Identification of risk factors and development of a predictive model for foetal growth restriction: a retrospective case–control study Authors: Xiu-Fang Shao Published: 2026 ABSTRACT AbstractObjectiveTo identify risk factors associated with foetal growth restriction (FGR) and develop a predictive model to support early diagnosis and individualized clinical management strategies.MethodsA retrospective case-control study was conducted at Wusibei Campus of Fujian Maternity and Child Health Hospital, Fuzhou, China involving 490 pregnant women (196 FGR cases and 294 controls) who delivered between January and December 2024. Participants were categorized into two groups: a control group with normal foetal growth and an FGR group with confirmed FGR (This retrospective case–control study selected participants according to pregnancy outcomes and analysed clinical data retrospectively). Clinical characteristics and maternal-neonatal outcomes were compared. Univariate and multivariate logistic regression analyses were performed using SPSS 26.0 to determine independent risk factors for FGR. A predictive nomogram was constructed using R software (version 4.1.0) with the rms package. Model performance was evaluated via bootstrap internal validation, and discrimination and calibration were assessed using receiver operating characteristic (ROC) curve analysis and calibration plots.ResultsThe FGR group demonstrated significantly higher incidences of foetal distress, caesarean delivery, preterm birth, and neonatal hospitalization compared to the control group (all p < 0.001). Multivariate analysis identified the following independent risk factors for FGR: reduced placental thickness on first-trimester nuchal translucency ultrasound, low pregnancy-associated plasma protein A levels, hypertensive disorders of pregnancy, gestational hypothyroidism, oligohydramnios, and absence of gestational diabetes mellitus. The resulting nomogram demonstrated moderate predictive capability with an area under the ROC curve of 0.722.ConclusionThe predictive model developed in this study may facilitate the early identification of high-risk pregnancies, enabling timely intervention to improve maternal and neonatal outcomes. However, external validation of this model is needed before clinical implementation.Keywords: Foetal growth restriction, high-risk factors, pregnancy outcome, predictive factors, maternal-neonatal prognosis Introduction Foetal growth restriction (FGR) is a prevalent obstetric complication, with a global incidence of approximately 10%, and higher rates reported in developing countries [1]. FGR is associated with increased risks of perinatal morbidity and mortality [2]. Globally, foetuses affected by growth restriction experience a foetal mortality rate of about 1.5%—nearly double that observed in appropriately grown foetuses—and this risk rises to ∼2.5% when foetal weight falls below the 5th percentile for gestational age [2,3]. Furthermore, epidemiological studies have demonstrated that FGR is linked to adverse long-term outcomes, including neurodevelopmental delays in childhood and increased adult susceptibility to cardiovascular and endocrine disorders such as coronary artery disease, type 2 diabetes mellitus, and obesity [4,5]. The aetiology of FGR is multifactorial, encompassing maternal, foetal, and placental factors; however, a substantial proportion of cases remain unexplained [3]. Despite advancements in obstetric care, more than 50% of FGR cases are not diagnosed prior to delivery, even in high-resource healthcare settings [6]. Currently, no curative treatments exist for FGR once diagnosed. Clinical management primarily focuses on close foetal surveillance, identification and mitigation of contributory factors, and timely delivery to minimize perinatal complications. Early recognition of at-risk pregnancies is critical for optimizing outcomes, and evidence suggests that proactive management can significantly reduce adverse perinatal events [7]. However, standard screening approaches—including maternal risk factor assessment, fundal height measurements, and uterine artery Doppler ultrasonography—have limited sensitivity for detecting intrauterine growth restriction. Placental dysfunction represents the most prevalent underlying cause of FGR. During early gestation, the placenta secretes multiple biomolecules into the maternal circulation, some of which correlate with aberrant foetal growth trajectories. First-trimester biochemical screening markers such as pregnancy-associated plasma protein A (PAPP-A), free β-human chorionic gonadotropin (β-hCG), and soluble fms-like tyrosine kinase-1 (sFlt-1) have shown potential as non-invasive indicators of placental insufficiency [8–10]. Despite these advances, there remains a lack of integrated early-pregnancy prediction models that combine a placental morphological marker (e.g. first-trimester placental thickness) with functional biochemical markers (such as PAPP-A) and maternal clinical factors, particularly in Asian populations. Current screening often relies on late-gestation foetal biometry, limiting opportunities for early intervention. To address this knowledge gap, the present study aimed to identify independent early-pregnancy risk factors for FGR and to develop an internally validated nomogram for early prediction of FGR. Specifically, our objectives were (1) to identify independent risk factors for FGR, and (2) to develop and internally validate a predictive model (nomogram) based on these factors for early detection of at-risk pregnancies. Materials and methods Study population This retrospective case–control study included 490 pregnant women who delivered at the Wusibei Campus of Fujian Maternity and Child Health Hospital between January and December 2024. Data were obtained from the hospital’s medical records department. The diagnostic criteria for FGR were defined as a pathological condition in which the fetus fails to achieve its full growth potential due to maternal, foetal, or placental abnormalities. The diagnosis was confirmed via ultrasonography using the Hadlock formula for estimated foetal weight, with an estimated foetal weight or abdominal circumference below the 10th percentile for gestational age, consistent with established criteria [3]. Inclusion criteria for the observation group (FGR group) were: ① singleton live birth; ② availability of complete medical records; ③ FGR diagnosis confirmed by ultrasound criteria; ④ neonatal birth weight below the 10th percentile for gestational age based on the national Chinese population birth weight standard. Inclusion criteria for the control group (non-FGR group) were: ① singleton live birth; ② availability of complete medical records. The control group (n = 294) was selected by simple random sampling from all eligible pregnancies with normal foetal growth delivered during the same period, aiming for approximately a 1.5:1 control-to-case ratio. This approach ensured that controls were representative of the base population and delivered in the same timeframe as FGR cases, minimizing selection bias. The following exclusion criteria were applied to both groups: ① incomplete or missing medical history; ② multiple gestation; ③ history of uterine surgery or presence of a uterine scar; ④ pre-existing maternal comorbidities (chronic hypertension, diabetes mellitus, cardiovascular disease, autoimmune disorders); ⑤ documented psychiatric illness. As incomplete records were excluded by these criteria, there were no missing data for the variables of interest in the final analysis dataset. Research methods A single-centre retrospective case-control study was conducted, enrolling 490 pregnant women. Participants were categorized into two groups according to foetal growth status. Group A (control group) included 294 women with normal foetal growth (222 primiparas and 72 multiparas). The median maternal age was 29.00 years (interquartile range [IQR]: 27.00–31.00), and the median gestational age at delivery was 39.57 weeks (IQR: 38.86–40.14). Group B (FGR group) included 196 women with pregnancies complicated by FGR (171 primiparas and 25 multiparas). The median maternal age in this group was 29.00 years (IQR: 26.00–31.00), and the median gestational age at delivery was 39.00 weeks (IQR: 38.14–39.43). The study protocol was approved by the Institutional Ethics Committee (Approval No. 2025KY031). Given the retrospective design and use of anonymized data, the requirement for informed consent was waived in accordance with institutional guidelines and applicable regulations. Observation indicators Clinical Characteristics: Maternal demographic and clinical data included age (years) and body mass index (BMI, kg/m2). First-trimester ultrasound parameters included placental thickness (cm) measured during the nuchal translucency (NT) examination between 11 and 13 weeks of gestation. Biochemical markers analysed were maternal serum free β-hCG (ng/mL) and PAPP-A (U/L) levels. Pregnancy-related complications recorded included gestational diabetes mellitus (GDM), hypertensive disorders of pregnancy (HDP), gestational hypothyroidism, intrahepatic cholestasis of pregnancy, premature rupture of membranes, polyhydramnios, oligohydramnios, and maternal anemia. Documented fetal and placental abnormalities included battledore placenta, velamentous cord insertion, nuchal cord, and umbilical cord torsion. Third-trimester Doppler ultrasound measurements of the systolic/diastolic ratios (S/D ratios) of the middle cerebral artery and umbilical artery were evaluated. Maternal and Neonatal Outcomes: Assessed obstetric outcomes included the presence of meconium-stained amniotic fluid, placental abruption, and foetal distress. Delivery-related outcomes included preterm birth, mode of delivery (caesarean section), and postpartum haemorrhage. Neonatal outcomes evaluated were neonatal asphyxia and neonatal hospitalization. Statistical methods Statistical analyses were performed using SPSS version 26.0. Continuous variables following a normal distribution are presented as mean ± standard deviation and compared using independent samples t-tests. Non-normally distributed data are expressed as median with IQR and analysed with the Mann-Whitney U test. Categorical variables are presented as frequencies and percentages (n (%)) and compared using chi-square (χ2) tests. Variables demonstrating statistically significant differences in univariate analyses were subsequently included in multivariable logistic regression models to determine independent risk factors for FGR. A predictive nomogram was developed based on the statistically significant predictors from the regression results using R software (version 4.1.0) with the rms package. The model was internally validated via the bootstrap resampling method, and its performance was evaluated through ROC curve analysis for discrimination and calibration curve analysis for model calibration. Post-hoc sample size adequacy was assessed using the Events Per Variable (EPV) criterion. With 196 FGR events and 6 predictors retained in the final nomogram model, the EPV was 32, exceeding the recommended minimum of 10–20 events per predictor, indicating a sufficient sample size to ensure model stability and minimize overfitting risk. A p-value < 0.05 was considered statistically significant. Results Comparison of clinical indicators In total, 490 women were included in the analysis (196 FGR cases and 294 controls). During the study period (January–December 2024), 11,657 women delivered at our hospital. Among these, 508 preg­nancies (4.3%) met the ultrasound criteria for suspected FGR. After applying the exclusion criteria (e.g. incomplete data or foetal anomalies), 196 cases were confirmed as FGR and included in the final analysis. During the same period, approximately 11,150 pregnancies with normal foetal growth were recorded. After applying the same exclusion criteria (eliminating roughly 650 cases with incomplete data or other exclusions), about 10,500 normal-growth pregnancies remained eligible. From these, 294 were randomly selected as the control group (approximately a 1.5:1 control-to-case ratio). The analysis revealed statistically significant differences between the FGR and control groups across multiple clinical indicators (all p < 0.05). The FGR group had a significantly reduced median placental thickness during the first-trimester NT ultrasound (1.35 cm vs 1.50 cm; p < 0.001), lower first-trimester free β-hCG levels (median: 60.25 vs 69.75; p = 0.015), lower PAPP-A levels (median: 1.36 vs 1.91; p < 0.001), and a higher umbilical artery S/D ratio in late pregnancy (median: 2.40 vs 2.17; p < 0.001). Regarding pregnancy complications, the FGR group had a lower incidence of GDM (8.16% vs 15.65%; p = 0.021), a higher prevalence of HDP (13.78% vs 0.34%; p < 0.001), increased rates of gestational hypothyroidism (4.59% vs 1.02%; p = 0.027), and a greater occurrence of oligohydramnios (16.84% vs 4.08%; p < 0.001). No significant differences were observed in other recorded clinical parameters (detailed data are presented in Table 1). Table 1.Comparison of clinical indicators between two groups.Clinical IndicatorGroup A (n = 294)Group B (n = 196)p-valueAge (years)29.00 (27.00, 31.00)29.00 (26.00, 31.00)0.757BMI (kg/m²)25.95 (24.13, 28.28)25.94 (23.73, 28.11)0.423Placental thickness (cm)1.50 (1.30, 1.70)1.35 (1.20, 1.60)<0.001Free β-hCG (ng/mL)69.75 (44.90, 105.00)60.25 (41.85, 85.55)0.015PAPP-A (U/L)1.91 (1.12, 3.07)1.36 (0.83, 2.36)<0.001Middle cerebral artery S/D ratio3.96 (3.30, 4.72)3.75 (3.25, 4.44)0.108Umbilical artery S/D ratio2.17 (2.00, 2.38)2.40 (2.16, 2.62)<0.001Gestational diabetes mellitus (%)46 (15.65%)16 (8.16%)0.021Hypertensive disorders of pregnancy (%)1 (0.34%)27 (13.78%)<0.001Gestational hypothyroidism (%)3 (1.02%)9 (4.59%)0.027Intrahepatic cholestasis of pregnancy (%)2 (0.68%)3 (1.53%)0.646Premature rupture of membranes (%)68 (23.13%)35 (17.86%)0.197Polyhydramnios (%)4 (1.36%)0 (0.00%)0.260Oligohydramnios (%)12 (4.08%)33 (16.84%)<0.001Anaemia in pregnancy (%)7 (2.38%)3 (1.53%)0.744Battledore placenta (%)12 (4.08%)10 (5.10%)0.755Velamentous cord insertion (%)5 (1.70%)2 (1.02%)0.816Nuchal cord (%)27 (9.18%)23 (11.73%)0.446Umbilical cord torsion (%)18 (6.12%)18 (9.18%)0.273 Comparison of clinical indicators between two groups. Comparison of maternal and neonatal outcomes Significant differences were observed in several key maternal and neonatal outcomes between the FGR and control groups (all p < 0.05). The FGR group had substantially higher rates of foetal distress (22.96% vs 4.08%; p < 0.001), caesarean section (54.08% vs 23.47%; p < 0.001), preterm birth (9.18% vs 0.34%; p < 0.001), and neonatal hospitalization (35.71% vs 12.24%; p < 0.001). However, there were no significant differences between the groups in the incidence of placental abruption, meconium-stained amniotic fluid, postpartum haemorrhage, or neonatal asphyxia (all p > 0.05). Complete outcome data are presented in Table 2. Table 2.Comparison of maternal and neonatal outcomes between two groups.OutcomeGroup A (n = 294)Group B (n = 196)p-valueFoetal distress (%)12 (4.08%)45 (22.96%)<0.001Caesarean section (%)69 (23.47%)106 (54.08%)<0.001Preterm birth (%)1 (0.34%)18 (9.18%)<0.001Placental abruption (%)1 (0.34%)4 (2.04%)0.169Meconium-stained amniotic fluid (%)19 (6.46%)14 (7.14%)0.912Postpartum haemorrhage (%)5 (1.70%)0 (0.00%)0.169Neonatal asphyxia (%)3 (1.02%)4 (2.04%)0.586Neonatal hospitalization (%)36 (12.24%)70 (35.71%)<0.001 Comparison of maternal and neonatal outcomes between two groups. Multivariate analysis of high-risk factors for FGR A multivariable logistic regression model was constructed with FGR status as the dependent variable (FGR = 1, Control = 0). Independent variables included all factors that were statistically significant in univariate analyses: placental thickness (NT ultrasound), first-trimester free β-hCG concentration, PAPP-A level, late-pregnancy umbilical artery S/D ratio, GDM, HDP, gestational hypothyroidism, and oligohydramnios. The analysis identified the following independent predictors of FGR: decreased placental thickness on NT ultrasound (p = 0.004); reduced first-trimester PAPP-A levels (p = 0.020); presence of HDP (p < 0.001); presence of gestational hypothyroidism (p = 0.030); presence of oligohydramnios (p < 0.001); and absence of GDM (p = 0.007). Notably, free β-hCG and the umbilical artery S/D ratio were significant in univariate analysis but did not remain significant in the multivariable model. Detailed regression coefficients and odds ratios are presented in Table 3. Table 3.Multivariate logistic regression analysis of risk factors for FGR (full multivariable model).Predictorβ coefficientSEWald χ²OR95% CIp-valuePlacental thickness (cm)–1.0730.3768.1620.3420.164–0.7140.004Free β-hCG (ng/mL)–0.0020.0021.7060.9980.994–1.0010.191PAPP-A (U/L)–0.1520.0655.4290.8590.756–0.9760.020Umbilical artery S/D ratio–0.0990.0891.2490.9050.761–1.0780.264Gestational diabetes mellitus–0.9370.3507.1590.3920.197–0.7780.007Hypertensive disorders of pregnancy3.8251.03713.60245.8506.004–350.117<0.001Gestational hypothyroidism1.5770.7294.6824.8401.160–20.1940.030Oligohydramnios1.5320.36817.3584.6282.251–9.516<0.001Note. FGR was coded as 1 and the control group as 0. Binary variables were coded as 1 = yes and 0 = no. The intercept of the complete multivariable logistic regression model corresponding to this table was 1.835. The full model equation is provided in the manuscript text. Multivariate logistic regression analysis of risk factors for FGR (full multivariable model). Note. FGR was coded as 1 and the control group as 0. Binary variables were coded as 1 = yes and 0 = no. The intercept of the complete multivariable logistic regression model corresponding to this table was 1.835. The full model equation is provided in the manuscript text. To facilitate model replication, the complete multivariable logistic regression equation corresponding to Table 3 was as follows: The predicted probability was calculated as P = exp[logit(P)]/{1 + exp[logit(P)]}. P denotes the predicted probability of FGR. Placental thickness was entered in cm, free β-hCG was entered in ng/mL, PAPP-A was entered in U/L, and the umbilical artery S/D ratio was entered as a continuous variable. GDM, HDP, gestational hypothyroidism, and oligohydramnios were coded as 1 = yes and 0 = no. FGR was coded as 1 and the control group as 0. Because GDM was coded as 1 = yes, its negative coefficient indicates that the presence of GDM was associated with a lower predicted probability of FGR in this cohort; equivalently, absence of GDM was associated with increased FGR risk. Development of a nomogram prediction model for FGR A nomogram was constructed based on the six statistically significant predictors identified in the multivariable analysis: placental thickness (NT ultrasound), PAPP-A level, presence of HDP, gestational hypothyroidism, oligohydramnios, and absence of GDM. The full multivariable regression equation, including all variables entered into the model, is reported above to facilitate replication; the nomogram focuses on the statistically significant predictors to provide a concise visual risk-assessment tool. In the nomogram (Figure 1), each predictor is assigned a point value, and the total score corresponds to a predicted probability of FGR. This visual tool facilitates individualized risk assessment and aids in clinical decision-making regarding early intervention strategies. Nomogram for predicting FGR risk based on six statistically significant predictors. Validation of the FGR nomogram model The model underwent internal validation using bootstrap resampling (n = 1000). Model performance was evaluated via ROC curve analysis and calibration plotting. The ROC analysis (Figure 2) yielded an area under the ROC curve (AUC) of 0.722 (95% CI: 0.675–0.768), demonstrating satisfactory discriminatory ability. The calibration plot (Figure 3) showed close agreement between predicted probabilities and observed outcomes, indicating good model calibration. Receiver operating characteristic (ROC) curve of the predictive model. Calibration curve of the FGR predictive nomogram. Discussion In our study, pregnancies complicated by FGR had markedly higher rates of intrapartum fetal distress, caesarean delivery, preterm birth, and neonatal hospitalization compared to normal-growth pregnancies. These adverse outcomes reflect the downstream effects of placental insufficiency and chronic intrauterine hypoxia diminishing the foetus’s tolerance to labour. Consequently, growth-restricted foetuses are more prone to intrapartum heart rate decelerations (foetal distress), often necessitating emergent caesarean delivery to prevent hypoxic injury [11,12]. Furthermore, the compromised intrauterine environment and associated maternal complications (e.g. HDP) frequently lead to medically indicated preterm delivery when foetal or maternal status deteriorates [13,14]. Previous studies have reported that FGR neonates are at increased risk of complications such as hypoglycaemia, respiratory distress, and infection, which may contribute to the increased neonatal hospitalization observed in our cohort [15]. This corresponds with the significantly increased neonatal hospitalization rate observed in the FGR group. Given these risks, pregnancies affected by FGR warrant enhanced surveillance (e.g. frequent foetal heart rate monitoring, serial growth ultrasounds) to enable timely intervention and improve perinatal outcomes. The pathogenesis of FGR is multifactorial and often involves placental dysfunction, maternal comorbidities, and foetal genetic abnormalities. In the absence of definitive clinical treatment, management primarily focuses on identifying underlying aetiologies, continuous foetal surveillance, and optimization of delivery timing. Early recognition of risk factors permits estimation of the likelihood of FGR development and facilitates implementation of individualized management strategies to potentially improve pregnancy outcomes. In this study, univariate and multivariate logistic regression analyses identified several independent risk factors for FGR, including reduced placental thickness measured during NT ultrasound, decreased first-trimester serum levels of PAPP-A, hypertensive disorders of pregnancy, gestational hypothyroidism, and oligohydramnios. These findings are consistent with previously published literature and align with established pathophysiological mechanisms underlying FGR[3,16]. Notably, although free β-hCG was associated with FGR in univariate analysis, it did not remain significant in the multivariable model. This may be due to overlapping predictive information with other variables—PAPP-A and placental thickness, in particular. PAPP-A appears to be a more specific marker of early placental insufficiency in our cohort, likely capturing much of the predictive variance that free β-hCG contributed in isolation. Consequently, when both markers were included, PAPP-A retained significance whereas free β-hCG did not. The placenta serves as the critical interface mediating nutrient and oxygen exchange between the maternal and foetal circulations. Reduced placental thickness likely reflects diminished villous development, poor vascularization, or other structural abnormalities, all of which compromise placental function and restrict foetal growth [17,18]. PAPP-A, a syncytiotrophoblast-derived protease, plays a key role in foetal development by regulating the bioavailability of insulin-like growth factors (IGFs). Decreased first-trimester serum levels of PAPP-A indicate underlying placental dysfunction, as this enzyme is crucial for IGF activation and nutrient transport. PAPP-A deficiency may disrupt these processes, leading to inadequate foetal nutrient supply and subsequent growth restriction [19]. An elevated umbilical artery S/D ratio indicates increased placental vascular resistance and compromised fetoplacental blood flow, contributing to FGR [20]. However, in our analysis, UA S/D was significant in univariate testing but lost significance in the multivariate model after adjusting for confounding factors (such as coexisting HDP) that independently affect both Doppler indices and foetal growth [21]. Thyroid hormones play a vital role in foetal development, particularly in neurological and skeletal maturation. Gestational hypothyroidism may disrupt foetal metabolic regulation and reduce the secretion of growth-promoting hormones, thereby increasing the risk of FGR [22]. Oligohydramnios often indicates placental insufficiency; a reduced amniotic fluid volume can limit foetal movement and cause cord compression, further impairing foetal blood flow and nutrient delivery and thereby exacerbating FGR [23]. Notably, this study identified GDM as a potential protective factor against FGR. The apparent protective effect of maternal hyperglycaemia likely occurs via foetal hyperinsulinemia, which promotes anabolic growth. This finding is consistent with reports from other populations; however, the apparent protective effect of GDM may be negated when GDM co-occurs with hypertensive disorders of pregnancy [24]. Based on the statistically significant predictors identified, we developed a nomogram predictive model for FGR. The model underwent internal validation using bootstrap sampling (n = 1000), and its predictive performance was assessed via ROC curve analysis and calibration curve analysis. The model demonstrated moderate discriminatory ability, with an AUC of 0.722 (95% CI: 0.675–0.768), indicating acceptable predictive performance for an early screening tool. This tool enables rapid, individualized estimation of FGR risk during pregnancy, offering practical utility in clinical settings. Strengths and limitations Strengths: Key strengths of this study include the integration of both biophysical and biochemical markers in early pregnancy (combining ultrasound-measured placental thickness with maternal serum analytes like PAPP-A), a relatively robust sample size relative to model complexity (reflected by an EPV of 32, well above recommended thresholds), and the use of internal bootstrap validation, which demonstrated excellent calibration of the model. Limitations: The study’s retrospective single-center design may introduce selection bias and limits the generalizability of our findings. Additionally, we lacked an external validation cohort, so the model’s performance in other populations remains untested. Finally, potential unmeasured confounders (for example, maternal nutritional status or smoking habits) were not captured in our dataset, which could influence the observed associations. The very wide confidence interval for HDP reflects sparse-data instability due to the small number of HDP cases in the control group; therefore, this estimate should be interpreted as directional rather than as a precise effect size. Conclusion This study identified six independent risk factors for FGR and used them to develop an early-pregnancy predictive nomogram, which achieved a moderate discriminatory ability upon internal validation (AUC = 0.722). While this model shows promise in facilitating early identification of high-risk pregnancies, it should be interpreted with caution and requires prospective external validation before it can be recommended for routine clinical use. Acknowledgements The author thanks all staff members of Fujian Maternity and Child Health Hospital who contributed to clinical care and medical record management associated with this study. Funding Statement No external funding was received for this study. Author’s contributions CRediT: Xiu-Fang Shao: Conceptualization, Data curation, Investigation, Methodology, Resources, Writing – original draft, Writing – review & editing. Ethics approval and consent to participate This study was conducted with approval from the Ethics Committee of Fujian Maternity and Child Health Hospital (No. 2025KY031). The study was performed in accordance with the Declaration of Helsinki. Given the retrospective design and use of anonymized data, the requirement for informed consent was waived in accordance with institutional guidelines and applicable regulations. Consent for publication Not applicable. Clinical trial registration Not applicable. Disclosure statement No potential conflict of interest was reported by the authors. Availability of data and materials The datasets used and analysed during the current study are available from the corresponding author on reasonable request. References Associated Data Data Availability Statement The datasets used and analysed during the current study are available from the corresponding author on reasonable request.