Recreational physical activity and biomarkers of breast cancer risk in a cohort of adolescent girls

Abstract

Background

Evidence suggests that recreational physical activity (RPA) during adolescence is associated with lower breast cancer (BC) risk, but the biological mechanisms underlying this relationship remain unclear. We examined the associations between RPA and three BC biomarkers–breast tissue composition (BTC), oxidative stress, and inflammation–in adolescent girls. We also investigated how oxidative stress and inflammation relate to BTC during this developmental stage.

Methods

We analyzed data from a population-based urban cohort of 191 Black/African American and Hispanic (Dominican) adolescent girls (ages 11–20 years). Participants reported the number of hours they participated in organized and unorganized RPA in the past week, categorized as none, < 2 h, and ≥ 2 h. We used optical spectroscopy to measure BTC: specifically, chromophores that are positively (percent water content and percent collagen content) or negatively (percent lipid content) correlated with mammographic breast density. We measured urinary concentrations of 15-F_2t-isoprostane, a sensitive and specific marker of oxidative stress, and blood biomarkers of inflammation, including TNF-α, IL-6, and high-sensitivity C-reactive protein. We used multivariable linear regression models to examine associations, adjusting for age, race, ethnicity, and percent body fat.

Results

51% of adolescent girls reported no past-week engagement in any type of RPA; 73% reported no participation in organized activities, and 66% reported no participation in unorganized activities. Girls who engaged in ≥ 2 h of organized RPA in the past week, compared to none, had lower percent water content in the breast (β: − 0.41, 95% CI: − 0.77, − 0.05), and this association was not modified by percent body fat; they also had lower urinary concentrations of 15-F_2t-isoprostane (β: − 0.50, 95% CI: − 0.95, − 0.05). An association was observed between higher urinary concentrations of 15-F_2t-isoprostane and higher percent collagen content in the breast (β: 0.15, 95% CI: 0.00, 0.31). RPA was not associated with the measured inflammatory biomarkers, and these biomarkers were not associated with BTC after adjusting for percent body fat.

Conclusions

These findings support that RPA is associated with BTC and oxidative stress in adolescent girls, independent of body fat. Additional longitudinal research is needed to understand the implications of these findings regarding subsequent BC risk.

Background

Recreational physical activity (RPA) in adulthood is associated with a lower risk of breast cancer (BC), with studies consistently reporting about a 20% reduction in risk when comparing women in the highest to the lowest category of RPA within a study sample [1–3]. While most studies have focused on RPA during adulthood, there is emerging evidence that RPA during earlier life periods may also be important for reducing BC risk, particularly risk of early-onset disease [4, 5]. In a recent study, we found that women in the highest versus lowest quartile of RPA during adolescence had a 12% (95% CI:2–22%) lower BC risk before age 40; this risk reduction increased to 22% (95% CI:11–32%) among women who were also in the highest quartile of RPA during early adulthood [5]. While these findings support the importance of physical activity across the life course, including during adolescence, for reducing BC risk, the biological mechanisms underlying these associations remain poorly understood. Elucidating the relationship between RPA and biomarkers of BC risk during adolescence is necessary for clarifying potential causal pathways. Further, understanding whether RPA is related to breast tissue composition (BTC) and other biomarkers is critical for guiding future prevention intervention strategies. Such research is especially urgent given the rising incidence of BC among younger women in the U.S. and globally [6–8].

One potential mechanism through which RPA during adolescence may influence subsequent BC risk is by shaping BTC. Having mammographically dense breasts, a common measure of BTC reflecting the amount of fibroglandular versus adipose tissue, is one of the strongest predictors of BC risk in adult women; those with extremely mammographic dense breasts have a 2- to 4-fold increased risk compared to those with non-dense breasts [9, 10]. BTC tracks across the life course, with studies showing that breast density measures predicting BC risk remain highly correlated over time within individuals [11, 12]. Therefore, studying BTC in adolescence is important for understanding BC risk, especially as this is a period characterized by rapid epithelial and stromal proliferation, during which the breast may be particularly susceptible to external influences, such as RPA [13–15]. Prior research has shown that adolescent BTC is associated with factors such as diet, body fat, and exposure to endocrine-disrupting chemicals, indicating that exogenous factors acting through hormonal and metabolic pathways may influence breast tissue characteristics during this developmental period [16–22]. Moreover, our prior work has shown that higher levels of physical activity in early childhood is associated with a later onset of breast development, a protective factor for BC risk, [23] further supporting the hypothesis that RPA may influence breast development and BTC. However, direct evidence linking RPA to BTC during adolescence remains limited, highlighting an important gap in understanding how modifiable behaviors during this critical developmental stage may shape long-term BC risk.

RPA during adolescence may also influence biological processes such as oxidative stress and chronic inflammation, which have been linked to BTC and BC risk in adult populations [24]. For example, several studies have found that higher circulating levels of inflammatory biomarkers, including interleukin-6 (IL-6), tumor necrosis factor-alpha (TNF-α), and C-reactive protein (CRP), are associated with greater mammographic breast density in women [25–28]. However, few studies have investigated these relationships in adolescents. To date, only one study has examined inflammation in relation to BTC in adolescent girls, finding no association after adjusting for body fatness [29]. Notably, that study did not consider the potential role of RPA, which may influence both inflammation and BTC. Therefore, additional research is needed to elucidate how RPA during adolescence relates to these potentially interconnected biological pathways that may underlie future BC risk.

In this study, we examined whether RPA is associated with BTC, oxidative stress, and chronic inflammation in adolescent girls. We also investigated whether inflammation and oxidative stress are related to BTC during this developmental period. We analyzed data from a well-characterized urban cohort of Hispanic (Dominican) and non-Hispanic Black/African American adolescent girls. This population is critical to include in BC research because they have been historically underrepresented and face a higher risk of developing BC at younger ages and with more aggressive subtypes [30–32]. In addition, Black/African American and Hispanic girls consistently report lower levels of RPA than their non-Hispanic White peers, underscoring the need to elucidate how early-life RPA influences biological pathways associated with BC risk, which may contribute to disparities in disease burden [33, 34].

Methods

Study sample

We used data from the population-based Columbia Breast Cancer and the Environment Research Program Study (Columbia-BCERP Study), which recruited participants from the Columbia Center for Children’s Environmental Health (CCCEH) Mothers and Newborns birth cohort [35–37]. The CCCEH birth cohort was established between 1998 and 2006. During this period, non-Hispanic Black/African American and Hispanic (Dominican) women living in New York City were recruited during their third trimester of pregnancy from prenatal clinics at New York Presbyterian and Harlem Hospitals, and satellite clinics [35–37]. These clinics provided care for women who resided in the study catchment area–the urban neighborhoods of Washington Heights, Central Harlem, and the South Bronx in New York City [38]. Enrollment to the CCCEH birth cohort was restricted to pregnant women who were not active cigarette smokers and were aged 18–35 years [35–37].

We initially recruited 845 pregnant women in the CCCEH cohort, with 725 followed through delivery [38]. Of the children born into the cohort, 543 were actively followed through at least 2016, including 285 girls [38]. Between 2016 and 2020, 217 of the 285 actively followed girls were enrolled in the Columbia-BCERP Study; boys were excluded because the primary aim was to investigate breast development. Participants were recruited to the Columbia-BCERP Study on a rolling basis, and each participant’s clinic visit was scheduled shortly after providing informed consent to ensure timely data collection. Participants completed a clinic visit during adolescence, which included questionnaires, blood and urine collection, anthropometry measurements, and breast optical spectroscopy assessments.

For this study, we restricted the sample to 191 girls, excluding 26 girls who were missing self-reported RPA data. All 191 girls had data on BTC and relevant covariates. For the analyses of oxidative stress, we excluded an additional 28 girls who did not provide a urine sample, as well as one participant whose urinary biomarker was more than three standard deviations (SDs) below the sample mean. For the inflammatory biomarker analyses, we excluded 11 girls who did not provide a blood sample and four girls with concentrations of at least one inflammatory biomarker exceeding three SDs above the sample mean, as such elevated levels may indicate acute rather than chronic inflammation (Fig. 1). This study was approved by the IRB of Columbia University. All methods in this study were conducted in accordance with relevant guidelines and regulations (Declaration of Helsinki).

Fig. 1.

Flow diagram of study sample selection. Overview of study sample selected from the CCCEH Mothers and Newborns birth cohort. Abbreviations: BTC: Breast tissue composition; Columbia-BCERP Study : Columbia breast cancer and the environment research program study; CCCEH :  Columbia center for children’s environmental health; CRP:  C-reactive protein; IL-6: Interleukin-6; RPA: Recreational physical activity; TNF-α: Tumor necrosis factor-alpha

Recreational physical activity

We used self-reported questionnaire data to evaluate past week participation in RPA during adolescence, including both organized and unorganized types of activities. To evaluate organized RPA, participants were asked, “In the past week, did you participate in any sports teams with practices and games, classes, or lessons? Examples include soccer team, basketball team, dance classes, martial arts lessons, ice skating classes, etc.” Participants who responded “yes” to this question were asked to report the specific types of organized activities that they participated in and the number of minutes that they engaged in each activity during the past week. Participants were also asked, “In the past week, did you do other physical activities, such as roller blading, riding a bike, working out at the gym, playing outside where you live or at a playground, playing tag, jumping rope, etc.?” Participants who responded “yes” were asked to report the total number of minutes that they engaged in unorganized activities during the past week.

Because previous research has documented differences in the types, intensities, and health benefits of organized versus unorganized physical activity in youth, [39–41] we analyzed past-week participation in organized and unorganized RPA separately, categorizing each into three groups: none, < 2 h, and ≥ 2 h. These cut points were selected based on the distribution of RPA frequency in our study sample. We also created a total RPA variable that combined participation in both types of RPA into three categories: none, < 2 h, and ≥ 2 h.

Breast tissue composition

We used optical spectroscopy to objectively measure BTC, as previously described in detail [42]. Briefly, optical spectroscopy is a non-imaging and minimally invasive method for measuring BTC that does not involve ionizing radiation or breast compression [42]. It quantifies the biochemical composition of the breast tissue by capturing the red and near-infrared light absorption of chromophores in the 650–1060 nm spectral range [42]. This includes water and collagen content, which positively correlate with mammographic breast density, and lipid content, which negatively correlates with mammographic breast density [42–44]. It also captures the optical parameters related to light scattering power and amplitude, which reflect breast tissue anatomy, including connective tissue. Spectra were collected for multiple source-detector distances placed at different locations, optically interrogating different but overlapping tissue volumes [42].

A fitting algorithm was developed to determine the concentrations of the chromophores and the two light-scattering parameters of the breast tissue [42]. For the fitting algorithm, simulated spectra were calculated using the known absorption spectra of the tissue chromophores, the expected range of breast light scattering properties, and look-up tables of expected detector signal values for different tissue absorption and scattering properties [42]. The look-up tables were generated using a Monte-Carlo simulation-based light propagation program (FullMonte) [45]. The fit parameters were constrained to within expected ranges based on previous studies, with sufficient flexibility to allow for population differences and ensure physiologically reasonable values [43]. The concentrations of water, lipid, and collagen were constrained to 4–90%, 10–95%, and 1–30%, respectively [42].

As in our previous research, [42] we included only spectra with data for at least seven wavelengths, including 985 nm and either 905–940 nm, which distinguish the primary lipid and water absorption peaks, for further analysis. Chromophore fitting was performed for multiple starting points, and the best fit was selected based on the lowest chi-square difference between the measured and fitted spectra, with the additional constraint that no more than two parameters reached a predetermined minimum or maximum [42]. Chromophore and light-scattering data were averaged over both breasts, which is supported by research that has demonstrated symmetry in the optical spectra between breasts in the absence of breast disease [46]. The fitting was performed using MATLAB (The MathWorks Inc., Natick, MA, USA).

In this analysis, we examined water, collagen, and lipid content separately. We also calculated a combined optical index, defined as [(collagen*water*scattering power)/ lipid], [43] which has been shown to be more strongly correlated with breast density than the individual chromophores [42, 47].

Oxidative stress

We measured 15-F_2t-isoprostane, a valid and reliable biomarker of oxidative stress, [48] in thawed urine samples collected during the same clinic visit during which BTC was assessed using optical spectroscopy. Urinary F₂-isoprostanes are chemically stable end-products of free radical-induced lipid peroxidation and reflect systemic oxidative stress, providing a precise and accurate measure of in vivo oxidative stress [49]. Our prior research supports that F₂-isoprostane levels are stable across pubertal development [50]. Urinary 15-F_2t-isoprostane levels were quantified using immunoassay kits from Oxford Biomedical Research (Product Number: EA85, Oxford, MI), following the manufacturer’s protocol in the Biomarkers Shared Resource of the Herbert Irving Comprehensive Cancer Center and NIEHS Center. To account for hydration status, we measured specific gravity (SG) using a handheld refractometer (TS 400, Reichert, Depew, NY). Urinary biomarker levels were adjusted for SG using the Levine-Fahy equation: SG-adjusted biomarker (nmol/L*SG) = biomarker (nmol/l) × [(overall mean SG – 1)/(sample SG – 1)] [50]. The overall mean SG was 5.88 (coefficient of variation = 54%).

Chronic inflammation

We measured three inflammatory biomarkers (CRP, TNF-α, and IL-6) from blood samples collected during the same clinic visit as optical spectroscopy. These markers have demonstrated high analytic stability in stored serum and plasma samples and strong intra-individual reliability across repeated measurements, supporting their use in epidemiologic studies to reflect chronic rather than acute inflammation [51, 52]. CRP was measured using a high-sensitivity automated assay on the Roche Integra400 machine, with a detection limit of 0.3 mg/L. IL-6 and TNF-α were analyzed using high-sensitivity ELISA kits. These inflammatory biomarkers were selected a priori because they are commonly used in studies of inflammation and BC risk, [53] and they were examined in the one previous study that examined the chronic inflammation and BTC in adolescent girls [29]. These assays were performed in the Biomarkers Core Laboratory at the Columbia University Irving Institute for Clinical and Translational Research.

Statistical analysis

We used multivariable linear regression models to evaluate whether adolescent RPA was associated with BTC, assessing individual associations with water content, lipid content, collagen content, and the optical index. We evaluated two nested models: Model 1, which was adjusted for age, race, ethnicity (non-Hispanic Black/African American versus Hispanic), and optical spectroscopy device type; and Model 2, which was further adjusted for percent body fat (clinically measured using an Omron Handheld HBF-360 C). We included quadratic and cubic terms for age in these models to account for the non-linear relationship between age and BTC. These covariates were selected a priori based on prior literature, theoretical relevance, and knowledge of this cohort and its measures of BTC. We considered adjustments for other potential confounders, including maternal education, maternal country of birth, and age at menarche. However, these variables did not change the estimates and were excluded from the final models for parsimony. We also evaluated models that adjusted for body mass index (BMI) instead of percent body fat, which did not change results. We tested for effect modification by percent body fat by including a cross-product term between RPA and percent body fat categorized into tertiles.

We also used multivariable linear regression to examine whether adolescent RPA was associated with the biomarkers of oxidative stress and chronic inflammation, adjusting for age (linear term only), race and ethnicity, and percent body fat. Finally, we examined whether the biomarkers of oxidative stress and chronic inflammation were associated with the measures of BTC, adjusting for age (including linear, quadratic, and cubic terms), race and ethnicity (non-Hispanic Black/African American vs. Hispanic), optical spectroscopy device type, and percent body fat. All dependent variables were log-transformed and standardized to a mean of zero and a standard deviation of one. Statistical significance was determined as p < 0.05 for a two-sided hypothesis test. Statistical analyses were conducted using Stata 15.1 (College Station, Texas) (Stata Corporation 2017).

Results

Sample characteristics

The average age of girls was 16.2 years (SD = 2.4), and 64% identified as Hispanic (Table 1). 51% reported no past-week engagement in any type of RPA; 73% reported no participation in organized activities, and 66% reported no participation in unorganized activities. Girls who were inactive in the past week were older (mean age = 16.8, SD = 2.2) than those who were active. The average body fat of the study sample was 29.9% (SD = 7.9), and there were no statistically significant differences in percent body fat across categories of total RPA in the past week. However, those who participated in ≥ 2 h of organized RPA in the past week had a lower average percentage body fat (mean = 27.4%, SD = 7.6) than those who were inactive (mean = 30.6%, SD = 7.6).

Table 1.

Characteristics of adolescent girls in the Columbia-BCERP Study, overall and by past-week participation in recreational physical activity, n = 191

Characteristic	Overall n = 191	Total RPA in the past weeka			p-valueb
		None n = 97	< 2 h n = 40	≥ 2 h n = 54
Age, mean (SD)	16.2 (2.4)	16.8 (2.2)	15.4 (2.2)	15.6 (2.6)	0.001
Race and ethnicity, n (%c)					0.01
Hispanic	123 (100)	57 (46.3)	34 (27.6)	32 (26.0)
Non-Hispanic Black/African American	68 (100)	40 (58.8)	6 (8.8)	22 (32.4)
Maternal country of birth, n (%c)					0.17
United Statesd	85 (100)	48 (56.5)	15 (17.6)	22 (25.9)
Outside the United States	98 (100)	42 (42.9)	25 (25.5)	31 (31.6)
Maternal education, n (%c)					0.87
≤ High school degree	84 (100)	43 (51.2)	18 (21.4)	23 (27.4)
> High school degree	99 (100)	47 (47.5)	22 (22.2)	30 (30.3)
Age at menarche, mean (SD)	11.5 (1.4)	11.6 (1.6)	11.6 (1.2)	11.3 (1.3)	0.34
Percent body fat, mean (SD)	29.9 (7.9)	30.0 (7.7)	30.1 (8.2)	29.4 (8.2)	0.89
Body mass index, mean (SD)	25.6 (7.1)	25.7 (6.7)	25.5 (7.2)	25.6 (7.8)	0.98
< 25 kg/m2, n (%c)	107 (100)	54 (50.5)	22 (20.6)	31 (29.0)	0.97
≥ 25 kg/m2, n (%c)	84 (100)	43 (51.2)	18 (21.4)	23 (27.4)

BCERP: Breast cancer and the environment research program; RPA: Recreational physical activity; SD: Standard deviation

^aIncludes both organized and unorganized types of recreational physical activity activity

^bEstimated from analysis of variance test for continuous measures and chi-squared test for categorical measures

^cRow percentages

^dExcludes Puerto Rico

Correlations between inflammatory and oxidative stress biomarkers

The three inflammatory biomarkers were correlated with one another (Pearson correlation coefficients: CRP and TNF-α = 0.23, CRP and IL-6 = 0.63, TNF-α and IL-6 = 0.30; all p-values < 0.01), but they were not correlated with 15-F_2t-isoprostane. As shown in Fig. 2, percent body fat was positively correlated with CRP (Pearson correlation coefficient = 0.59, p-value < 0.001) and IL-6 (Pearson correlation coefficient = 0.48, p-value < 0.001), but not with TNF-α or 15-F_2t-isoprostane.

Fig. 2.

Correlations between percent body fat and biomarkers of oxidative stress and chronic inflammation among adolescent girls in the Columbia-BCERP study. Scatter plot with fitted values from a linear regression of percent body fat on each inflammatory blood biomarker

Associations of recreational physical activity with breast tissue composition

In the fully adjusted model, participating in ≥ 2 h of total RPA in the past week versus none was associated with lower percent water content in the breast tissue of adolescent girls (Table 2, Model 2: β: − 0.34, 95% CI: − 0.63, − 0.05). We also found an association between participating in ≥ 2 h of organized RPA in the past week versus none and lower percent water content (Model 2: β: − 0.41, 95% CI: − 0.77, − 0.05); this equates to about a 2.5% difference in the marginal mean percent water content (Fig. 3, Panel A). This association did not statistically significantly differ across tertiles of percent body (Fig. 3, Panel B; interaction p-value = 0.64). No association was found between participation in unorganized RPA and percent water content. There were no associations between RPA and the other measures of BTC (percent lipid content, percent collagen content, and the optical index).

Table 2.

Associations of hours of recreational physical activity in the past week and breast tissue composition among adolescent girls in the Columbia-BCERP Study, n = 191

Measure of RPA	Measure of breast tissue composition
	% Water content	% Lipid content	% Collagen content	Optical index
Model	β (95% CI)	β (95% CI)	β (95% CI)	β (95% CI)
Total RPA in the past week
Model 1a
none	ref.	ref.	ref.	ref.
< 2 h	− 0.19 (− 0.52, 0.14)	0.09 (− 0.22, 0.40)	− 0.32 (− 0.68, 0.05)	− 0.02 (− 0.36, 0.32)
≥ 2 h	− 0.32 (− 0.62, − 0.03)	0.12 (− 0.15, 0.40)	0.05 (− 0.28, 0.38)	− 0.13 (− 0.43, 0.18)
Model 2b
none	ref.	ref.	ref.	ref.
< 2 h	− 0.17 (− 0.49, 0.15)	0.07 (− 0.22, 0.36)	− 0.31 (− 0.67, 0.06)	− 0.01 (− 0.35, 0.33)
≥ 2 h	− 0.34 (− 0.63, − 0.05)	0.14 (− 0.12, 0.40)	0.05 (− 0.28, 0.37)	− 0.14 (− 0.44, 0.17)
Organized RPA in the past week
Model 1a
none	ref.	ref.	ref.	ref.
< 2 h	− 0.15 (− 0.54, 0.25)	0.19 (− 0.18, 0.56)	− 0.09 (− 0.53, 0.35)	0.15 (− 0.25, 0.56)
≥ 2 h	− 0.31 (− 0.68, 0.06)	− 0.01 (− 0.35, 0.33)	0.10 (− 0.30, 0.51)	0.01 (− 0.37, 0.38)
Model 2b
none	ref.	ref.	ref.	ref.
< 2 h	− 0.21 (− 0.60, 0.18)	0.27 (− 0.08, 0.62)	− 0.13 (− 0.57, 0.31)	0.11 (− 0.30, 0.52)
≥ 2 h	− 0.41 (− 0.77, − 0.05)	0.12 (− 0.21, 0.44)	0.05 (− 0.36, 0.46)	− 0.06 (− 0.44, 0.32)
Unorganized RPA in the past week
Model 1a
none	ref.	ref.	ref.	ref.
< 2 h	− 0.15 (− 0.50, 0.20)	0.07 (− 0.25, 0.39)	− 0.36 (− 0.74, 0.03)	− 0.08 (− 0.44, 0.28)
≥ 2 h	− 0.25 (− 0.57, 0.08)	0.21 (− 0.10, 0.51)	0.09 (− 0.27, 0.45)	− 0.12 (− 0.46, 0.22)
Model 2b
none	ref.	ref.	ref.	ref.
< 2 h	− 0.10 (− 0.44, 0.24)	0.00 (− 0.31, 0.31)	− 0.33 (− 0.71, 0.06)	− 0.05 (− 0.40, 0.31)
≥ 2 h	− 0.20 (− 0.53, 0.12)	0.15 (− 0.14, 0.44)	0.12 (− 0.24, 0.48)	− 0.09 (− 0.42, 0.25)

RPA: Recreational physical activity

^aLinear regression models are adjusted for age (linear, quadratic, and cubic terms), race, ethnicity, and optical spectroscopy device type. Outcome measures are log-transformed and standardized to a mean of zero and a standard deviation of one

^bLinear regression models are adjusted for age (linear, quadratic, and cubic terms), race, ethnicity, optical spectroscopy device type, and percent body fat. Outcome measures are log-transformed and standardized to a mean of zero and a standard deviation of one

Fig. 3.

Estimated marginal geometric mean percent breast water content by hours of organized recreational physical activity in the past week, overall and stratified by percent body fat, among adolescent girls in the Columbia-BCERP Study, n = 191. Marginal geometric means and 95% confidence intervals are estimated from linear regression models adjusted for age (modeled using linear, quadratic, and cubic terms), race, ethnicity, optical spectroscopy device type, and percent body fat. The model presented in Panel B also includes an interaction term between percent body fat, categorized into tertiles, and organized recreational physical activity ( p-value for interaction = 0.64). The outcome variable (percent water content) was log-transformed and standardized to a mean of zero and a standard deviation of one in the models; marginal means were back-transformed to the original scale. Abbreviations: hrs = hours; RPA = recreational physical activity

Associations of recreational physical activity, oxidative stress, and chronic inflammation

Participating in ≥ 2 h of organized RPA in the past week, compared to none, was associated with lower urinary concentrations of 15-F_2t-isoprostane in adolescent girls, both before and after adjustment for percent body fat (Table 3; Model 2: β: − 0.50, 95% CI: − 0.95, − 0.05). RPA was not associated with any of the three biomarkers of chronic inflammation (CRP, TNF-α, or IL-6).

Table 3.

Associations of hours of recreational physical activity in the past week and biomarkers of oxidative stress and chronic inflammation among adolescent girls in the Columbia-BCERP study

Measure of RPA	15-F2t-isoprostane n = 162	CRP n = 176	TNF-α n = 176	IL-6 n = 176
Model	β (95% CI)	β (95% CI)	β (95% CI)	β (95% CI)
Total RPA in the past week
Model 1a
none	ref.	ref.	ref.	ref.
< 2 h	− 0.15 (− 0.54, 0.25)	0.17 (− 0.21, 0.54)	0.01 (− 0.34, 0.36)	0.10 (− 0.27, 0.47)
≥ 2 h	− 0.16 (− 0.53, 0.21)	0.07 (− 0.26, 0.41)	0.11 (− 0.20, 0.42)	− 0.05 (− 0.38, 0.28)
Model 2b
none	ref.	ref.	ref.	ref.
< 2 h	− 0.15 (− 0.54, 0.25)	0.10 (− 0.20, 0.40)	0.00 (− 0.35, 0.35)	0.05 (− 0.27, 0.36)
≥ 2 h	− 0.16 (− 0.53, 0.21)	0.07 (− 0.19, 0.34)	0.11 (− 0.20, 0.42)	− 0.05 (− 0.33, 0.23)
Organized RPA in the past week
Model 1a
none	ref.	ref.	ref.	ref.
< 2 h	− 0.25 (− 0.70, 0.20)	0.10 (− 0.36, 0.56)	0.07 (− 0.35, 0.49)	0.01 (− 0.37, 0.40)
≥ 2 h	− 0.48 (− 0.93, − 0.04)	0.02 (− 0.41, 0.45)	− 0.07 (− 0.47, 0.33)	0.04 (− 0.33, 0.40)
Model 2b
none	ref.	ref.	ref.	ref.
< 2 h	− 0.26 (− 0.71, 0.19)	0.21 (− 0.15, 0.57)	0.07 (− 0.35, 0.49)	0.13 (− 0.25, 0.51)
≥ 2 h	− 0.50 (− 0.95, − 0.05)	0.21 (− 0.13, 0.55)	− 0.07 (− 0.47, 0.33)	− 0.17 (− 0.53, 0.20)
Unorganized RPA in the past week
Model 1a
none	ref.	ref.	ref.	ref.
< 2 h	0.03 (− 0.38, 0.45)	0.07 (− 0.33, 0.47)	0.07 (− 0.29, 0.44)	0.04 (− 0.40, 0.49)
≥ 2 h	0.09 (− 0.31, 0.49)	0.01 (− 0.36, 0.39)	0.16 (− 0.18, 0.51)	− 0.32 (− 0.74, 0.10)
Model 2b
none	ref.	ref.	ref.	ref.
< 2 h	0.04 (− 0.38, 0.46)	− 0.09 (− 0.40, 0.23)	0.07 (− 0.29, 0.44)	− 0.11 (− 0.45, 0.22)
≥ 2 h	0.09 (− 0.31, 0.49)	− 0.14 (− 0.44, 0.16)	0.16 (− 0.18, 0.51)	− 0.09 (− 0.41, 0.23)

CRP: C-reactive protein; IL-6: Interleukin-6; RPA: Recreational physical activity; TNF-α: Tumor necrosis factor-alpha

^aLinear regression models are adjusted for age (linear term), race, and ethnicity. Outcome measures are log-transformed and standardized to a mean of zero and a standard deviation of one

^bLinear regression models are adjusted for age (linear term), race, ethnicity, and percent body fat. Outcome measures are log-transformed and standardized to a mean of zero and a standard deviation of one

Associations of oxidative stress and chronic inflammation with breast tissue composition

Higher urinary concentrations of 15-F_2t-isoprostane were associated with higher collagen content in the breast tissue of adolescent girls, both before and after adjusting for percent body fat (Table 4, Model 2: β: 0.15, 95% CI: 0.00,0.31). No associations were observed between 15-F_2t-isoprostane and the other measures of BTC. Before adjusting for percent body fat, higher circulating levels of CRP were associated with higher lipid content (Model 1: β: 0.24, 95% CI: 0.11,0.36), lower collagen content (Model 1: β: − 0.19, 95% CI: − 0.34, − 0.03), and lower optical index (Model 1: β: − 0.15, 95% CI:− 0.29, − 0.01). However, these associations were attenuated and no longer statistically significant after adjusting for percent body fat. No associations were observed between TNF-α or IL-6 with BTC.

Table 4.

Associations of biomarkers of oxidative stress and chronic inflammation with breast tissue composition among adolescent girls in the Columbia-BCERP study

Biomarker	n	Measure of breast tissue composition
Model		% Water content β (95% CI)	% Lipid content β (95% CI)	% Collagen content β (95% CI)	Optical index β (95% CI)
15-F2t-isoprostane	162
Model 1a		− 0.07 (− 0.20, 0.06)	0.01 (− 0.12, 0.14)	0.16 (0.00, 0.31)	0.03 (− 0.08, 0.15)
Model 2b		− 0.07 (− 0.20, 0.05)	0.01 (− 0.11, 0.14)	0.15 (0.00, 0.31)	0.03 (− 0.08, 0.14)
CRP	176
Model 1a		− 0.09 (− 0.23, 0.05)	0.24 (0.11, 0.36)	− 0.19 (− 0.34, − 0.03)	− 0.15 (− 0.29, − 0.01)
Model 2b		0.06 (− 0.12, 0.23)	0.13 (− 0.02, 0.29)	− 0.19 (− 0.38, 0.01)	− 0.10 (− 0.28, 0.08)
TNF-α	176
Model 1a		0.09 (− 0.06, 0.24)	0.07 (− 0.07, 0.21)	− 0.09 (− 0.26, 0.08)	− 0.05 (− 0.21, 0.11)
Model 2b		0.12 (− 0.02, 0.27)	0.03 (− 0.10, 0.17)	− 0.07 (− 0.24, 0.10)	− 0.03 (− 0.19, 0.13)
IL-6	176
Model 1a		− 0.09 (− 0.23, 0.05)	0.10 (− 0.03, 0.23)	− 0.14 (− 0.30, 0.02)	− 0.08 (− 0.23, 0.07)
Model 2b		0.03 (− 0.14, 0.19)	− 0.05 (− 0.20, 0.10)	− 0.11 (− 0.30, 0.08)	− 0.01 (− 0.18, 0.17)

CRP: C-reactive protein; IL-6: Interleukin-6; TNF-α: Tumor necrosis factor-alpha

^aLinear regression models are adjusted for age (linear, quadratic, and cubic terms), race, ethnicity, and optical spectroscopy device type. Outcome measures are log-transformed and standardized to a mean of zero and a standard deviation of one

^bLinear regression models are adjusted for age (linear, quadratic, and cubic terms), race, ethnicity, optical spectroscopy device type, and percent body fat. Outcome measures are log-transformed and standardized to a mean of zero and a standard deviation of one

Discussion

This study provides some of the first data on the relationship between RPA and BTC in adolescent girls. We found that girls who reported at least two hours of RPA in the past week versus none had lower percent water content in the breast tissue, an indicator of lower breast density [43, 54–56]. These findings align with prior research in adult women, where several studies have shown that higher levels of RPA are associated with lower mammographic breast density [57–60]. For example, a cohort study of Hispanic women aged 40 years and older found that physically inactive women had higher percent breast density than active women, [59] and a longitudinal study reported that active women experienced greater reductions in dense breast area over time compared to inactive women [57]. Together, these findings suggest that RPA may promote favorable changes in BTC that persist across the life course, highlighting the importance of examining these relationships earlier in development. However, given that other studies in adult women have reported null associations between RPA and BTC, [61] and that our study is among the first to investigate this relationship in adolescent girls, additional longitudinal research is needed to clarify the role of RPA in shaping BTC across the life course and its potential implications for BC risk.

When we examined associations between RPA and BTC by activity type, only participation in organized activities was associated with lower percent water content in adolescent girls; no association was found for participation in unorganized activities (e.g., free play, bike riding). These findings may suggest that more structured types of activities, which often involve higher intensity and physical fitness, [62, 63] may be more important for reducing breast density during adolescence. However, our classification of organized and unorganized activities might have been imprecise, and we were unable to assess the intensity levels of various activities. Our findings may thus reflect possible non-differential measurement error, especially as certain types of organized activities may be easier to accurately recall, especially compared to unorganized activities. Therefore, further studies using more objective measures of RPA, such as actigraphy, are needed to better understand how exercise intensity, duration, and type influence BTC during adolescence. Nonetheless, BTC was measured objectively using optical spectroscopy, and all assessments were conducted blinded to participants’ reported RPA, reducing the potential for measurement bias. As a result, this study provides important new insights into the relationship between RPA and BTC during adolescence, a critical period of development.

We also found that participation in organized RPA was associated with lower oxidative stress levels in adolescent girls. Oxidative stress occurs when there is an imbalance between oxidants and antioxidants, leading to potential DNA damage and increased BC risk [64]. Higher levels of oxidative stress have been observed in women with BC compared to those without the disease, [65–67] and oxidative stress has been linked to established BC risk factors, including elevated estrogen levels [50, 68, 69]. In our analysis, we found that adolescent girls who participated in ≥ 2 h of organized RPA in the past week had lower urinary levels of 15-F_2t-isoprostane, a well-established biomarker of lipid peroxidation, [48] compared to inactive girls. This aligns with previous research demonstrating that regular exercise enhances antioxidant defenses and mitigates oxidative stress in children and adolescents [70, 71]. A recent accelerometer-based study in children and adolescents found that higher moderate-to-vigorous physical activity was associated with lower urinary concentrations of oxidative stress markers, including isoprostane F2α and 8-hydroxy-2′-deoxyguanosine [72]. Additionally, studies comparing children who participate in sports to those who do not suggest that engaging in structured activities may lead to higher antioxidant levels [73, 74]. Our findings add to this body of evidence, but additional longitudinal research is still needed to clarify the long-term health implications of this association, particularly in relation to long-term BC risk.

When we examined the relationship between oxidative stress and BTC, we found that higher urinary levels of 15-F_2t-isoprostane were associated with higher collagen content in the breast tissue of adolescent girls, which is positively associated with breast density [43, 54–56]. These findings provide some of the first evidence linking oxidative stress to BTC during adolescence and are consistent with previous studies that have found an association between oxidative stress and higher mammographic breast density in adult women [75]. One study found that urinary malondialdehyde excretion, another biomarker of oxidative stress, was 23–30% higher among women in the highest versus the lowest quintile of mammographic breast density, adjusting for BMI or waist circumference [76]. Experimental research has also shown that F2-isoprostanes generated by lipid peroxidation in hepatocytes can drive hepatic stellate cell proliferation and excessive collagen production, contributing to liver fibrosis [77]. The parallels between oxidative stress-driven fibrosis in the liver and increased collagen content in breast tissue suggest that oxidative stress may play a broader role in tissue remodeling and fibrotic processes across different organ systems. Together, these findings support a positive relationship between oxidative stress and collagen content in the breast, including during adolescence. However, we did not find evidence to support that RPA contributes to this relationship, as RPA was not associated with collagen content in our study sample.

We did not find an association between RPA and chronic inflammation in adolescent girls, although our study was limited to only three inflammatory biomarkers. Nevertheless, these results contrast with previous research showing that physical activity can improve inflammatory profiles, including lower circulating levels of CRP and IL-6, in children and adolescents [78, 79]. The lack of an observed association in our study may be due to several factors, including the timing, intensity, and overall distribution of RPA within our sample. Notably, 73% of participants in our sample reported no organized RPA in the past week, and fewer than 10% engaged in three or more hours per week. As a result, we were limited in our ability to evaluate how longer durations of RPA may impact chronic inflammation. We also did not find an association between chronic inflammation and BTC in adolescent girls after adjusting for percent body fat, which is consistent with the only other previous study to examine this relationship in adolescent girls [29]. However, because these studies focused on circulating markers of inflammation, the role of inflammation within the breast tissue itself, and whether it is influenced by RPA, remains largely unexplored in adolescent girls.

This study has several strengths, including the use of multiple objective biomarkers measured in urine, blood, and breast tissue of adolescent girls. Notably, we measured biomarkers of oxidative stress and chronic inflammation that are among the most widely validated and commonly used in epidemiologic research, [49, 53, 80] enhancing the validity and cross-study comparability of our findings. Additionally, this research was conducted in a population-based urban cohort of Black/African American and Hispanic girls, thus focusing on groups that are historically underrepresented in research and face persistent disparities in both physical activity and BC outcomes [33, 34, 81]. However, it is not without limitations.

First, RPA was assessed using self-reported data that captured activity over a limited time frame (past week). Therefore, our exposure variables are susceptible to non-differential measurement error and may not fully capture habitual patterns, as both organized and unorganized activities can fluctuate substantially across seasons and over time. However, past-week RPA was correlated with self-reported general activity levels in our study, such that only 15% of girls who reported no past-week RPA described themselves as active “most or all of the time,” compared with 42% and 67% of those reporting < 2 h and ≥ 2 h of RPA, respectively. This provides some support that past-week RPA reflects broader, habitual activity in our study sample. Moreover, our measures of RPA were adapted from the widely utilized Physical Activity Questionnaire for Adolescents (PAQ-A), which is a reliable and valid tool for assessing physical activity behaviors in youth [82–84]. We also acknowledge that the high proportion of inactive participants in our study limited our ability to examine more homogeneous activity groups, such as those defined by organized activity type. As a result, activities of varying intensity were likely combined into a single category. Future research with larger and more active samples is needed to evaluate how different types and intensities of RPA influence biomarkers during adolescence.

Another limitation of this study is that, although we leveraged a prospective cohort study, our analysis used cross-sectional data, which prevents us establishing temporality or inferring causal relationships between RPA and adolescent biomarkers of BC risk. We were also limited by a relatively small sample size, as this study was conducted within a subset of an existing birth cohort. This reduced our statistical power to detect associations. Nevertheless, with our sample size we were able to detect modest associations between recreational physical activity and biomarkers of breast tissue composition and oxidative stress, suggesting that these relationships are robust and warrant further investigation in larger studies. Moreover, our findings may not be generalizable to populations with different demographics or higher levels of physical activity, as RPA was notably low in our cohort. Finally, although we examined several validated biomarkers of BC risk, other mechanisms, such as hormonal regulation and insulin sensitivity, may also be important for understanding the relationship between adolescent RPA and BC risk [24]. Given these considerations, additional longitudinal research on the role of adolescent RPA in mechanisms linked to BC risk over time is warranted.

Conclusions

This study provides some of the first evidence that RPA may be associated with BTC and oxidative stress in adolescent girls, offering new insights into the potential biological pathways through which RPA operates during this critical developmental period. Further longitudinal studies are needed to understand the implications of these findings regarding subsequent BC risk. The importance and urgency of this research are underscored by the rising incidence of BC in young women [6–8] and the alarmingly low levels of RPA observed both in this study and in adolescent populations more broadly [33, 85].

Abbreviations

BC

breast cancer

BMI

body mass index

BTC

breast tissue composition

CCCEH

Columbia center for children’s environmental health

Columbia-BCERP Study

Columbia breast cancer and the environment research program study

CRP

C-reactive protein

IL-6

interleukin-6

RPA

recreational physical activity

SD

standard deviations

SG

specific gravity

TNF-α

tumor necrosis factor-alpha

Author contributions

FP designed the original Columbia Center for Children’s Environmental Health Mothers and Newborns birth cohort and supervised the collection of the pregnancy-related exposure measures. MBT designed and supervised the field activities of the Columbia-BCERP follow-up study conducted in adolescent girls and their mothers from the original birth cohort. RDK, RLM, and JBH gave critical input throughout the follow-up study fieldwork. MLW acquired the data for this analysis. RMS oversaw the execution and interpretation of all laboratory assays. RDK and EJW conducted statistical analyses. RDK, EJW, LL, and MBT contributed to data interpretation and critical evaluation of statistical models. RDK initially drafted to manuscript. All authors reviewed the manuscript and provided critical input into the manuscript.

Funding

This research was supported by grants U01ES026122 and P30ES009089 from the National Institute of Environmental Health Sciences, as well as grant R00CA263024 from the National Cancer Institute.

Data availability

The dataset used and analyzed during this study is available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the Institutional Review Board at Columbia University. Written informed consent was obtained from all mothers for themselves and for their daughters under age 18 years at the clinic visit, along with written informed assent from daughters. Written informed consent was also obtained from daughters who were 18 years or older at enrollment in the Columbia-BCERP Study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.