Restricted Cubic Spline Regression Research Articles

Assessing the longitudinal association between the GGT/HDL-C ratio and NAFLD: a cohort study in a non-obese Chinese population

BackgroundA cross-sectional association between the combination indicator of high-density lipoprotein cholesterol (HDL-C) and gamma-glutamyl transferase (GGT) and fatty liver has been described in several recent studies, and this study aims to further evaluate the longitudinal relationship between the ratio of GGT to HDL-C (GGT/HDL-C ratio) and nonalcoholic fatty liver disease (NAFLD).MethodsThis cohort study included 12,126 individuals without NAFLD at baseline, followed prospectively for 5 years, and the endpoint of interest was new-onset NAFLD. The relationship of the GGT/HDL-C ratio with new-onset NAFLD and the shape of the association was assessed by Cox regression models and restricted cubic spline (RCS) regression, respectively. Time-dependent receiver operator characteristics (ROC) curves were constructed to evaluate the predictive value of GGT, HDL-C, GGT/HDL-C ratio and BMI for the occurrence of NAFLD at different time points in the future.ResultsThe prevalence of NAFLD was 72.46/1000 person-years during the 5-year follow-up period. Results of multivariate Cox regression analysis showed a positive association of the GGT/HDL-C ratio with new-onset NAFLD after adequate adjustment of the related confounding factors, and the degree of correlation was slightly higher than that of GGT, and further subgroup analysis found that this association was more significant in the population with elevated systolic blood pressure (SBP). In addition, we also found a nonlinear relationship of the GGT/HDL-C ratio with the risk of new-onset NAFLD using the RCS regression, where the saturation threshold was about 31.79 U/mmol. Time-dependent ROC analysis results showed that the GGT/HDL-C ratio was increasingly valuable in predicting NAFLD over time, and was better than HDL-C in predicting NAFLD in the early stage (1–3 years), but was not superior to BMI and GGT.ConclusionsIn this large longitudinal cohort study based on a Chinese population, our results supported that the GGT/HDL-C ratio was positively and nonlinearly associated with the risk of new-onset NAFLD in a non-obese population. In the assessment of future NAFLD risk, the GGT/HDL-C ratio was slightly better than GGT alone; However, the GGT/HDL-C ratio did not appear to have a significant advantage over GGT and BMI alone in predicting NAFLD.

The impact of sleep-corrected social jetlag on mental health, behavioral problems, and daytime sleepiness in adolescents

Study objectivesThis study aimed to examine the effect of sleep-corrected social jetlag (SJLsc) on mental health, behavioral problems, and daytime sleepiness in adolescents. MethodsThis was a cross-sectional study which included 4787 adolescents (Mean age: 14.83±1.6y, 56.0% girls) recruited from 15 secondary schools in Hong Kong. SJLsc was defined as the absolute difference between sleep-corrected midsleep on weekdays and weekends, at which the sleep debt has been considered. It was classified into three groups: low-level (“LSJLsc”, <1h), mid-level (“MSJLsc”, ≥1h and <2h), and high-level of SJLsc (“HSJLsc”, ≥2h). Adolescents’ mental health, behavioral problems and daytime sleepiness were measured by the General Health Questionnaire (GHQ-12), the Strengths and Difficulties Questionnaire (SDQ) and the Pediatric Daytime Sleepiness Scale (PDSS). Logistic regression analysis and restricted cubic spline regression (RCS) analysis were applied with consideration of confounders including age, gender, puberty and sleep problems. ResultsNearly half (46.9%) of adolescents had SJLsc for at least 1 h. Greater SJLsc was associated with more behavioral difficulties (MSJLsc: OR: 1.20, p = 0.03; HSJLsc: OR: 1.34, p = 0.02) when controlling for age, sex, puberty, chronotype, insomnia, and time in bed. There was a dose-response relationship in which higher SJLsc had an increased risk of conduct problems and hyperactivity, while only high-level SJLsc was associated with a peer relationship problem. In RCS analysis, SJLsc was associated with a higher likelihood of behavioral difficulties (p = 0.03) but not poor mental health or daytime sleepiness. ConclusionsSleep-corrected social jetlag was a unique risk factor for behavioral problems in adolescents. Our findings highlighted the need for interventions to promote healthy sleep-wake patterns in school adolescents.

Modeling non-linear relationships in epidemiological data: The application and interpretation of spline models.

Traditional methods to deal with non-linearity in regression analysis often result in loss of information or compromised interpretability of the results. A recommended but underutilized method for modeling non-linear associations in regression models is spline functions. We explain spline functions in a non-mathematical way and illustrate the application and interpretation to an empirical data example. Using data from the Amsterdam Growth and Health Longitudinal Study, we examined the non-linear relationship between the sum of four skinfolds and VO2max, which are measures of body fat and cardiorespiratory fitness, respectively. We compared traditional methods (i.e., quadratic regression and categorization) to spline methods [1- and 3-knot linear spline (LSP) models and a 3-knot restricted cubic spline (RCS) model] in terms of the interpretability of the results and their explained variance (). The spline models fitted the data better than the traditional methods. Increasing the number of knots in the LSP model increased the explained variance (from for the 1-knot model to for the 3-knot model). The RCS model fitted the data best (), but results in regression coefficients that are harder to interpret. Spline functions should be considered more often as they are flexible and can be applied in commonly used regression analysis. RCS regression is generally recommended for prediction research (i.e., to obtain the predicted outcome for a specific exposure value), whereas LSP regression is recommended if one is interested in the effects in a population.

Association between kaolin-induced maximum amplitude and slow-flow/no-reflow in ST elevation myocardial infarction patients treated with primary percutaneous coronary intervention

BackgroundST-segment elevation myocardial infarction (STEMI) patients with a high thrombus burden have a relatively high slow-flow/no-reflow risk. However, the association between kaolin-induced maximum amplitude (MAthrombin) and slow-flow/no-reflow has been scarcely explored. MethodsSTEMI patients treated with primary percutaneous coronary intervention (PCI) were retrospectively enrolled from January 2015 to December 2019 at China-Japan Friendship Hospital. MAthrombin levels were measured using thromboelastography before the PCI procedure. The patients were divided into two groups according to thrombolysis in myocardial infarction (TIMI) flow grade after primary PCI: the normal flow group (TIMI flow grade = 3) and slow-flow/no-reflow (TIMI flow grade ≤ 2). The logistic regression model and restricted cubic spline regression (RCS) were used to analyze the predictive value of MAthrombin for slow-flow/no-reflow. All patients were followed up after discharge and observed the adverse cardiovascular events between the two groups. ResultsA total of 690 patients were enrolled, with 108(15.7%) having slow-flow/no-reflow. The multivariate logistic regression model analysis showed that MAthrombin level was an independent risk factor for slow-flow/no-reflow. The RCS analysis showed a nonlinear relationship between MAthrombin levels and slow-flow/no-reflow. The cut-off value of MAthrombin levels for predicting slow-flow/no-reflow was 68 mm. During a median follow-up time of 4.4 years, slow-flow/no-reflow (hazard ratio 1.93, 95% confidence interval 1.27–2.93, P = 0.002) and MAthrombin levels (hazard ratio 1.06, 95% confidence interval 1.03–1.08, P < 0.001) were independent risk factors for predicting the long-term of adverse clinical cardiovascular events. ConclusionMAthrombin was an independent risk factor for predicting slow-flow/ no-reflow in STEMI patients who underwent primary PCI.

Association of aldehyde exposure with bone mineral density in the national health and nutrition examination survey (NHANES 2013-2014).

The association between aldehyde exposure and bone health in humans remains unclear. This study was to evaluate the association of serum aldehydes with bone mineral density (BMD) and osteopenia/osteoporosis. We analyzed the US National Health and Nutrition Examination Survey cross-sectional data from 2013 to 2014. Weighted multivariate-adjusted linear regression and logistic regression models were used to assess the association between specific aldehydes and osteopenia/osteoporosis. Associations between aldehyde combinations and BMD were also evaluated using the restricted cubic spline (RCS) method. Compared with men in the first tertile, those in the third tertile of propanaldehyde concentration were negatively associated with proximal femur and lumbar spine BMD. Significant inverse associations were observed between benzaldehyde exposure and trochanter BMD in women. Benzaldehyde increased the risk of osteopenia/osteoporosis 2.75-fold [95% confidence interval (CI) = 1.06, 7.11] in the highest tertile in women compared to the lowest tertile concentration. In males, the prevalence of total femur, femur neck, and trochanter osteopenia/osteoporosis was significantly higher in the highest versus the lowest tertile of propanaldehyde exposure, with odds ratios (ORs) of 6.84 (95% CI = 2.33, 20.04), 2.72 (95% CI = 1.18, 6.27), and 3.26 (95% CI = 1.25, 8.56), respectively. RCS regression also showed decreased BMD continuously with increasing serum mixed aldehyde levels. Serum aldehyde concentrations were associated with low BMD and high osteopenia/osteoporosis risk in adults, with propanaldehyde and benzaldehyde being the most critical. Co-exposure to aldehyde combinations was negatively correlated with BMD.

Relationships between sodium levels, haemodynamics and metalloproteinases in heart failure patients.

To estimate the associations between dysnatraemia and inflammatory marker [including interleukin-6 (IL-6)], and tissue remodelling marker [matrix metalloproteinase (MMP)-9 and tissue inhibitor of MMP (TIMP)-1], the pulmonary capillary wedge pressure (PCWP), mean pulmonary artery pressure (PAP), and left ventricular end-diastolic pressure (EDP), and the prognostic relevance in patients with heart failure. The serum sodium level and circulating levels of IL-6, MMP-9, and TIMP-1 were measured in 173 heart failure patients. Dual heart catheterisation was performed to measure PCWP, mean PAP, and EDP. All-cause mortality was assessed during the follow-up period (mean 88 ± 49months). Restricted cubic spline (RCS) regression showed a U-shaped association of serum sodium level with TIMP-1, with the lowest values in the 138-140mmol/L range (P for effect = 0.042, P for non-linearity = 0.017). IL-6 and MMP-9 levels showed non-significant associations with serum sodium level. U-shaped associations of serum sodium level with PCWP (P for effect = 0.004, P for non-linearity = 0.001) and mean PAP (P for effect = 0.042, P for non-linearity = 0.017) were found with the RCS regression model. The random forest model revealed that TIMP-1, MMP-9, and IL-6 were important predictors for serum sodium levels. Restricted cubic spline Cox regressions demonstrated that TIMP-1 levels indicated a U-shaped, concaved, non-linear association with all-cause mortality (P for effect = 0.011, P for non-linearity = 0.022). Dysnatraemia is an index of TIMP-1 aggravation and elevated PCWP, mean PAP; hence, it is associated with worsening all-cause mortality.Clinical Trial Registration: UMIN000023840.

Association of waist-to-height ratio with estimated glomerular filtration rate in middle-aged and elderly Chinese.

This study aimed to assess the relationship between waist-to-height (WHtR) and estimated glomerular filtration rate (eGFR) in men, non-menopausal, and postmenopausal women among middle-aged and elderly Chinese. This study analyzed the data of 7,807 participants in a cross-sectional survey, ie, the third wave of the China Health and Retirement Longitudinal Study. Restrictive cubic-spline regression with three knots was used to assess the dose-response association of WHtR with eGFR. Piecewise linear regression models were further established to calculate the slope of each segment and their 95% confidence interval (CI). After adjusting for potential confounders, an inverse L-shaped dose-response relationship was found between WHtR and eGFR among men (Pnonlinear = 0.024, threshold = 0.513) and postmenopausal women (Pnonlinear = 0.009, threshold = 0.503). The slopes on the right sides of the threshold were statistically significant among men (β2 = -33.77, 95% CI: -53.23 to -14.31) and postmenopausal women (β2 = -36.53, 95% CI: -49.71 to -23.35), respectively. A weak negative linear relationship existed between WHtR and eGFR in non-menopausal women. The relationship between WHtR and eGFR tended to be inverse-L-shaped in men and postmenopausal women, but may vary with postmenopausal status in women.

Pre-pregnancy body mass index and time to pregnancy among couples pregnant within a year: A China cohort study.

BackgroundExtreme pre-pregnancy body mass index (BMI) values have been associated with reduced fecundability and prolonged time to pregnancy in previous studies. However, the effect in fertile couples is unclear.ObjectivesThis study aimed to evaluate the association between pre-pregnancy BMI and fecundability, measured as time to pregnancy (TTP), among couples that achieved pregnancy within 1 year.MethodsThis was a retrospective cohort study of 50,927 couples wishing to conceive, enrolled in the National Free Preconception Health Examination Project (NFPHEP) in Chongqing, China, during 2012–2016. Participants’ weight and height were measured by NFPHEP-trained preconception guidance physicians. TTP measured in months was used to determine subfecundity (TTP >6 months). The strength of association between BMI and TTP/subfecundity was measured with fecundability odds ratios (FOR)/odds ratios (OR) and their corresponding 95% confidence intervals (CI), calculated with Cox and logistic regression analysis. We used restricted cubic spline regression (RCS) to test the observed FOR trends.ResultsCompared to women with normal BMI, women with pre-pregnancy overweight/obesity had longer TTP (FOR = 0.96, 95% CI: 0.94–0.99) and increased risk of subfecundity (OR = 1.08, 95% CI: 1.00–1.17). There was no association between TTP and male BMI. RCS trends varied when data were stratified by male pre-pregnancy BMI, with the greatest change detected in pre-pregnancy underweight men.ConclusionsPre-pregnancy overweight/obesity was associated with longer TTP and subfecundity among women who became pregnant within 1 year; this effect was likely mediated by their partners’ pre-pregnancy BMI. These findings indicate that BMI could affect fecundability, independently of affecting the risk of sterility. Advice on weight management and maintaining healthy weight should be included in couples’ preconception guidance.

Exploring the associations of serum concentrations of PCBs, PCDDs, and PCDFs with walking speed in the U.S. general population: Beyond standard linear models

Studies have shown that persistent organic pollutants (POPs) can have various health effects. However, little is known about the effects of multiple chemicals with possible common sources of exposure on walking speed, a proxy index reflecting lower limb neuromuscular function and physical function. We simultaneously applied multiple linear and nonlinear statistical models to explore the complex exposure-response relationship between a mixture of 22 selected POPs and walking speed. A total of 14 polychlorinated biphenyls (PCBs), 3 polychlorinated dibenzo-p-dioxins (PCDDs), and 5 polychlorinated dibenzofurans (PCDFs) were measured in the serum of participants in the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2002. Walking speed was measured during a physical examination. Linear regression (LR), least absolute shrinkage and selection operator (LASSO), and group LASSO were used to evaluate the linearity of mixtures, while restricted cubic spline (RCS) regression, random forest (RF), and Bayesian kernel machine regression (BKMR) models were used to evaluate the nonlinearity of mixtures. Potential confounders were adjusted in the above models. A total of 436 subjects were included in our final analysis. The results of the LR model did not identify any POP exposure that was significantly associated with walking speed. The LASSO results revealed an inverse association of one PCDD congener and two PCDF congeners with walking speed, while the group LASSO analysis identified PCDFs at the exposure level and at the group level. In the RCS analysis, two PCB congeners presented significant overall associations with walking speed. The PCB congener PCB194 showed statistically significant effects on the outcome (P = 0.01) when a permutation-based RF was used. The BKMR analysis suggested that PCBs and PCDFs (probabilities = 0.887 and 0.909, respectively) are potentially associated with walking speed. Complex statistical models, such as RCS regression, RF and BKMR models, can detect the nonlinear and nonadditive relationships between PCBs and walking speed, while LASSO and group LASSO can identify only the linear relationships between PCDFs and walking speed. Fully considering the influence of collinearity in each method during modelling can increase the comprehensiveness and reliability of conclusions in studies of multiple chemicals.

Correlation between fasting blood glucose and hepatocarcinogenesis: a multicentre retrospective study (A report of 94 264 cases)

Correlation between fasting blood glucose and hepatocarcinogenesis: a multicentre retrospective study (A report of 94 264 cases)

Correlation between different body mass indexes and incidence of digestive carcinoma: a multicentre retrospective study (A report of 95 177 cases)

Correlation between different body mass indexes and incidence of digestive carcinoma: a multicentre retrospective study (A report of 95 177 cases)

Predictive value of serum uric acid on new-onset cholelithiasis (a report of 97 469 cases)

Objective To explore the predictive value of serum uric acid on new-onset cholelithiasis. Methods The retrospective cohort study was conducted. The data of 97 469 subjects who participated health examination at the Kailuan General Hospital Affiliated to the North China University of Science and Technology, Kailuan Linxi Hospital, Kailuan Zhaogezhuang Hospital, Kailuan Tangjiazhuang Hospital, Kailuan Fan′gezhuang Hospital, Kailuan Lyujiatuo Hospital, Kailuan Jinggezhuang Hospital, Kailuan Linnancang Hospital, Kailuan Qianjiaying Hospital, Kailuan Majiagou Hospital and Kailuan Branch Hospital from June 2006 to December 2015 were collected. Epidemiological investigation, anthropometric parameters and biochemical indicators were collected. All the subjects were allocated into 4 groups according to squartiles of serum uric acid: 24 140 with serum uric acid <232 μmol/L in the Q1 group, 24 473 with 232 μmol/L≤ serum uric acid <282 μmol/L in the Q2 group, 24 382 with 282 μmol/L≤ serum uric acid <338 μmol/L in the Q3 group and 24 474 with serum uric acid ≥338 μmol/L in the Q4 group. Observation indicators: (1) comparisons of clinical characteristics among the 4 groups; (2) incidence of cholelithiasis in the 4 groups; (3) effects of serum uric acid on the new-onset cholelithiasis: ① the dose-response relationship between serum uric acid and the risk of cholelithiasis, ② comparisons of the fitting degree of serum uric acid on the cholelithiasis model, ③ effects of different serum uric acid levels on incidence of cholelithiasis after stratification by sex, ④ serum uric acid of different gender on the boxplots, ⑤ effects of different serum uric acid levels on the incidence of cholelithiasis after stratification by age. Measurement data with normal distribution were expressed as ±s, and comparisons among groups were analyzed using the one-way ANOVA. Measurement data with skewed distribution is expressed by M (Q), and comparisons among groups were analyzed using the nonparametric Krustal-willis test. Count data were represented by percentage, and comparisons among groups were analyzed using chi-square test. The incidences of cholethiasis in 4 groups of different serum uric acid were calculated by person-year incidence. Restrictive cubic spline regression was used to calculate the dose-response relation between the continuous variable and the risks of new-onset cholelithiasis and 95% confidence interval (CI). COX regression model was used to analyze the hazard ratio (HR) and 95%CI of different serum uric acid levels on new-onset cholelithiasis. Likelihood ratio test and akaike information criterion (AIC) were used to calculate the fitting degree of serum uric acid on new-onset cholelithiasis model. Boxplots were used to describe serum uric acid in different genders. Results (1) comparisons of clinical characteristics among the 4 groups: sex (male), age, body mass index (BMI), systolic pressure, diastolic pressure, fasting plasma glucose (FPG), total cholesterol (TC), triglyceride (TG), high sensitive C-reactive protein, diabetes, hypertension, smoking, drinking and physical exercise were 15 162, (50±11)years, (24±3)kg/m2, (123±21)mmHg (1 mmHg=0.133 kPa), (82±12)mmHg, (5.6±2.0)mmol/L, (4.8±1.2)mmol/L, 1.14 mmol/L (range, 0.81-1.63 mmol/L), 0.70 mmol/L (range, 0.23-2.23 mmol/L), 2 537, 9 415, 4 575, 2 380, 2 649 in the Q1 group, 19 079, (51±12)years, (25±3)kg/m2, (130±21)mmHg, (83±12)mmHg, (5.5±1.7)mmol/L, (4.9±1.2)mmol/L, 1.20 mmol/L (range, 0.86-1.76 mmol/L), 0.71 mmol/L (range, 0.28-1.98 mmol/L), 2 287, 10 124, 6 918, 3 649, 3 288 in the Q2 group, 21 132, (52±13)years, (25±3)kg/m2, (132±21)mmHg, (84±12)mmHg, (5.5±1.6)mmol/L, (5.0±1.2)mmol/L, 1.29 mmol/L (range, 0.91-1.94 mmol/L), 0.80 mmol/L (range, 0.30-2.06 mmol/L), 2 027, 10 755, 8 259, 4 730, 3 958 in the Q3 group, 22 651, (53±14)years, (26±3)kg/m2, (134±21)mmHg, (85±12)mmHg, (5.4±1.5)mmol/L, (5.1±1.2)mmol/L, 1.54 mmol/L (range, 1.05-2.35 mmol/L), 1.02 mmol/L (range, 0.43-2.50 mmol/L), 1 981, 12 082, 9 562, 6 209, 4 758 in the Q4 group, respectively, with statistically significant differences among the 4 groups (χ2=7 624.63, F=279.93, 961.91, 330.84, 271.40, 38.25, 353.18, H=3 406.30, 912.23, χ2=108.15, 590.49, 2567.07, 2 209.21, 760.15, P 0.05). In male participants, adjusted for age, BMI, TC, TG, diabetes, hypertension, smoking, drinking and physical exercise, risks of new-onset cholelithiasis in Q2, Q3 and Q4 groups were increased compared with Q1 group (HR=1.17, 1.24, 1.30, 95%CI: 1.06-1.30, 1.12-1.37, 1.18-1.44, P 0.05), however, risk of new-onset cholelithiasis was increased in the Q4 group (HR=1.15, 95%CI: 1.02-1.28, P 60 years old, adjusted for sex, BMI, TC, TG, diabetes, hypertension, smoking, drinking and physical exercise, risk of new-onset cholelithiasis in the Q2 groups was not increased compared with Q1 group (HR=1.16, 95%CI: 0.99-1.36, P>0.05), however, risks of new-onset cholelithiasis were increased in the Q3 and Q4 groups (HR=1.19, 1.21, 95%CI: 1.02-1.40, 1.04-1.41, P<0.05). Conclusion Elevated serum uric acid is an independent risk factor for the new-onset cholelithiasis. Key words: Cholelithiasis; Serum uric acid; Incidence; Risk factor