TWTiansheng Wang

Tiansheng Wang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Tiansheng Wang has a total of 183 co-authors. Here are the top co-authors by number of shared publications:

  • Til Stürmer

    Department of Epidemiology, Gillings School of Global Public Health University of North Carolina at Chapel Hill Chapel Hill North Carolina USA

    15 shared publications
  • John B. Buse

    University of North Carolina School of Medicine, Chapel Hill, USA

    11 shared publications
  • Virginia Pate

    Department of Epidemiology, The University of North Carolina at Chapel Hill Gillings School of Global Public Health, Chapel Hill, North Carolina, USA

    9 shared publications
  • Huilin Tang

    Center for Preventative Ophthalmology and Biostatistics, Department of Ophthalmology University of Pennsylvania Perelman School of Medicine Philadelphia Pennsylvania USA

    8 shared publications
  • Til Stürmer

    Department of Epidemiology, Gillings School of Global Public Health University of North Carolina at Chapel Hill Chapel Hill North Carolina USA

    7 shared publications
  • Jeff Yufeng Yang

    Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA

    6 shared publications

Latest Publications

  • Adapting the prevalent new-user study design and clone-censor-weighting approach to evaluate drug discontinuation

    by Qoua Her, Til Stürmer, Robert J Glynn, Emilie Duchesneau, Tiansheng Wang, Charles Poole, Jennifer Leigh Lund, Michael Webster-Clark

    We described how to adapt the prevalent new-user (PNU) study design to estimate the effect of drug discontinuation among patients who discontinue and clone-censor-weighting (CCW) approach to estimate the effect of drug discontinuation among an eligible population at a pre-specified time point, focusing on the differences between the target trials the 2 approaches emulate. While both approaches produced unbiased estimates of the effect of drug discontinuation when all confounders were adjusted for in simulations, the true estimands differed, particularly when treatment effect heterogeneity was introduced. We also applied the approaches to study the effect of statin discontinuation among United States Medicare beneficiaries who discontinued statin therapy after at least 5 years (PNU) or at approximately 5 years (CCW) of continuous use. The PNU analysis estimated a mortality rate difference (MRD) of 3.6 (95% CI, 3.3-3.9) per 100 person-years and a mortality rate ratio (MRR) of 1.41 (95% CI, 1.37-1.45), while the CCW analysis estimated a MRD of 3.4 (95% CI, 2.8-4.0) per 100 person-years and MRR of 1.55 (1.45-1.66). Researchers studying drug discontinuation should carefully consider their causal question and target population when choosing between these 2 newer approaches.

  • Open Access

    The heterogeneous treatment effects of statins on dementia: a target trial emulation with causal machine learning using integrated genetic and real‐world data

    by Yongjie Lai, Qoua L Her, Yue Zhang, Hong Xu, Anna Kucharska‐Newton, Dongze Ji, Guorong Wu, Gwenn Garden, Til Stürmer, Tiansheng Wang, et al.

    INTRODUCTION: Given the complexity of dementia, the inconsistent evidence on statins and dementia highlights the need for robust methods to assess heterogeneous treatment effects (HTEs). METHODS: We emulated a target trial using UK Biobank comparing statin initiators and non‐initiators aged ≥55 years. Marginal structural models were fitted to estimate 5‐year adjusted risk difference (aRD). We used iterative causal forest, a causal machine learning subgrouping algorithm, to identify subgroups with HTEs. RESULTS: Among 18,366 participants, the overall aRD for all‐cause dementia was −1.0‰ (95% CI: −4.2‰ to 2.3‰). We identified subgroups by polygenic risk score for Alzheimer's disease (AD) excluding apolipoprotein E ( APOE ) genotype (“non‐ APOE PRS”). Participants with high non‐ APOE PRS showed cognitive benefit (all‐cause dementia: aRD −5.9‰, 95% CI: −8.1‰ to 1.2‰; AD: aRD −5.0‰, 95% CI: −8.2‰ to −0.2‰). DISCUSSION: Participants with high non‐ APOE PRS may benefit from statins, suggesting genetic susceptibility beyond APOE could modify statins' cognitive effects.


  • Active Comparator New User Cohort Studies and Matching

    by Til Stürmer, Tiansheng Wang

    This Guide to Statistics and Methods provides an overview of the active comparator, new user design and the use of propensity score matching.


  • Open Access

    Integrate Real‐World Data and Genetics: a causal machine learning approach to assess Statin’s Effect on Incident Dementia

    by Tiansheng Wang, Yongjie Lai, Elizabeth C. Mormino, Guorong Wu, Gwenn A Garden, Til Stürmer, Yang Xu

    Background: The impact of statins on dementia has been debated for over two decades. Population differences may modify statins effects, contributing to inconsistent findings. Approaches to identify patients most likely to benefit are needed. We aim to apply iterative causal forest (iCF) causal machine learning algorithm in emulated trials in UK Biobank (UKB) to assess heterogeneous treatment effects (HTEs) of statins on incident dementia. Method: We conducted an emulated target trial using 36 monthly sequential trials (2009–2011) with linked UKB primary care data comparing statins initiators and non‐initiators aged 55+. Incident dementia was identified using UKB’s algorithmically‐defined outcomes and Read Version 2 codes. Patients were followed from baseline until dementia, death, loss to follow‐up, or end of primary care (5/31/2016–8/31/2017). Confounders and potential effect modifiers include genetics, socioeconomic status, lifestyle, comorbidities, and medications. Genetic modifiers included Apolipoprotein E (APOE) genotype, the polygenic risk score (PRS) for Alzheimer’s disease excluding the APOE region (nonAPOE‐PRS), and the PRS for statins response predicting low‐density lipoprotein cholesterol reduction after statins use (statin‐PRS). We performed propensity score (PS)‐matching on the 36 (stacked) sequential trials, censored upon treatment deviation in per‐protocol analysis, accounting for selection using inverse probability of censoring weights (IPCW). Next, we applied iCF to PS‐matched trials to identify subgroups with HTEs, and estimated 5‐year adjusted risk difference (aRD) by inverse probability treatment and censoring weights in each identified subgroup. Result: The overall aRD was ‐0.01% (‐0.20% to 0.22%), risks of 0.68% vs. 0.69%, in PS matched cohort (9175 statins initiators vs 9175 non‐statins initiators). iCF revealed HTEs ( P ‐value = 0.001) and identified patients with a high [> second tertile] nonAPOE‐PRS and healthy dietary habits (³ 4 of 7 criteria, see Figure 2. legend) benefit most from statins (aRD ‐1.18% [‐2.20% to ‐0.21%], risk 0.59% vs 1.77%), while those with high nonAPOE‐PRS alone exhibited a smaller benefit (aRD ‐0.47% [‐0.87% to ‐0.05%], risk 0.74% vs 1.21%). Conclusion: Our emulated trial indicated that statins did not demonstrate a protective effect against incident dementia in overall population. However, a subpopulation with high nonAPOE‐PRS (which predicts AD dementia) and healthy dietary habits may benefit from statins use.

  • Open Access

    Investigating the association between incretin-based therapies and thyroid cancer incidence among US Medicare beneficiaries with diabetes

    by Clement O Acheampong, John B Buse, Klara R Klein, Lawrence T Kim, Joshua Evron, Anna R Kahkoska, Caroline A Thompson, Tiansheng Wang, Virginia Pate, Peter Leese, et al.

    Introduction: Preclinical studies suggest a potential link between glucagon-like peptide 1 receptor agonists (GLP-1RA) and thyroid cancer (TC), yet it is unclear if this risk translates to humans. Research design and methods: We estimated the comparative effect of incretin-based therapies (GLP-1RA and dipeptidyl-peptidase-4 inhibitors (DPP-4i)) versus sodium-glucose cotransporter-2 inhibitors (SGLT-2i) on TC incidence among US older adults with type 2 diabetes. We defined TC as a thyroidectomy followed by ≥2 separate diagnoses codes for malignant neoplasm of thyroid gland within 90 days. We estimated adjusted 3-year cumulative risk differences of TC (aRDs) with 95% CIs using weighted Kaplan-Meier survival functions, and adjusted HRs using weighted Cox models. Results: We included 73 388 new users in the GLP-1RA versus SGLT-2i cohort (mean age 72.4 years, men: 48.3%) and 106 274 in the DPP-4i versus SGLT-2i cohort (mean age 74.6 years, men: 44.9%). At 3 years and a median duration of treatment of 0.82–1.15 years, the aRD for GLP-1RA versus SGLT-2i for TC was −23 per 10 000 (95% CI: −51 to 4) and the aRD for DPP-4i versus SGLT-2i was −2 per 10 000 (95% CI: −17 to 13). Secondary and sensitivity analyses were consistent. Conclusions: Our study of US Medicare beneficiaries with type 2 diabetes suggests that the initiation of incretin-based therapies may not increase the 3-year risk of TC compared with initiation of SGLT-2i. This finding offers reassurance for short-term use but does not eliminate the possibility of increased long-term or subtype-specific risks.


  • Open Access

    Estimating cardiovascular effects of influenza vaccination in older adults: a target trial emulation using proximal causal inference

    by Jinxin Guo, Tiansheng Wang, Zhike Liu, Weihong Zeng, Peng Shen, Yexiang Sun, Siyan Zhan, Yang Xu


  • Risk of suicidal ideation and suicidality among adults prescribed semaglutide for weight management: A population‐based cohort study

    by Qoua L. Her, Tiansheng Wang, Til Stürmer, John B. Buse, Michele Jonsson Funk, Virginia Pate, Michael Webster‐Clark

    Aims: To evaluate the risk of incident suicidal ideation (SI) and suicidality (SI , suicide attempt or intentional self‐harm) among overweight or obese adults who were prescribed semaglutide for weight management compared with active comparators (weight management drugs: bupropion/naltrexone, orlistat, phentermine or phentermine/topiramate ). Materials and Methods: Using the Merative® MarketScan Databases, we conducted a retrospective cohort study of adults initiating semaglutide or an active comparator between 01 June 2021, and 31 December 2022. Patients were required to have ≥1 year of continuous health and pharmacy plan coverage, a diagnosis of overweight or obesity, no prior suicidality or type 2 diabetes, and new users of semaglutide or active comparators for inclusion. SI and suicidality were identified using diagnosis codes within 183 days of drug initiation. Kaplan–Meier analysis was used to estimate risks, risk differences (RDs), and risk ratios (RRs), controlling for confounding with standardized mortality ratio weights. Results: Among 16 822 semaglutide and 11 986 active comparator new users, the adjusted risk of incident SI was 0.08% (13.0 events) for semaglutide and 0.05% (8.9 events) for active comparators, and for incident suicidality was 0.08% (14.0 events) for semaglutide and 0.07% (11.0 events) for active comparators. Adjusted RDs were 0.02% (95% CI: [−0.02%, 0.07%]) for SI and 0.02% (−0.03%, 0.07%) for suicidality, with corresponding RRs of 1.36 (0.62, 6.14) and 1.18 (0.57, 3.63). Conclusion: Among patients initiating semaglutide for weight management, semaglutide did not appear to increase the risk of suicide‐related outcomes. However, the CIs of the RDs are compatible with a 0.07% increase in 6‐month risk and the true RD could be larger due to the potentially low sensitivity of ICD‐10 diagnosis codes for and underreporting of suicide‐related outcomes. Additional studies are critical to confirm these findings and obtain more precise estimates of any potential benefits or harms of semaglutide.


  • Developing a novel algorithm to identify incident and prevalent dementia in Medicare claims—the ARIC Study

    by Tiansheng Wang, Virginia Pate, Dae Hyun Kim, Melinda C Power, Gwenn Garden, Priya Palta, David Knopman, Michelle Jonsson-Funk, Til Stürmer, Anna M Kucharska-Newton

    There is an urgent need to improve dementia ascertainment robustness in real-world studies assessing drug effects on dementia risk. We developed algorithms to dementia identification algorithms using Medicare claims (inpatient/outpatient/prescription) from 3318 Visit 5 (2011-2013) and 1828 Visit 6 (2016-2017) participants of the Atherosclerosis Risk in Communities (ARIC) Study, validated against ARIC's rigorous syndromic dementia classification. Algorithm performance was compared to existing algorithms (Jain, Bynum, Lee). We further evaluated algorithms effectiveness in a 20% random Medicare sample aged ≥70 years who initiating liraglutide or dipeptidyl peptidase 4 inhibitors (DPP4i) to assess 3-year adjusted risk difference (aRD) for dementia. Our incident dementia algorithm required two dementia diagnostic codes within 1-year, or one dementia code plus a new dementia prescription within 90 days. It achieved a positive predictive value (PPV) of 69.2%, specificity of 99.0%, and sensitivity of 34.6% (population prevalence: 8.8%), comparable to extant algorithms (PPV, 58.7–68.6%; sensitivity 25.5–40.4%). Prevalent dementia algorithm (without requiring incident diagnoses/prescriptions) demonstrated similar performance. In the Medicare sample, dementia risk ranged from 3.0% to 12.5%, aRD comparing liraglutide to DPP4i varied −1.2% to −3.6%, with our algorithm closely matching the Bynum algorithm. Algorithm selection significantly impacts treatment effect estimates, highlighting its importance in in pharmacoepidemiologic research.


  • Glucagon‐like Peptide‐1 Receptor Agonists in Asthma Exacerbations: An Application of High‐Dimensional Iterative Causal Forest to Identify Subgroups

    by Tiansheng Wang, Jeanny H. Wang, Alan C. Kinlaw, Richard Wyss, Virginia Pate, Zhuoyue Gou, John B. Buse, Corinne A. Keet, Michael R. Kosorok, Til Stürmer

    Background: Glucagon‐like Peptide‐1 Receptor Agonists (GLP1RA) may reduce asthma exacerbation (AE) risk, but it is unclear which populations benefit most. Recent pharmacoepidemiologic studies have employed iterative causal forest (iCF), a machine learning (ML) algorithm, to identify subgroups with heterogeneous treatment effects (HTEs). While iCF does not rely on prior knowledge of treatment‐variable interactions, it may be constrained by missing or poorly defined variables in pharmacoepidemiologic studies.Methods: We applied the high‐dimensional iterative causal forest (hdiCF)—a causal ML algorithm requiring predefined variables—to MarketScan 2016–2020 claims data to identify populations with asthma that might benefit most from GLP1RA in reducing AE risk. We built a GLP1RA vs. sulfonylurea new‐user cohort with ≥ 1 inpatient or two outpatient asthma encounters, excluding patients with nonasthma indications for systemic steroids. The outcome was acute AE (hospital admission or emergency department visit for asthma), assessed over 6 months using 599 high‐dimensional features from inpatient/outpatient services and pharmacy claims.Results: In the overall population, GLP1RA decreased AE risk relative to sulfonylurea: aRD −1.4% (−2.0%, −0.8%). hdiCF identified three subgroups based on the quantity of systemic steroid prescription fills (0, 1, and ≥ 2): patients with ≥ 2 prescriptions (GLP1RA: 34 events/1367 individuals; sulfonylurea: 53/1013) benefited most from GLP1RA: aRD −3.8% (−5.3%, −2.2%).Conclusions: This study demonstrates how automated feature identification can pinpoint clinically relevant subgroups with HTEs. The quantity of systemic steroid prescriptions, as a proxy for severe asthma, may guide personalized predictions of GLP1RA's short‐term benefits on acute AE.


  • A novel high dimensional approach to assess heterogeneous treatment effect in claims data

    by Tiansheng Wang, Virginia Pate, Richard Wyss, John B Buse, Michael R Kosorok, Til Stürmer

    We tested a novel high-dimensional approach (using 1 ordinal variable per code with up to four levels: zero, occurred once, sporadically, or frequent) against the standard high-dimensional propensity score (hdPS) method (up to 3 binary variables per code) for detecting heterogeneous treatment effects (HTE). Using the iterative causal forest (iCF) subgrouping algorithm, we analyzed a new-user cohort of 8075 sodium–glucose cotransporter-2 inhibitors and 7313 glucagon-like peptide-1 receptor agonists from a 20% random Medicare sample (2015–2019) with ≥1-year parts A/B/D enrollment and without severe renal disease. We extracted the top 200 prevalent codes across diagnoses, procedures, and prescriptions during the 1-year baseline. Subgroup-specific conditional average treatment effects (CATEs) were assessed for 2-year risk differences (aRD) in hospitalized heart failure using inverse-probability treatment weighting. The overall population exhibited an aRD of −0.4% (95% CI, −1.1% to 0.2%). Our high-dimensional setting identified patients with ≥2 loop diuretic prescriptions (aRD, −2.6%, 95% CI, −5.0% to -0.2%) as the subgroup with the largest CATE. In contrast, the high-dimensional setting from hdPS identified patients with chronic kidney disease (aRD, −1.7%, 95% CI, −3.6% to 0.2%). Across various sensitivity analyses, our high-dimensional approach more accurately identified expected subgroups with HTE that aligns with prior clinical evidence.