Amiodarone

The data for the entire cohort of 1340 subjects including all fixed and time-dependent variables was represented in counting process notation suitable for Cox regression of time-dependent exposure [11 ]. Multiple records (with consecutive start and end times) were created for each subject to account for every change in exposure to any of the time-dependent variables over the study period.
The hazard ratio (HR) of pacemaker insertion associated with amiodarone doses >200 mg per day was estimated using a Cox proportional-hazards model including all fixed and time-dependent covariates. The timescale used in the model was time since first prescription of amiodarone. Non-significant variables (other than age and sex) were sequentially removed if the resultant model had no significant increase in Akaike Information Criteria (AIC) and no significant change in the HR for amiodarone dose.
The nested case-control approach was also used to estimate the HR of pacemaker insertion associated with amiodarone doses >200 mg per day. Cases of pacemaker insertion were identified and controls were randomly selected from the risk-set of each case (i.e. subjects present in the cohort at the time the case is defined). After selecting all controls and recording their index dates (i.e. the time, in cohort time, at which the respective case is defined) the relevant time-dependent covariate information was retrieved by merging with the database configured in counting process notation. The relevant subject record was selected by requiring that the index date fall within the start and end time of the subject record for each control.
The nested case-control approach was repeated using 4, 8, 16, 32, and 64 controls per case. For each number of controls per case, random sampling of controls for all cases and conditional logistic regression analysis was repeated 100 times using the OUTEST option in the PROC PHREG statement to create an output SAS data containing all the parameter estimates [11 ]. The mean and standard deviation (SD) of the parameter estimates for each number of controls per case was calculated.
Computational times for regression models of time-dependent exposures using nested case-control and survival analysis methodologies were compared. The nested case-control samples with 4 and 32 controls per case were analyzed using conditional logistic regression with the PHREG procedure in SAS Release 8.2 [12 ]. The full cohort was analyzed using Cox regression adapted for analysis of time-dependent covariates with the PHREG procedure in SAS Release 8.2. Ties were handled using the TIES = EFRON option in the PHREG procedure [11 ]. All analyses were performed using an Intel Pentium 4 computer with a 1.80 GHz central processing unit (CPU) and 256 MB of random access memory (RAM).
Relative computational efficiencies were evaluated by comparing the CPU times of the three regression models used to analyze the cohort. Relative increases in computational time as a function of sample size were quantified by repeating the analyses on progressively larger cohorts. This was done by progressive doubling of the original cohort to 2, 4, 8, and 32 times its original size. Given that the objective was to compare computational times, all fixed and time-dependent exposures were included in all models regardless of statistical significance.
The computational efficiency of the Cox regression model was also compared to the conditional logistic regression model where the nested case-control sample included all possible controls for each case. This comparison was performed for the original cohort of 1340 subjects with 53 cases (including 2 ties). All analyses were performed using the PHREG procedure. Ties were handled using the TIES = EFRON option in the PHREG procedure, and subsequently using the TIES = DISCRETE option in the PHREG procedure for comparison.

Thyroid parameters (PTFQI, TSH, and fT4) of a sample of patients with atrial fibrillation were compared to a reference sample of healthcare patients (case-control study). Patients with atrial fibrillation (paroxistic or persistent) who required, at least, one hospital admission were included as cases. The control sample PTFQI distribution tertiles were used as cut-offs for dividing cases and controls. We compared atrial fibrillation odds across the tertiles.
Within the cases sample, additionally, a cross-sectional study of association of the thyroid parameters with clinical characteristics was conducted (cross-sectional study).
We reviewed the 165 clinical records of the patients with atrial fibrillation who were admitted to the Miguel Servet University Hospital in Zaragoza (Spain) from July 2017 to June 2019. Exclusion criteria included previous thyroid disease or abnormal TSH and fT4 (which was the main cause of exclusion, 72%) and diseases or pharmacological treatments that interfere with the thyroid axis, among which amiodarone use accounted for the 35% of excluded pacients (Supplementary Material Table 1). Only patients who had TSH and fT4 values available before medical intervention (electrical cardioversion, catheter ablation, or treatment with amiodarone) were included. The number of patients identified for the analysis was 84 (Supplementary Material Figure 1).
Thyroid parameters of healthcare patients older than 18 years who underwent a thyroid hormone measurement across three months of 2018, in the central laboratory of the Miguel Servet University Hospital (6051 patients) were used to calculate the reference parameters for the PTFQI formula (14 (link)). Euthyroid participants of this sample constituted the control group (n=5256).
This study protocol was approved by the Ethical Committee of Clinical Research of Aragón (CEICA) (expedient numbers 19-041 and 19-519) who authorized review of clinical records.

This study was designed as a multicenter retrospective study comprising 180,582 adults from 5 health check-up centers from 2010 to 2018. All individuals had access to serum FT3, FT4, and TSH concentrations, as well as MAFLD diagnosis. Moreover, there were 3 centers in the north of China and 2 in the south of China. Participants who 1) had previously been diagnosed with liver cancer, liver cirrhosis, or had a history of liver surgery (N=227); 2) had the administration of drugs influencing serum TH levels, such as methimazole, propylthiouracil, levothyroxine, and amiodarone (N=24); 3) had a severe medical illness, such as acute infection, acute heart failure, acute coronary syndrome, stroke, severe kidney diseases, and malignancy (N=2791) were excluded. Finally, 177540 adults were enrolled in our multicentered retrospective study. The flow chart of participant selection is shown in Figure 1A.
This study was approved by the ethical review committee of Renmin Hospital of Wuhan University and followed by acceptance by the ethics center in each collaborating hospital. The ethics committees granted a waiver of the requirement for documentation of informed consent for just analyzing existing data after anonymization without individual identification.

De-duplication of identical documents from multiple databases through literature management software such as Endnote. Selection criteria for duplicate literature. (1) Selection of the one with the largest sample size. (2) Selection of the one with the longest follow-up period. (3) Select the one with the most comprehensive study outcomes. The following were the conditions for inclusion: (1) adult population; (2) a study that explored the association between dairy product consumption and the likelihood of developing NAFLD, such as cohort studies, case–control studies or cross-sectional studies; (3) a diagnosis of NAFLD made by ultrasound, magnetic resonance imaging (MRI), controlled attenuation parameter (CAP), fibro-scan, fatty liver index, or liver biopsy; (4) consumption of dairy products: total dairy, milk, yogurt, cheese, or other types of dairy products; (5) outcome was NAFLD; (6) language was restricted to English. Exclusion criteria were as follows: (1) teenaged or pregnant participants; (2) surface antigens for hepatitis B, antibodies against hepatitis C, or HIV antibodies present; (3) abnormally high alcohol or drug intake that could be hazardous to the liver (tamoxifen, steroids, and amiodarone); and (4) reviews, comments, editorials, letters, interviews, or reports.