Investigating the relationship between birthweight and breast cancer from a non-linear and mediation perspective

doi:10.21203/rs.3.rs-1611382/v1

Download PDF

Research Article

Investigating the relationship between birthweight and breast cancer from a non-linear and mediation perspective

https://doi.org/10.21203/rs.3.rs-1611382/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Purpose: Epidemiological studies have shown a positive relationship between birthweight and breast cancer; however, inconsistent, sometimes even controversial, observations also emerged. We here attempted to re-explore the association between them in the UK Biobank cohort.

Methods: Relying on large-scale data available from the UK Biobank cohort (5,760 cancer cases and 162,778 controls), we evaluated the causal linear mediation between birthweight and breast cancer, with age of menarche and age at menopause as two potential mediators under the traditional mediation analysis framework. The non-linear relationship between birthweight and breast cancer was also investigated by including the square of birthweight or discretized birthweight categories (<2.5, 2.5~4.0, or >4.0). Furthermore, we performed a stratification analysis in terms of the menopause status.

Results: We revealed that birthweight can indirectly influence breast cancer risk in adulthood via the path of age of menarche or age at menopause, and found statistical evidence supporting the existence of suggestive non-linear association between birthweight and breast cancer (β=0.062 and P=0.004 for the square of birthweight) although failing to discover a linear relationship (P=0.230). We also demonstrated that such non-linear association seemed more pronounced and robust for premenopausal women compared to postmenopausal ones (27.5% vs. 19.5% increase in breast cancer risk).

Conclusions: This study provides an in-depth insight into the observed relationship between birthweight and breast cancer and reveals that non-linear impact and causal mediation commonly drive the connection between the two traits.

breast cancer

birthweight

UK Biobank

non-linear association

mediation analysis

We explicitly revealed the existence non-linear association between birthweight and breast cancer and showed that women birth with both low and high weight had higher risk to develop breast cancer in adulthood, reinforcing previous findings.
We discovered that the non-linear effect of birthweight on breast cancer was more evident and robust for premenopausal women compared to postmenopausal ones.
We demonstrated that birthweight can indirectly influence breast cancer in adulthood via the path of age of menarche or age at menopause.

Over the past few decades, the relationship between breast cancer and early growth/development, perinatal intrauterine environments has been attracted much research attention, forming the well-known hypothesis that breast cancer might, to a certain extent, originate in utero [1–3]. The intrauterine origin of breast cancer has important implication on the connection of its causal risk factors to estrogens and other hormones generated during pregnancy or to other pregnancy-relevant life processes. Indeed, it has been revealed that exposing to higher levels of endogenous estrogen in utero is pathophysiologically associated with increased risk of breast cancer [1, 2, 4], offering a meaningful complementary interpretation to the natural etiology of breast cancer.

As it is difficult or unrealistic to obtain pregnancy estrogen measurements retrospectively for women who have already developed breast cancer in adulthood, some indicators of high estrogen exposure are therefore employed as surrogate measures. Among those, weight at birth, which is positively correlated with pregnancy estrogen concentrations, is intensively studied in the literature (Table S1). For example, a positive correlation between women’s birthweight and breast cancer risk was discovered in the cohorts of two USA nurses’ health studies [5]; a meta-analysis of 18 epidemiological studies indicated that women born with weight > 4kg had 20% (95% confidence intervals [CIs] 8–34%) higher risk of breast cancer than those born with weight < 3kg [6], or 7% (95%CIs 2–12%) increased risk per 1kg, in agreement with results obtained from other meta-studies [7]. It was also detected that the risk of breast cancer for individuals with high birthweight (≥ 4kg) almost doubled compared to those with low birthweight (< 2.5kg) after adjusting for age of menarche, birth times, adult body mass index (BMI) and other covariates in terms of a population-based case-control study of Asian-American in Los Angeles [8]. Similar findings were identified in other countries and regions including a Norway study [9], Denmark studies [10–12], Britain studies [13, 14], as well as a black women’s health study [15]. This positive association between birthweight and breast cancer risk was further supported by the observation in the study of opposite-sexed twins [4] and the animal experiment [16].

Although the studies described above have demonstrated the existence of a relationship between high birthweight and increased risk of breast cancer, some others failed to replicate such connection or even demonstrated inconsistent correlations in effect direction (Table S1), making it difficult to draw a definitive conclusion on the causal association between birthweight and breast cancer. Furthermore, previous work focused primarily on the linear relationship between birthweight and breast cancer (Table S1), it is not clear whether there exists a non-linear or mediating connection between them [13, 17]. For instance, it was shown that individuals with birthweight < 2.5kg had a 30% higher risk and individuals with birthweight > 4.0kg had a 70% higher risk of breast cancer compared to those with normal birthweight (2.5-4.0kg) [3], implying that both low and high birthweight might increase the risk of this type of cancer [3, 10, 16]; however, a formal analysis is relatively lacking for such relationship.

The present work attempts to assess the relationship between birthweight and breast cancer using large-scale data available from the UK Biobank cohort [18]. To this aim, we first conducted a mediation analysis with age of menarche and age at menopause as two candidate mediators that stand on the path from birthweight to breast cancer. Then, we evaluated the non-linear relationship between birthweight and breast cancer to study the influence of birthweight on breast cancer. The flow diagram of data process and statistical analysis for the present study is illustrated in Fig. 1.

2.1 Individual-level UK Biobank dataset

We applied the UK Biobank data to investigate the influence of birthweight on breast cancer [18], with age of menarche and age at menopause as two candidate mediators. After performing the similar quality control procedure described in prior work [19], we obtained a total of 337,198 independent individuals of white British ancestry aged 48–82 years. We only kept female individuals with breast cancer as cases, leaving 7,350 breast cancer patients. To guarantee the temporal ordering between age of menarche, age at menopause and breast cancer, which is necessary for the causal interpretation of effects in the causal inference [20, 21], we ensured that no breast cancer patients reported their age at menopause earlier than age at menarche, and excluded 1,590 patients whose age at diagnosis was prior to age at menarche or age at menopause. Afterwards, we reserved 5,760 breast cancer patients. To maximize the sample size for boosting power, we included all female individuals without breast cancer as controls, leading to a total of 162,778 controls. Besides birthweight, age of menarche, age at menopause and status of breast cancer, we primarily incorporated age (by the end of the last data collection), menopause or not, ever smoked or not and BMI as potential covariates (Table 1).

Table 1

Descriptive statistics of the UK Biobank data after quality control used in the mediation analysis
variable	N	mean ± sd (yes/no)
birthweight (kg)	108,956	3.2 ± 0.6
age at menarche (year)	163,763	13.0 ± 1.6
age at menopause (year)	96,098	49.8 ± 5.0
age (year)	168,539	66.5 ± 7.9
BMI (m/kg²)	130,635	27.0 ± 5.1
breast cancer	168,538	5,760/162,778
menopause or not	141,969	102,889/39,080
ever smoked or not	167,983	93,396/74,587
Note: BMI: body mass index; N: the sample size of diverse variables, which is different due to the distinct settings of missing value; sd: standard deviation.

2.2 Mediation association from birthweight to breast cancer mediated by age at menarche or age at menopause

Using the UK Biobank dataset, we aimed to explore the association between birthweight (the exposure X) and breast cancer (the outcome Y), with age at menarche (the first mediator M₁) and age at menopause (the second mediator M₂) as two potential mediators (Fig. 2). We implemented our mediation analysis under the traditional framework with varying covariates in different mediation models [22]. The basic principle of incorporating covariates in these models was that we considered covariates if they were measured at the same time of collecting the outcome of focus. For example, when it came to age at menarche, we would not choose any covariates as none of them was measured at the age at menarche for a woman in the UK Biobank dataset; when it came to breast cancer, we considered BMI and smoking but would not include age at menopause if we only analyzed age at menarche. Moreover, we fit a linear model when analyzing continuous outcomes (e.g., age at menarche or age at menopause), whereas we fit a logistic model when analyzing binary outcomes (e.g., breast cancer).

With the two mediators under consideration, we here highlight that age at menarche can affect age at menopause but not vice versa [23]. As a consequence, there are eight association possibilities, with four potentially consecutive paths from birthweight to breast cancer (Fig. 2): (i) a direct association between birthweight and breast cancer but not mediated by neither age at menarche nor age at menopause; (ii) birthweight affects breast cancer through age at menarche alone; (iii) birthweight impacts breast cancer through age at menopause alone; (iv) birthweight influences breast cancer through age at menarche, subsequently by age at menopause. The summation of effect sizes on these paths is equal to the total causal effect of birthweight on breast cancer, and the summation of the last three effect sizes can be viewed as the indirect effect of birthweight. Note that, because the temporal ordering between these variables is determinative, these estimated effects have a causal interpretation if additional sequential ignorability assumptions are assumed to be satisfied [20].

2.3 Evaluating linear and non-linear relationship between birthweight and breast cancer

As would be shown below, though failing to detect an obvious linear relationship between birthweight and breast cancer in Model 1 and only identifying a marginally significant association between them in Model 8, we cannot completely exclude the likelihood that there might exist a non-linear association between birthweight and breast cancer [3, 10, 16]. To assess such relationship, we performed a logistic regression by including the square of birthweight to examine its non-linear association with breast cancer. In addition, we also discretized birthweight into three categories (< 2.5, 2.5 ~ 4.0, or > 4.0 by following prior work; see for instance in Table S1) and carried out the similar logistic analysis above. Furthermore, we performed a stratification analysis in terms of the menopause status in each analysis setting. In the present work, all statistical analyses were implemented under the R software computing environment [24], with a significance level of 0.05.

3.1 Mediation paths between birthweight and breast cancer

To explore potential mediation paths, we carefully selected a total of 168,539 (5,760 cases and 162,778 controls) female individuals from the UK Biobank cohort and examine each possible relationship shown in Fig. 2. First, we implemented a logistic regression analysis to explore the linear effect of birthweight on breast cancer while adjusting for three available covariates (Model 1), but find a null association between them (P = 0.230) (Table 2). Note that, in terms of a simple simulation study implemented with these estimated parameters, we have approximately 100% statistical power, indicating this null finding is not likely a consequence of lack of power. Therefore, we consider the model as a complete mediation model in our analysis. Second, we performed two separate linear regressions to study the association between birthweight and age at menarche (Model 2) or between birthweight and age at menopause (Model 3), and display that birthweight has a positive influence on each of two ages, with the effect size being β = 0.041 (95% confidence intervals [CIs]: 0.025 ~ 0.057) for age at menarche and β = 0.212 (95%CIs: 0.149 ~ 0.275) for age at menopause. At the same time, we find that age at menarche can positively affect age at menopause (Model 4), with the effect size being β = 0.038 (95%CIs: 0.018 ~ 0.058).

Table 2

Potential mediation paths and effect sizes between birthweight and breast cancer with two mediators
	X→Y (Model 1) (case = 3,376 and control = 105,012)	X→M₁ (Model 2) (n = 106,811)	X→M₂ (Model 3) (n = 58,311)	M₁→M₂ (Model 4) (n = 93,768)
	β (se, P)	β (se, P)	β (se, P)	β (se, P)
X	0.033 (0.027, 0.230)	0.041 (0.008, 6.42×10^− 8)	0.212 (0.032, 2.92×10^− 11)
M₁				0.038 (0.010, 2.49×10^− 4)
M₂
C₁	0.717 (0.039, 2.21×10^− 75)
C₂	0.069 (0.035, 0.053)		-0.326 (0.042, 5.82×10^− 15)	-0.329 (0.033, 4.56×10^− 23)
C₃	0.004 (0.003, 0.206)
	X + M₁→M₂ (Model 5) (n = 57,514)	X + M₁→Y (Model 6) (case = 3,314 and control = 102,948)	X + M₂→Y (Model 7) (case = 2,274 and control = 55,871)	X + M₁ + M₂→Y (Model 8) (case = 2,247 and control = 55,102)
	β (se, P)	β (se, P)	β (se, P)	β (se, P)
X	0.214 (0.032, 2.86×10^− 11)	0.035 (0.028, 0.205)	0.055 (0.033, 0.096)	0.061 (0.033, 0.067)
M₁	0.045 (0.013, 6.62×10^− 4)	-0.052 (0.011, 5.81×10^− 6)		-0.050 (0.014, 3.38×10^− 4)
M₂			-0.039 (0.004, 8.01×10^− 24)	-0.039 (0.004, 3.48×10^− 24)
C₁		0.708 (0.039, 3.50×10^− 72)
C₂	-0.326 (0.042, 8.97×10^− 15)	0.079 (0.036, 0.028)	0.044 (0.043, 0.311)	0.056 (0.044, 0.199)
C₃		0.002 (0.003, 0.593)	0.005 (0.004, 0.268)	0.003 (0.004, 0.515)
Note: X: birthweight, M₁: age at menarche, M₂: menopausal age; C₁: menopause; C₂: smoke; C₃: BMI. The intercept is not included here.

Third, we find age at menarche can mediate the influence of birthweight on age at menopause (Model 5), with the effect size of age at menarche on age at menopause estimated to be β = 0.045 (95%CIs: 0.020 ~ 0.070) conditional on birthweight; meanwhile, birthweight also has a direct effect on age at menopause (β = 0.214, 95%CIs: 0.151 ~ 0.277). In addition, with age at menarche or age at menopause as a single mediator (Model 6 and Model 7), we discover that both the two ages have a significantly negative impact on the risk of breast cancer, with odds ratio [OR] = 0.95 (95%CIs: 0.92 ~ 0.97) for age at menarche and OR = 0.96 (95%CIs: 0.95 ~ 0.97) for age at menopause, indicating the earlier the age at menarche or age at menopause, the higher the risk of breast cancer. However, in both mediation models, we do not detect the presence of direct effect of birthweight on breast cancer (P = 0.205 and 0.096, respectively) after controlling for age at menarche or age at menopause. Finally, we considered to incorporate age at menarche and age at menopause in the mediation model simultaneously (Model 8), and discover that both the two ages are still significantly associated with breast cancer (OR = 0.95, 95%CIs: 0.93 ~ 0.98 for age at menarche; OR = 0.96, 95%CIs: 0.95 ~ 0.97 for age at menopause), and that birthweight has a positive effect on breast cancer at the marginally significant level (OR = 1.06, 95%CIs: 1.00 ~ 1.13). To be more concise and evident, we demonstrate these estimated effects and relationships in Fig. 3. Relying on the naïve principle of mediation analysis [20–22, 25], we can conclude that birthweight can indirectly affect adult breast cancer through age at menarche or age at menopause.

3.2 Non-linear relationship between birthweight and breast cancer

Though not detecting a linear relationship between birthweight and breast cancer in Model 1 (P = 0.230) and only discovering a marginally significant association between them in Model 8 (P = 0.067), there likely exists a non-linear association with breast cancer. To evaluate such relationship, we further performed a logistic regression by including the square of birthweight into models (Table 3). As demonstrated, we observe a positive and significant association between the square of birthweight and breast cancer in terms of Model 1 (β = 0.062, 95%CIs: 0.019 ~ 0.105), but such association becomes weaken and is no longer significant according to Model 8 (β = 0.037, 95%CIs: -0.016 ~ 0.090). The stratification analysis in terms of the menopause status leads to the similar results in both models. An interesting finding here is that the effect sizes of birthweight and its square is much stronger and robust for premenopausal women compared with postmenopausal women (Table 3). For example, the effect size of the square of birthweight is 0.109 (95%CIs: 0.033 ~ 0.185) for premenopausal women in Model 1, compared to 0.044 (95%CIs: -0.007 ~ 0.095) for postmenopausal women.

Table 3

Relationship between birthweight and breast cancer in Model 1 and Model 8 using continuous measurement of birthweight and discrete birthweight
	X→Y (Model 1) (case = 3,376, control = 105,012)	X + M₁ + M₂→Y (Model 8) (case = 2,247, control = 55,102)		X→Y (Model 1) case = 3,376, control = 105,012	X + M₁ + M₂→Y (Model 8) case = 2,247, control = 55,102
	beta (se, P)	beta (se, P)		beta (se, P)	beta (se, P)
X	-0.366 (0.141, 0.009)	-0.175 (0.176, 0.320)	X < 2.5	0.037 (0.054, 0.494)	-0.092 (0.090, 0.305)
X²	0.062 (0.022, 0.004)	0.037 (0.027, 0.173)	X > 4	0.202 (0.053, 1.37×10^− 4)	0.225 (0.102, 0.027)
C₁	0.714 (0.039, 9.09×10^− 75)		C₁	0.713 (0.039, 1.46×10^− 74)
C₂	0.068 (0.035, 0.054)	0.056 (0.044, 0.199)	C₂	0.067 (0.035, 0.057)	0.055 (0.044, 0.204)
C₃	0.004 (0.003, 0.292)	0.002 (0.004, 0.581)	C₃	0.004 (0.003, 0.273)	0.003 (0.004, 0.551)
M₁		-0.050 (0.014, 3.19×10^− 4)	M₁		-0.050 (0.014, 3.33×10^− 4)
M₂		-0.039 (0.004, 5.67×10^− 24)	M₂		-0.039 (0.004, 5.18×10^− 24)
C₁ = 0	case = 932, control = 46,098	case = 915, control = 45,038	C₁ = 0	case = 932, control = 46,098	case = 915, control = 45,038
X	-0.677 (0.253, 0.008)	-0.624 (0.261, 0.017)	X < 2.5	0.183 (0.103, 0.076)	0.172 (0.104, 0.099)
X²	0.109 (0.039, 0.006)	0.100 (0.040, 0.014)	X > 4	0.238 (0.103, 0.021)	0.243 (0.104, 0.019)
C₂	0.043 (0.066, 0.519)	0.056 (0.067, 0.404)	C₂	0.045 (0.066, 0.497)	0.058 (0.067, 0.390)
C₃	0.001 (0.006, 0.870)	-0.003 (0.006, 0.673)	C₃	0.001 (0.006, 0.839)	-0.003 (0.006, 0.693)
M₁		-0.054 (0.021, 0.012)	M₁		-0.054 (0.021, 0.012)
C₁ = 1	case = 2,444, control = 58,914	case = 2,247, control = 55,102	C₁ = 1	case = 2,444, control = 58,914	case = 2,247, control = 55,102
X	-0.244 (0.168, 0.146)	-0.175 (0.176, 0.320)	X < 2.5	-0.015 (0.064, 0.812)	-0.056 (0.067, 0.405)
X²	0.044 (0.026, 0.085)	0.037 (0.027, 0.173)	X > 4	0.189 (0.062, 0.002)	0.178 (0.065, 0.006)
C₂	0.078 (0.042, 0.062)	0.056 (0.044, 0.199)	C₂	0.076 (0.042, 0.068)	0.055 (0.044, 0.208)
C₃	0.005 (0.004, 0.205)	0.002 (0.004, 0.581)	C₃	0.005 (0.004, 0.246)	0.002 (0.004, 0.561)
M₁		-0.050 (0.014, 3.33×10^− 4)	M₁		-0.050 (0.014, 3.28×10^− 4)
M₂		-0.039 (0.004, 5.18×10^− 24)	M₂		-0.039 (0.004, 4.67×10^− 24)
Note: The left side shows results obtained with continuous birthweight, while the right side shows results obtained with discretized birthweight. X: birthweight, M₁: age at menarche, M₂: menopausal age; C₁: menopause; C₂: smoke; C₃: BMI. The intercept is not included here.

When incorporating discretized birthweight into the two models, we discover that women birth with weight < 2.5 or > 4.0 have a higher risk of breast cancer in adulthood compared to those with normal birthweight (2.5 ~ 4.0) in both Model 1 and Model 8 (Table 3). For instance, in terms of Model 8 the breast cancer risk increases approximately 25.2% (95%CIs: 2.54 ~ 52.9%) for women birth with weight > 4.0 across all age groups, which is more pronounced among premenopausal women compared to postmenopausal women (27.5% vs. 19.5%). These above findings indicate that relationship between birthweight and breast cancer is curved in a U-shaped form and varies in distinct age groups [3, 16].

4.1 Summary of the results

A larger number of prior studies have demonstrated the existence of possibly positive association between birthweight and breast cancer; however, inconsistent, sometimes even controversial, findings still emerged. Definitively determining that relationship is not a trivial task; in the present we intended to handle this challenging problem. As a result, we offer implicit answers for some key questions regarding the association between the two traits. First, to examine whether the observed relationship represents a linear causality, we applied the logistic regression but did not identify a linear causal association between birthweight and breast cancer, which is in agreement with the null finding obtained from another study published recently [26]. Second, to determine whether some growth traits and life processes may mediate the long-term impact of birthweight on breast cancer, we depended on the principle of mediation analysis [20, 21, 25, 27] and demonstrated that birthweight can indirectly influence breast cancer risk in adulthood via the path of age of menarche or age at menopause.

4.2 New contributions from our study

Compared with existing studies, the present work makes three new contributions to the relationship between birthweight and breast cancer. First, although it was implied by previous observations [3, 10, 16], there is no fully definitive conclusion regarding the non-linear association between the two traits. Here we explicitly revealed the existence of such non-linear association and showed that women birth with both low and high weight had higher risk to develop breast cancer in adulthood, reinforcing previous findings. However, the biological mechanisms underlying the association between birthweight and breast cancer are complex and have many possible explanations. For instance, birthweight is positively correlated with the estrogen level and the activity of the insulin-like growth factor-1 (IGF-1) level [14, 28]; whereas both estrogen and IGF-1 are thought to have important effect on fetal growth and breast development [29]; therefore, women with higher birthweight are more likely to develop breast cancer in later life [3, 30–33]. Besides the impact of high concentration of pregnancy estrogen, the role of other pregnancy hormones or intrauterine factors in the observed relationship cannot be excluded [4, 11, 34].

Second, prior studies suggested the relationship between birthweight and breast cancer might be distinct among various age groups [3], but many of them only analyzed premenopausal or postmenopausal cases of breast cancer [35]; it is not clear whether conclusions obtained from one age group can be generalized to other age groups. We here indeed found that the association was diverse between premenopausal and postmenopausal women. More specifically, we discovered that the effect size of birthweight was more evident and robust for premenopausal women compared to postmenopausal ones.

Third, we demonstrated that birthweight can indirectly influence breast cancer in adulthood via the path of age of menarche or age at menopause. This observation indicates that the fetal environments and growths during childhood and adolescence are important for the development of breast cancer in adult life [10, 12]. However, it needs to point out that we found that increased age at menopause had a protective influence on the development of breast cancer (β=-0.039 and P = 3.48×10^− 24). This finding is inconsistent with prior observations [36–38], which might be due to the biases by some unmeasured confounders (e.g., unreported oophorectomy). For instance, it was shown that oophorectomy had an evidently protective effect against breast cancer risk in BRCA1 and BRCA2 carriers and there existed a substantial trend in reducing hazard with increasing time since oophorectomy [39].

4.3 Limitations of this study

Our study is not without limitations. First, due to unavailability of many other early growth indicators (e.g., childhood obesity [13, 40]) in the UK Biobank cohort, we cannot further study their mediating role in the path from birthweight to adult breast cancer; therefore, the comprehensive causal path between them remains unclear. Second, like any retrospective studies and as mentioned above, our observational analysis might be biased by confounding factors such as unknown/unmeasured covariates, information and recall bias, which can undermine the validity of our results. Third, we focused only on European population; extending to other ancestral groups to validate our conclusions is warranted in the future.

This study provides an in-depth insight into the observed relationship between birthweight and breast cancer in later life, and reveals that non-linear impact and causal mediation commonly drive the connection between the two traits.

Acknowledgements

This study was also based on the UK Biobank resource under application number 57471. The UK Biobank was established by the Wellcome Trust medical charity, Medical Research Council, Department of Health, Scottish Government, and the Northwest Regional Development Agency. It has also had funding from the Welsh Assembly Government, British Heart Foundation and Diabetes UK. The data analyses in the present study were carried out with the high-performance computing cluster that was supported by the special central finance project of local universities for Xuzhou Medical University.

Funding

The research of Ping Zeng was supported in part by the National Natural Science Foundation of China (82173630 and 81402765), the Youth Foundation of Humanity and Social Science funded by Ministry of Education of China (18YJC910002), the Natural Science Foundation of Jiangsu Province of China (BK20181472), the China Postdoctoral Science Foundation (2018M630607 and 2019T120465), the QingLan Research Project of Jiangsu Province for Outstanding Young Teachers, the Six-Talent Peaks Project in Jiangsu Province of China (WSN-087), the Training Project for Youth Teams of Science and Technology Innovation at Xuzhou Medical University (TD202008), the Postdoctoral Science Foundation of Xuzhou Medical University, and the Statistical Science Research Project from National Bureau of Statistics of China (2014LY112). The research of Zhuanzhuan Liu was supported in part by the National Natural Science Foundation of China (2002157), the Natural Science Foundation of Jiangsu Province of China (BK20180994), and the China Postdoctoral Science Foundation (2018M632382).

Availability of data and materials

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Conflict of interest

The authors declare that they have no competing interests.

Author Contribution

PZ and ZL conceived the idea for the study. PZ obtained the data. PZ, JQ and MZ cleared up the datasets; PZ, JQ and MZ performed the data analyses. PZ and MZ interpreted the results of the data analyses. PZ, JQ, MZ and ZL wrote the manuscript with the participation of other authors.

Adami HO et al (1995) The aetiology and pathogenesis of human breast cancer. Mutat Res 333(1–2):29–35
Anbazhagan R, Gusterson BA (1994) Prenatal factors may influence predisposition to breast cancer. Eur J cancer (Oxford England: 1990) 30A(1):1–3
Sanderson M et al (1996) Perinatal factors and risk of breast cancer. Epidemiology, vol 7. Cambridge, Mass.),, pp 34–37. 1
Kaijser M et al (2001) In utero exposures and breast cancer: a study of opposite-sexed twins. J Natl Cancer Inst 93(1):60–62
Michels KB et al (2006) Longitudinal study of birthweight and the incidence of breast cancer in adulthood. Carcinogenesis 27(12):2464–2468
Xu X Birth weight as a risk factor for breast cancer: a meta-analysis of 18 epidemiological studies. Journal of women's health et al (2002) 2009. 18(8): p. 1169–1178
Zhou W et al (2020) Birth Weight and Incidence of Breast Cancer: Dose-Response Meta-analysis of Prospective Studies. Clin Breast Cancer 20(5):e555–e568
Wu AH, McKean-Cowdin R, Tseng C-C (2011) Birth weight and other prenatal factors and risk of breast cancer in Asian-Americans. Breast Cancer Res Treat 130(3):917–925
Vatten LJ et al (2002) Birth weight as a predictor of breast cancer: a case-control study in Norway. Br J Cancer 86(1):89–91
Mellemkjær L et al (2003) Birth weight and risk of early-onset breast cancer (Denmark). Cancer Causes Control 14(1):61–64
Ahlgren M et al (2003) Birth weight and risk of breast cancer in a cohort of 106,504 women.International Journal of Cancer, 107(6)
Ahlgren M et al (2004) Growth Patterns and the Risk of Breast Cancer in Women. N Engl J Med 351(16):1619–1626
dos Santos Silva I et al (2004) Is the association of birth weight with premenopausal breast cancer risk mediated through childhood growth? Br J Cancer 91(3):519–524
Swerdlow AJ et al (2018) Maternal breast cancer risk in relation to birthweight and gestation of her offspring. Breast cancer research: BCR 20(1):110
Barber LE et al (2019) Pre- and perinatal factors and incidence of breast cancer in the Black Women's Health Study. Cancer causes & control: CCC 30(1):87–95
da Cruz RS et al (2018) Paternal malnutrition programs breast cancer risk and tumor metabolism in offspring. Breast cancer research: BCR 20(1):99
Bukowski R et al (2012) Birth weight, breast cancer and the potential mediating hormonal environment. PLoS ONE 7(7):e40199
Sudlow C et al (2015) UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 12(3):e1001779
Yang S, Zhou X (2020) Accurate and Scalable Construction of Polygenic Scores in Large Biobank Data Sets. Am J Hum Genet 106(5):679–693
VanderWeele T (2015) Explanation in causal inference: methods for mediation and interaction. Oxford University Press
Zeng P, Shao Z, Zhou X (2021) Statistical methods for mediation analysis in the era of high-throughput genomics: current successes and future challenges. Comput Struct Biotechnol J 19:3209–3224
Baron RM, Kenny DA (1986) The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Personal Soc Psychol 51(6):1173–1182
Daniel RM et al (2015) Causal mediation analysis with multiple mediators. Biometrics 71(1):1–14
Ihaka R, Gentleman R (1996) R: A Language for Data Analysis and Graphics. J Comput Graphical Stat 5(3):299–314
MacKinnon DP (2008) Introduction to statistical mediation analysis. Routledge
Kar SP et al (2019) The association between weight at birth and breast cancer risk revisited using Mendelian randomisation. Eur J Epidemiol 34(6):591–600
MacKinnon DP et al (2002) A comparison of methods to test mediation and other intervening variable effects. Psychol Methods 7(1):83–104
Petridou E et al (1990) Tobacco smoking, pregnancy estrogens, and birth weight. Epidemiology 1(3):247–250
Wang HS, Chard T (1992) The role of insulin-like growth factor-I and insulin-like growth factor-binding protein-1 in the control of human fetal growth. J Endocrinol 132(1):11–19
Kazer RR (1995) Insulin resistance, insulin-like growth factor I and breast cancer: a hypothesis. Int J Cancer 62(4):403–406
Pezzino V et al (1996) Insulin-like growth factor-I (IGF-I) receptors in breast cancer. Ann N Y Acad Sci 784:189–201
McCormack VA et al (2003) Fetal growth and subsequent risk of breast cancer: results from long term follow up of Swedish cohort. BMJ (Clinical research ed. 326:248–2487383
Park SK et al (2006) Intrauterine environment and breast cancer risk in a population-based case‐control study in Poland. Int J Cancer 119(9):2136–2141
Ekbom A et al (1997) Intrauterine environment and breast cancer risk in women: a population-based study. J Natl Cancer Inst 89(1):71–76
Luo J et al (2020) Birth weight, weight over the adult life course and risk of breast cancer. Int J Cancer 147(1):65–75
Monninkhof EM, van der Schouw YT, Peeters PH (1999) Early age at menopause and breast cancer: are leaner women more protected? A prospective analysis of the Dutch DOM cohort. Breast Cancer Res Treat 55(3):285–291
Parks RM Breast Cancer Epidemiology, in Breast Cancer Management for Surgeons: A European Multidisciplinary Textbook, L. Wyld, Editors et al (2018) Springer International Publishing: Cham. p. 19–29
Lipworth L (1995) Epidemiology of breast cancer.European journal of cancer prevention: the official journal of the European Cancer Prevention Organisation (ECP), 4(1)
Chang-Claude J et al (2007) Age at menarche and menopause and breast cancer risk in the International BRCA1/2 Carrier Cohort Study. Cancer epidemiology, biomarkers & prevention: a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. 16:740–7464
Ruder EH et al (2008) Examining breast cancer growth and lifestyle risk factors: early life, childhood, and adolescence. Clin Breast Cancer 8(4):334–342

SupplementaryFile.docx
Additional FileSupplementary Files.

Download PDF

Version 1

posted

You are reading this latest preprint version

Investigating the relationship between birthweight and breast cancer from a non-linear and mediation perspective

Status:

Version 1

Abstract

Figures

Highlights

1. Introduction

2. Methods

2.1 Individual-level UK Biobank dataset

2.3 Evaluating linear and non-linear relationship between birthweight and breast cancer

3. Results

3.1 Mediation paths between birthweight and breast cancer

3.2 Non-linear relationship between birthweight and breast cancer

4. Discussion

4.1 Summary of the results

4.2 New contributions from our study

4.3 Limitations of this study

5. Conclusions

Declarations

References

Supplementary Files

Status:

Version 1