with improved physical fitness, cardiometabolic health, bone health, cognitive outcomes, mental health, and reduced body fat [1]. Studies suggest that changes in PA patterns during adolescence are associated with concurrent changes in cardiometabolic risk factors. Favourable blood pressure changes were found among PA increasers, and unfavourable changes in BMI, insulin, glucose, and HDL cholesterol were observed in groups with decreasing PA [10]. This evidence highlights the importance of understanding not only static levels of PA but also the dynamic patterns of change that occur throughout adolescence. Despite the growing evidence base, three important gaps remain. First, no comprehensive synthesis has evaluated the strength and consistency of associations between adolescent PA trajectories and a broad range of NCD risk factors across diverse populations and settings. Second, it remains unclear which trajectory characteristics, such as timing, magnitude, or direction of change, are most consequential for downstream disease risk. Third, existing literature does not clearly delineate which NCD risk factor domains (anthropometric, cardiovascular, metabolic, or inflammatory) are most sensitive to adolescent activity trajectory patterns, nor during which developmental windows these associations are most pronounced. This systematic review addresses these gaps by synthesizing evidence from longitudinal studies examining the association between PA trajectories during adolescence and the development of NCD risk factors. Methods Search strategy A comprehensive systematic search was conducted across multiple electronic databases to identify relevant studies. Based on established methodological approaches for systematic reviews examining PA in adolescents, five electronic databases were searched: PubMed, Embase, Scopus, Ovid Medline, and Web of Science. The search strategy was developed in consultation with a librarian and incorporated both Medical Subject Headings (MeSH) terms and free-text keywords. The search terms were grouped into three main concepts. The first concept focused on the population of interest and included terms such as "adolescent∗", "teenage∗", "youth", and "young adult". The second concept addressed the exposure of interest, incorporating terms related to PA such as "physical activity", "exercise", and "fitness". The third concept encompassed outcomes related to NCD risk factors, including "cardiovascular disease", "diabetes", "metabolic syndrome", "obesity", and specific risk factor terms such as "blood pressure", "cholesterol", "insulin resistance", "BMI", "body composition", and "cardiometabolic risk". The complete search strategy is provided in the Supplementary Material. The search strategy was adapted for each database to account for differences in indexing systems and search functionality. No language restrictions were applied initially, though non-English studies were evaluated during the screening process based on translation availability. To minimize publication bias and ensure comprehensive coverage, supplementary searches for grey literature were conducted. Reference lists of all included studies and relevant systematic reviews were manually examined to identify additional eligible studies. Forward citation searching was performed using Google Scholar and Web of Science Citation Index. Trial registries were searched to identify completed but unpublished longitudinal studies; the WHO International Clinical Trials Registry Platform (ICTRP; www.who.int/ictrp) and ClinicalTrials.gov (www.clinicaltrials.gov) were searched. All searches were conducted in August 2025. Eligibility criteria Studies were included if they met all of the specified inclusion criteria. Regarding the population, studies needed to include adolescents aged 10-19 years at baseline as defined by the World Health Organization, though studies could include participants outside this age range if the majority of baseline assessments occurred during adolescence. For the exposure component, studies required longitudinal measurement of PA patterns or trajectories during adolescence, utilizing trajectory modelling approaches such as group-based trajectory modelling, latent class growth analysis, or growth mixture modelling, or examining changes in PA over at least two time points. PA can be measured using objective methods, such as accelerometry or pedometry, or validated subjective methods, such as questionnaires or interviews. Regarding outcomes, studies needed to measure NCD risk factors in adolescence, young adulthood, or later life. These risk factors included cardiovascular risk factors such as blood pressure and lipid profiles; metabolic risk factors, including glucose, insulin, and metabolic syndrome; anthropometric measures such as body mass index, waist circumference, and body composition; and inflammatory markers. Studies were required to be longitudinal in design, including cohort and panel studies, with a minimum follow-up of 1 year and at least 2 measurement time points. Acceptable publication types included peer-reviewed journal articles and conference abstracts with sufficient methodological detail. Studies were excluded if they were cross-sectional in design, focused exclusively on clinical populations such as those with diagnosed diabetes or cardiovascular disease, examined only acute effects of PA interventions, did not report original data such as editorials, commentaries, or narrative reviews, or were duplicate publications of the same cohort and outcomes, in which case the most comprehensive publication was retained. Study selection process The study selection process was conducted in two distinct phases. In the first phase, two reviewers independently screened all identified titles and abstracts against the established eligibility criteria. The second phase involved independent assessment of full-text articles of potentially eligible studies by two reviewers. Reasons for exclusion at the full-text stage were systematically recorded. Disagreements at both stages were resolved through discussion, and a third reviewer was consulted when consensus could not be reached. All screening activities were conducted using Rayyan software (https://www.rayyan.ai/) to facilitate collaboration between reviewers and maintain a comprehensive audit trail of decision-making processes. Data extraction Data extraction was performed independently by two reviewers using a standardized, pilot-tested extraction form developed specifically for this review. The extraction form captured comprehensive information about study characteristics, including author details, publication year, country of origin, study design, sample size, demographic characteristics, follow-up duration, and assessment time points. Participant characteristics were systematically extracted, including age at baseline and follow-up assessments, sex and gender distribution, socioeconomic status indicators, and ethnicity or race, as reported in the original studies. Detailed information was extracted regarding PA assessment methods, including whether objective or subjective measurement approaches were used, the specific instruments employed, the PA domains assessed, the trajectory modelling approaches used where applicable, and the number and characteristics of identified trajectory groups. For NCD risk factor assessment, data extraction included the specific risk factors measured, the measurement methods and assessment timing, and the laboratory procedures and standardization protocols employed. Results data were comprehensively extracted, including main findings regarding associations between PA trajectories and NCD risk factors; effect sizes and confidence intervals, where available; adjustment for confounding variables; and any subgroup analyses conducted. Disagreements in data extraction between reviewers were resolved through discussion, with arbitration by a third reviewer when consensus could not be achieved. Risk of bias assessment The risk of bias in included studies was assessed independently by two reviewers using the Risk of Bias in Non-randomized Studies of Exposures (ROBINS-E) tool, which is specifically designed to evaluate observational studies examining the effects of exposures [11]. The ROBINS-E tool evaluates seven domains of bias across the study process. Each domain was rated as having low risk, some concerns, or high risk of bias using predefined signalling questions and criteria. Studies were assigned an overall risk of bias rating corresponding to the highest risk rating in any individual domain, following the structured approach recommended by the ROBINS-E guidance. Two reviewers independently assessed the risk of bias for each included study, with disagreements resolved through discussion and consultation with a third reviewer when consensus could not be reached. The risk of bias assessment results were summarized and presented visually using traffic light plots generated with the robvis (Risk of Bias Visualization) tool in R statistical software, allowing for a clear presentation of bias risk across studies and domains [12]. Narrative synthesis All included studies underwent qualitative synthesis, with results systematically organized according to the types of PA trajectory patterns identified, the specific NCD risk factors examined, the direction and strength of reported associations, and the population characteristics and study settings. Narrative synthesis, rather than meta-analysis, was chosen to present the results due to substantial clinical and methodological heterogeneity across the included studies. PA was measured using markedly different methods across studies, including accelerometry and self-report questionnaires, which were not directly comparable in the activity constructs they captured. The trajectory modelling approaches also varied considerably across studies, making the exposure definitions fundamentally incomparable. Furthermore, the outcome measures differed substantially in terms of the specific NCD risk factors assessed, the instruments and cut-offs used, and the timing of outcome assessment relative to the exposure period, with follow-up durations ranging from 4 to 25 years. Finally, effect measures were heterogeneous across studies, including beta coefficients, odds ratios, and composite cardiometabolic risk scores, and could not be meaningfully pooled into a common metric. Results were therefore organized by outcome domain and synthesized narratively to allow transparent comparison across studies. Results Study selection The systematic search across five electronic databases initially identified 11,080 records. After removing 4230 duplicates, 6851 records underwent title and abstract screening. Following this initial screening, 6836 records were excluded for not meeting the eligibility criteria, leaving 14 full-text articles for detailed assessment. During the full-text review, 4 studies were excluded. Ultimately, 10 longitudinal studies met all inclusion criteria and were included in the qualitative synthesis. The study selection process is summarized in Fig. 1. PRISMA flow chart. Study characteristics The 10 included studies were published between 2006 and 2025, with sample sizes ranging from 116 to 9155 participants. The total number of participants across all studies was 20,133. Studies were conducted in various countries, including the United Kingdom (n = 4), Norway (n = 2), Finland (n = 2), the USA (n = 1), and the Netherlands (n = 1). Follow-up periods ranged from 4 to 25 years, with most studies following participants from adolescence into young adulthood. The median baseline age ranged from 11 to 16 years across studies. PA was measured using diverse approaches across the included studies. Objective measurements using accelerometry were employed in six studies, while four studies used self-reported questionnaires or validated PA instruments. Several studies specifically examined sedentary time, light PA, and moderate-to-vigorous PA as distinct behavioural patterns. Three studies explicitly used trajectory modelling approaches, including group-based trajectory modelling and latent class growth analysis. The studies examined various NCD risk factors as outcomes, including anthropometric measures (body mass index, waist circumference, fat mass index), cardiovascular risk factors (blood pressure, cholesterol levels), metabolic markers (glucose, insulin, HOMA-IR), inflammatory markers (high-sensitivity C-reactive protein), and composite measures such as clustered cardiometabolic risk scores. Study characteristics are detailed in Table 1. Table 1Characteristics of the studies included in the review.Table 1Author, year, countryStudy designStudy objectivesPopulation characteristicsFollow-up durationPA measurementOutcome measuredMain findingsViner et al. [13], 2006, United KingdomLOTo examine whether factors common to obesity prevention programmes in adolescence (exercise, sedentary activities, eating behaviours, psychological factors) predict change in BMI in 'free-living' adolescents followed into adulthood.Sample size: 4461.2946 female subjects.Age: 16 years at baseline.8.2% obese at baseline.14 yearsExercise (frequency of playing sport). Sedentary activities (hours of sedentary activity). Eating behaviours (eating takeaway meals, consuming carbonated drinks). Psychological factors (dieting history, psychological distress, self-esteem)Change in BMI z-score (zBMI) between 16 and 30 yearsLoss of BMI between adolescence and adulthood is predicted by female sex (B = 0.07, p = 0.01), higher social class (B = 0.16, P < 0.0001), and higher frequency of playing sport (B = 0.07, p = 0.05).Increase in BMI is predicted by sedentary activities (B = 0.15, p = 0.01), frequent takeaway meals (B = 0.21, p = 0.009), consumption of carbonated drinks (B = 0.13, p = 0.04), and a history of dieting (B = 0.09, p = 0.04).Reducing sedentary behaviours and unhealthy eating patterns is crucial for preventing obesity persistence from childhood to adulthood.Boone et al.[14], 2007, USALOAssess the joint impact of sedentary behaviour and physical activity on obesitySample size: 9155.4276 female subjects.Age: 15.9 ± 0.12 years at baseline.8.2% obese at baseline.Obesity prevalence: 11.5% males, 10.9% females at Wave II; 21.1% males, 23.9% females at baseline. Lower household income and lower parental education associated with higher obesity prevalence. Race and ethnicity varied among females.6 yearsMVPA (moderate to vigorous physical activity) and screen time through self-report questionnaires.Obesity prevalence and incidence (BMI ≥30 kg/m 2)In males, higher MVPA was associated with lower odds of obesity [OR = 0.50 (0.40, 0.62)], while in females, both higher MVPA [OR = 0.67 (0.49, 0.91)] and lower screen time [OR = 0.67 (0.53, 0.85)] were associated with lower odds of obesity. Screen time during adolescence had a stronger influence on obesity incidence in females [OR = 0.58 (0.43, 0.80)] compared to males OR = 0.78 (0.61, 0.99)]. Reducing screen time during adolescence and into adulthood is a promising strategy for reducing obesity incidence, particularly in females.Lagestad et al.[15], 2018, NorwayLOThe aim of this study was to provide new evidence of longitudinal changes in PA level, body mass index, and oxygen uptake among adolescents.Sample size: 116.Boys: 61.Girls: 55Norwegian adolescents aged 14 years at baseline. From both urban and rural areas. Students.5 yearsSelf-reported and accelerometer physical activity level (minutes of moderate and/or vigorous PA per day).BMI and oxygen uptakeThe study found a significant decrease in physical activity levels among adolescents from age 14 to 19, as measured by accelerometers (F = 4.922, p = 0.004). There was a significant increase in BMI and a decrease in oxygen uptake over the study period (F ≥ 26.8, p < 0.001). The proportion of boys who reported being physically active during school breaks decreased substantially from 61% at age 14 to 11% at age 19.Kemper et al.[16], 2019, NetherlandsLOTo detect changes in health and lifestyle over time during the teenage and young adult periods. To determine the longitudinal relationship between daily physical activity, daily food intake, and health outcomes such as aerobic fitness, bone mineral density, and body fatness.Sample size: 617Age: 12 years at baseline.Recruited from two secondary schools in Amsterdam and Purmerend, Netherlands.25 yearsDaily physical activity (DPA) measured by heart rate integrators and converted to metabolic (MET) scores.Daily food intake (DFI) measured by a cross-check dietary historyInterview.aerobic fitness (VO 2 peak), bone mineral density (BMD), obesity (BMI), body fatness (S4S)Aerobic fitness increased more significantly in physically active individuals compared to those who were less active (β = 0.07, p < 0.01). Physical activity with high mechanical load increased bone formation in the hip (11%) and lumbar (15%) region. More daily physical activity resulted in lower fat mass and less overweight and obesity, with no significant relationship with dietary intake.Le et al.[17], 2023, Finland, DenmarkLOInvestigate directional influences in the association between adiposity and physical activity from pre-puberty to early adulthood.Participants from two longitudinal cohorts.Calex-study: 396 Finnish girls aged 11.2 years at baseline.European Youth Heart Study (EYHS): 399 Danish boys and girls. aged 9.6 years at baseline.Calex-study: approximately 7.1 years. EYHS: approximately 12.2 yearsLeisure-time physical activity (LTPA) measured using modified version of a WHO validated self-administrated physical activity questionnaire.Habitual physical activity (PA) using accelerometer.Body Mass Index (BMI). Fat Mass Index (FMI).Adiposity measures (BMI and FMI) were associated with physical activity levels at different ages, but these associations were not consistent across all ages or genders. The study found no clear bidirectional associations between adiposity and physical activity, indicating that while adiposity may predict future physical activity levels, the relationship is not strongly reciprocal.Aira et al.[18], 2023, FinlandLOTo examine the associations between longitudinal physical activity (PA) patterns and the development of cardiometabolic risk factors from adolescence to young adulthood.Total: 250.60% female.Mean age 15 at baseline. Majority from families with high affluence (68%). Mostly engaged in education at follow-up (64% studying, 20% working). More active than average Finnish young people.4 yearsPhysical activity patterns measured using a Hookie accelerometer.Glucose (mmol/L). Insulin (mU/L). Homeostasis model assessment for insulin resistance (HOMA-IR). Total cholesterol (mmol/L). HDL cholesterol (mmol/L). LDL cholesterol (mmol/L). Triglycerides (mmol/L). Blood pressure (mm Hg). Body mass index (kg/m2)Decreasing physical activity from moderate to low levels is associated with unfavourable changes in insulin (β = 0.23, p < 0.05) and BMI (β = 0.90, p < 0.05). Decreasing physical activity from high to moderate levels is linked to unfavourable changes in glucose (β = 0.18, p < 0.05) and HDL cholesterol (β = 0.18, p < 0.01). Increasing physical activity is associated with favourable changes in systolic (β = −6.43, p < 0.05) and diastolic (β = −6.72, p < 0.01) blood pressure.Agbaje et al.[19], 2023, United KingdomLOExamine the mediating role of fat mass, lipids, and insulin resistance on the associations of cumulative sedentary time (ST), light physical activity (LPA), and moderate-to-vigorous physical activity (MVPA) with inflammation.Total: 792, 58% female.Children aged 11 years at baseline.13 yearsSedentary time (ST), light physical activity (LPA), moderate-to-vigorous physical activity (MVPA) measured using ActiGraph accelerometer.Inflammation measured by high-sensitivity C-reactive protein (hsCRP).ST increased over the 13-year follow-up and was associated with increased inflammation. LPA was associated with decreased inflammation, with fat mass mediating this relationship by 30%. MVPA was associated with decreased inflammation, but fat mass mediated this relationship by 77%, indicating a stronger attenuation effect compared to LPA.Agbaje et al.[20], 2024, United KingdomLOExamined the longitudinal associations of cumulative accelerometer-measured ST, LPA, and MVPA with repeated measures of systolic and diastolic. Assessed the extent to which the longitudinal associations of movement behaviours with BP are mediated by total fat mass, lean mass, insulin resistance, lipid metabolism, and inflammation.Total = 2513Children aged approximately 11.72 years at the start of the study. Majority female (61%).13 yearsSedentary time (ST), light physical activity (LPA), moderate-to-vigorous physical activity (MVPA) measured using ActiGraph accelerometer.Systolic blood pressure (mmHg), diastolic blood pressure (mmHg)ST is positively associated with increased systolic and diastolic blood pressure in children from ages 11 to 24 (β = 0.009, p < 0.001). LPA is inversely associated with systolic blood pressure (β = −0.007, p < 0.001), suggesting it can help lower systolic BP. Replacing sedentary time with LPA can significantly lower blood pressure, with a theoretical reduction of −3 mmHg in systolic BP and −2 mmHg in diastolic BP.Agbaje et al.[21], 2025, United KingdomLOExamine the effects of cumulative ST, LPA, and MVPA on glucose, insulin, and HOMA-IR. Assess the extent to which the associations of movement behaviours with glucose, insulin, and HOMA-IR are mediated by fat mass, muscle mass, lipid, and inflammation. Examine the temporal interrelations among activity levels and glucose, insulin, and HOMA-IR in the total cohort and participants with overweight/obesity.Total: 792455 females.Children from the Avon Longitudinal Study of Parents and Children in the UK.Age = 11 years at baseline.13 yearsSedentary time (ST), light physical activity (LPA), moderate-to-vigorous physical activity (MVPA) measured using ActiGraph accelerometer.Glucose, insulin, and homeostatic model assessment for insulin resistance (HOMA-IR).Cumulative ST from ages 11 to 24 years increased the risk of hyperinsulinemia in participants with overweight/obesity (OR = 1.20, 95% CI = 1.00-1.44, p = 0.047). Cumulative LPA (OR = 0.80, CI = 0.66-0.96, p = 0.017) was associated with a decreased risk of hyperinsulinemia, suggesting a protective effect against metabolic disorders. Promoting LPA while decreasing body fat mass and sedentary time may be crucial intervention targets to attenuate the risk of hyperinsulinemia and insulin resistance.Husøy et al.[22], 2025, NorwayLOExamine the prospective associations of sedentary time and physical activity at age 15 with cardiometabolic risk markers at age 24Total: 64247% females.Age = 15 years at baseline.9 yearsSedentary time (ST), moderate-to-vigorous physical activity (MVPA), vigorous physical activity (VPA) measured using ActiGraph accelerometers.Body mass index (BMI) (kg/m2), waist circumference (cm), visceral fat (g), maximal oxygen uptake (VO2max), systolic blood pressure (mm Hg), LDL-cholesterol (mmol/L), insulin (pmol/L), high-sensitivity C-reactive protein (mg/L) and Clustered risk Z-score.Sedentary time at age 15 was inversely associated with VO2max at age 24 (β = −1.6 mL/kg/min,95% CI, = −2.8 to −0.5, p = 0.005). MVPA (β = 1.6 mL/kg/min, 95% CI = 0.8 to 2.4, p < 0.001) and VPA (β = 1.6 mL/kg/min, 95% CI = 0.8 to 2.4, p < 0.001) at age 15 were positively associated with VO2max at age 24. VPA at age 15 was inversely associated with visceral fat mass (β = −41g, 95% CI = −78 to −3, p = 0.034), insulin levels (β = −4.3 pmol/L, 95% CI = −8.2 to −0.4, p = 0.027), and clustered cardiometabolic risk in young adulthood (β = −0.09, 95% CI = −0.18 to −0.01, p = 0.034). Characteristics of the studies included in the review. Risk of bias assessment Risk of bias assessment using the ROBINS-E tool revealed considerable variability in methodological quality across the included studies. Two studies [13,23] demonstrated a low overall risk of bias across all seven domains. Four studies [[14], [15], [16], [17]] raised concerns overall, while four studies[10,[18], [19], [20]] were rated as having a high overall risk of bias. The traffic plot of the risk-of-bias assessment results is shown in Fig. 2. Risk of bias assessment using the ROBINS-E tool. Synthesis of findings Physical activity trajectory patterns The included studies identified several distinct PA trajectory patterns during adolescence. The most commonly observed pattern was a general decline in PA levels from early to late adolescence, reported in eight of the ten studies. Viner et al. [14] found that the frequency of playing sports decreased over the 14-year follow-up period, while Lagestad et al. [16] demonstrated significant decreases in accelerometer-measured PA from age 14 to 19 years (F = 4.922, p = 0.004). Studies employing formal trajectory modelling identified three to four distinct trajectory groups. Aira et al. [10] characterized participants into high, moderate, and low PA trajectory groups, with decreasing patterns being most prevalent. Similarly, studies examining specific activity intensities revealed heterogeneous patterns, with sedentary time generally increasing and moderate-to-vigorous PA decreasing throughout adolescence. Associations with anthropometric outcomes Seven studies examined the relationship between PA trajectories and body composition measures. Consistent inverse associations were observed between PA levels and the development of body mass index. Viner et al. [14] reported that a higher frequency of playing sports predicted lower BMI z-score changes (β = 0.07, p = 0.05), while increased sedentary activities predicted BMI increases (β = 0.15, p = 0.01). Boone et al. [15] demonstrated sex-specific associations, with higher moderate-to-vigorous PA associated with lower odds of obesity in both males (OR = 0.50, 95% CI: 0.40-0.62) and females (OR = 0.67, 95% CI: 0.49-0.91). The protective effect of PA against obesity development was particularly pronounced when examining screen time reduction, with stronger associations observed in females compared to males. Kemper et al. [18] found that individuals with higher daily PA levels showed significantly lower fat mass and reduced risk of overweight and obesity over the 25-year follow-up period. The study demonstrated dose-response relationships, with greater PA levels associated with progressively better body composition outcomes. Cardiovascular risk factor development Six studies investigated associations between PA trajectories and cardiovascular risk factors. Blood pressure outcomes showed consistent inverse relationships with PA levels. Aira et al. [10] reported that increasing PA was associated with favourable changes in both systolic (β = −6.43, p < 0.05) and diastolic (β = −6.72, p < 0.01) blood pressure over the 4-year follow-up period. Agbaje et al. [13] demonstrated that cumulative sedentary time from ages 11 to 24 was positively associated with increased systolic and diastolic blood pressure (β = 0.009, p < 0.001). At the same time, light PA showed an inverse association with systolic blood pressure (β = −0.007, p < 0.001). The study estimated that replacing sedentary time with light PA could theoretically reduce systolic blood pressure by 3 mmHg and diastolic blood pressure by 2 mmHg. Husøy et al. [20] found that vigorous PA at age 15 was inversely associated with clustered cardiometabolic risk in young adulthood (β = −0.09, 95% CI: −0.18 to −0.01, p = 0.034), while sedentary time showed inverse associations with maximal oxygen uptake (β = −1.6 mL/kg/min, 95% CI: −2.8 to −0.5, p = 0.005). Metabolic risk factor outcomes Five studies examined metabolic risk factors, including glucose, insulin, and markers of insulin resistance. Aira et al. [10] demonstrated that decreasing PA from moderate to low levels was associated with unfavourable changes in insulin (β = 0.23, p < 0.05) and glucose levels (β = 0.18, p < 0.05). Conversely, decreasing activity from high to moderate levels was associated with unfavourable changes in HDL cholesterol (β = 0.18, p < 0.01). Agbaje et al. [23] found that cumulative sedentary time from ages 11 to 24 increased the risk of hyperinsulinemia in participants with overweight/obesity (OR = 1.20, 95% CI: 1.00-1.44, p = 0.047). Light PA showed protective effects against hyperinsulinemia (OR = 0.80, 95% CI: 0.66-0.96, p = 0.017), suggesting potential metabolic benefits of even low-intensity movement behaviours. Husøy et al. [20] reported that vigorous PA at age 15 was inversely associated with visceral fat mass (β = −41g, 95% CI: −78 to −3, p = 0.034) and insulin levels (β = −4.3 pmol/L, 95% CI: −8.2 to −0.4, p = 0.027) measured nine years later in young adulthood. Inflammatory markers Two studies specifically examined inflammatory outcomes. Agbaje et al. [17] investigated high-sensitivity C-reactive protein levels and found that sedentary time increased over the 13-year follow-up and was associated with increased inflammation. Light PA was associated with decreased inflammation, with fat mass mediating 30% of this relationship. Moderate-to-vigorous PA also showed anti-inflammatory effects, but fat mass accounted for 77% of this association, indicating stronger attenuation than with light PA. The studies suggest that the anti-inflammatory effects of PA may be partially mediated by improvements in body composition, particularly reductions in fat mass during adolescence. Sex-specific and mediating factors Several studies reported sex-specific associations between PA trajectories and NCD risk factors. Boone et al. [15] found that reduced screen time had stronger protective effects against obesity in females than in males. Viner et al. [14] reported that female sex was a significant predictor of BMI z-score increases (β = 0.07, p = 0.01), suggesting potential sex differences in the relationship between activity behaviours and weight gain. Multiple studies identified body composition, particularly fat mass, as an important mediating factor in the relationship between PA and various health outcomes. Agbaje et al. 13,17,23, 13,17 demonstrated substantial mediation effects of fat mass on the associations between PA and both inflammatory markers and metabolic outcomes, highlighting the interconnected nature of movement behaviours, body composition, and cardiometabolic health. Critical periods and trajectory characteristics The included studies provided insights into critical periods for intervention. Husøy et al. [20] demonstrated that PA patterns established at age 15 had significant associations with cardiometabolic risk factors measured nearly a decade later, suggesting that mid-adolescence may represent a particularly important intervention window. Studies with longer follow-up periods (>10 years) consistently showed that PA trajectories established during adolescence had persistent effects on the development of NCD risk factors into young adulthood. The magnitude of associations appeared to be influenced by both the timing and intensity of changes in PA, with earlier and more substantial decreases in activity associated with stronger relationships with adverse health outcomes. Discussion This systematic review synthesized evidence from ten longitudinal studies examining the associations between adolescent PA trajectories and later NCD risk factors. The findings provide compelling evidence that PA patterns established during adolescence have significant and persistent effects on cardiometabolic health outcomes extending into young adulthood. Our results align with and extend previous systematic reviews documenting the decline in PA during adolescence and its health implications. The consistent decline in PA levels from early to late adolescence observed across eight of our included studies[10,[13], [14], [15], [16], [17],20,23] corroborates findings from Dumith et al.'s landmark systematic review and pooled analysis, which demonstrated an average annual decline of 7% in PA during adolescence [21]. This decline pattern has been consistently reported globally, with recent surveillance data showing that the majority of adolescents do not meet current PA guidelines, requiring the urgent scaling up of known effective policies and programs [8]. Our identification of heterogeneous trajectory patterns and distinct subgroups builds on the emerging body of trajectory modelling research. A 2019 systematic review by Lounassalo et al. [22] found that researchers have increasingly applied trajectory approaches to identify homogeneous subgroups of PA in heterogeneous populations across different life phases. However, few studies have used group-based trajectory modelling to examine the effects of longitudinal patterns of PA during childhood and adolescence on adult health outcomes [6], making our synthesis particularly valuable for understanding the long-term health implications. Trajectory patterns and health outcomes A central conceptual contribution of this review lies in its focus on PA trajectories rather than single-timepoint or simple cross-sectional assessments of PA during adolescence. Studies linking PA measured at one point in time to later NCD risk are informative but inherently limited; they capture whether being active or inactive at a particular moment predicts future disease, yet they cannot account for how an individual arrived at that level, the direction of subsequent change, or the cumulative pattern of exposure across the adolescent years. Traditional longitudinal analyses have similarly been criticized for examining population-level changes over time without accounting for the subgroups of individuals who may follow distinct patterns of behaviour, a limitation that trajectory-based approaches are specifically designed to address [6,22]. Latent class analysis and other trajectory modelling approaches allow researchers to summarise complex patterns of behaviour across the life course, rather than relying on single data points, enabling the overall developmental course of behaviour to be examined and used to predict later health outcomes [6]. This distinction matters for NCD risk for several reasons. First, cumulative exposure to physical inactivity across the adolescent years rather than inactivity at any single assessment is likely the more biologically relevant driver of cardiometabolic risk factor development, given that processes such as adiposity accrual, insulin resistance, and vascular remodelling are time-dependent [24,25]. Evidence from life-course modelling supports a cumulative exposure hypothesis, in which PA at multiple points in development affects later health outcomes additively rather than through a single critical or sensitive period [26]. Second, trajectory approaches can identify when, during adolescence, activity change is most consequential, enabling the detection of critical intervention windows that single-timepoint studies cannot reveal. Group-based trajectory modelling is particularly valuable for examining the timing and magnitude of PA decline, overcoming the limitations of studies that use only two time points and cannot characterize the shape or progression of change [27]. Third, by identifying distinct subgroups such as consistently low-active, steeply declining, and persistently active adolescents, trajectory modelling enables targeted identification of the highest-risk individuals for preventive intervention, rather than treating adolescents as a homogeneous population [9,22]. The findings of this review, demonstrating that declining trajectories through mid-adolescence are most strongly associated with cardiometabolic risk factors measured up to a decade later, exemplify the kind of actionable, temporally specific evidence that trajectory modelling uniquely affords. The associations between declining PA trajectories and adverse changes in anthropometric, cardiovascular, metabolic, and inflammatory markers identified in our review extend beyond previous cross-sectional or short-term longitudinal studies. While systematic reviews have established that PA provides health benefits for school-aged children and youth [2], our findings demonstrate that the temporal patterns of activity change are equally important as absolute activity levels. The sex-specific associations observed in our review, particularly the stronger protective effects of PA against obesity in females [15], align with previous research documenting gender differences in PA correlates and outcomes. Systematic reviews have identified gender, age, socioeconomic status, and parental and peer influences as among the most-researched correlates of PA in youth [28]. Still, our findings suggest that intervention approaches may need to be tailored differently for males and females to optimize effectiveness. Critical periods and intervention timing Our identification of mid-adolescence as a critical intervention period, particularly around age 15, is supported by developmental research emphasizing adolescence as a period of neurobiological and behavioural plasticity [29]. This finding is consistent with trajectory research in other health behaviours, which has identified similar critical periods during adolescence when intervention may have a particularly lasting impact [30]. The persistence of associations between adolescent activity patterns and young adult health outcomes observed in studies with longer follow-up periods supports the concept of PA "tracking" from adolescence to adulthood. This tracking phenomenon has been documented in previous longitudinal research. However, the development of leisure-time PA from adolescence to adulthood is heterogeneous, with the largest trajectory group characterized by low levels throughout the life course [9]. Mechanisms and mediating pathways The identification of body composition, particularly fat mass, as a critical mediating pathway aligns with existing research on the mechanisms linking PA to health outcomes. Previous systematic reviews of longitudinal studies have documented long-term relationships between PA and weight gain, obesity, coronary heart disease, and type 2 diabetes mellitus [31] However, most have focused on adult populations rather than trajectories established during adolescence. The finding that even light PA showed protective effects is particularly noteworthy given current PA guidelines that emphasize moderate-to-vigorous activity. This supports emerging research suggesting that any movement above sedentary behaviour may yield health benefits, potentially making interventions more achievable for adolescents with low baseline activity levels [32]. Strengths, limitations, and future directions This systematic review has several important strengths relative to previous reviews in this area. The comprehensive search strategy and focus on longitudinal studies with trajectory-based approaches provided insights into dynamic patterns of activity change rather than static assessments. The use of the ROBINS-E tool for bias assessment provided a rigorous evaluation of study quality, specifically designed for observational studies examining exposures. However, limitations should be acknowledged. The inability to conduct meta-analysis due to substantial clinical and methodological heterogeneity limits the precision of conclusions that can be drawn from this review. The likelihood of publication bias cannot be fully excluded. At the same time, comprehensive search methods were employed to minimize this risk; a formal statistical assessment was not feasible given the use of narrative synthesis and the small number of included studies. Future research should prioritize sophisticated trajectory modelling to characterize heterogeneity in adolescent PA patterns better and identify predictors of distinct trajectory types. Understanding factors that distinguish adolescents following different activity trajectories could inform targeted intervention development. Conclusion This systematic review provides compelling evidence that PA trajectories established during adolescence have significant and persistent associations with the development of NCD risk factors. Our findings build upon and extend previous research by demonstrating that the temporal patterns of activity change during adolescence are critical determinants of long-term health outcomes. The consistent finding of declining PA during adolescence, coupled with evidence for long-term health implications of these patterns, reinforces the critical importance of this developmental period for lifelong health promotion. The identification of distinct trajectory patterns and critical periods around mid-adolescence provides new insights that could inform more effective, targeted intervention approaches. Ethical statement Ethical approval was not required for this study, as it was a review paper. Protocol registration and reporting guidelines This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [33]. The review protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) before commencing data extraction (Registration number: CRD420251131034). Authorship Conceptualization: VK; Data curation: DS; Funding acquisition: VK; Methodology: SS, DS, PP, VK; Supervision: BC, HP, LD; Validation: SK, PS, TB, RR; Visualization: WD, AP, NP; Writing – original draft: DS, SS, VK; Writing – review & editing: SR, SM, KK, NB. Funding This research was supported by the Indian Council of Medical Research (ICMR), Delivery Research Division under Intermediate Grants (Project ID: IIRPIG-2024-01-01648). Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.puhip.2026.100811. Contributor Information Poovitha Shruthi Paramashiva, Email: poovitha.sp@manipal.edu. Suresh Sukumar, Email: suresh.sugumar@manipal.edu. Dilip Shettigar, Email: dilip.mchpmpl2023@learner.manipal.edu. K. Vaishali, Email: vaishali.kh@manipal.edu. Baskaran Chandrasekaran, Email: baskaran.chandrasekaran@ssss.edu.in. Hari Prakash Palaniswamy, Email: hari.prakash@manipal.edu. Winniecia Dkhar, Email: winniecia.dkhar@manipal.edu. Abhimanyu Pradhan, Email: abhimanyu.pradhan@manipal.edu. Nitika C. Panakkal, Email: nitika.cp@manipal.edu. Leena R. David, Email: ldavid@sharjah.ac.ae. Shivashankar Kaniyoor Nagari, Email: shi.sha@manipal.edu. Prateek Srivastav, Email: prateek.srivastav@manipal.edu. Tulasiram Bommasamudram, Email: tulasiram.b@manipal.edu. Revati Amin, Email: revati.a@manipal.edu. K.S. Naga Padmaja, Email: naga.mchpmpl2023@learner.manipal.edu. Sneha Ravichandran, Email: sneha.ravichandran@learner.manipal.edu. Sathya Sabina Muthu, Email: sathya.mchpmpl2023@learner.manipal.edu. Koustubh Kamath, Email: koustubh.mchpmpl2024@learner.manipal.edu. Neil Abraham Barnes, Email: neil.mchpmpl2023@learner.manipal.edu. Supplementary data The following are the Supplementary data to this article: Multimedia component 1 Multimedia component 2 References Associated Data Supplementary Materials Multimedia component 1 Multimedia component 2