pith. sign in

arxiv: 2606.18809 · v1 · pith:AQSSN56Unew · submitted 2026-06-17 · 📊 stat.ME · stat.AP

Applying the Weibull Shape Parameter test for signal detection in pharmacovigilance using the R package WSPsignal

Julia Dyck , Odile Sauzet This is my paper

Pith reviewed 2026-06-26 20:20 UTC · model grok-4.3

classification 📊 stat.ME stat.AP
keywords Weibull shape parameter testpharmacovigilancesignal detectionR packagetime-to-event datahazard functionadverse drug reactionBayesian estimation
0
0 comments X

The pith

An R package called WSPsignal unifies tools for Weibull shape parameter tests that detect drug safety signals from time-to-event data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces the R package WSPsignal to make Weibull shape parameter testing practical for pharmacovigilance. The tests model time until an adverse event after drug initiation and flag a signal when the hazard rate deviates from constancy. The package supports frequentist or Bayesian estimation, several Weibull-family distributions, and lets users either apply preset specifications or run simulations to tune the test for a given dataset. Examples show a frequentist version on roughly 20,000 observations and a Bayesian version on roughly 1,000 observations. The goal is to give practitioners and researchers a single open-source interface that encourages wider use of temporal information in signal detection.

Core claim

The WSPsignal package consolidates all required functionalities for the family of Weibull shape parameter tests into one open-source R interface, allowing either default test specifications or simulation-based tuning to select the most suitable test for a given data scenario in pharmacovigilance.

What carries the argument

The WSP test family, which fits Weibull, double Weibull or power generalized Weibull distributions to time-to-event data and tests for statistically significant departure from constant hazard.

If this is right

  • Users can run a frequentist WSP test directly on large datasets of approximately 20,000 observations.
  • A Bayesian WSP test becomes available for smaller datasets of approximately 1,000 observations.
  • Simulation-based tuning inside the package identifies optimal test specifications for a user's specific data characteristics.
  • The package supplies a single interface that removes the need to assemble separate code for estimation, testing, and tuning.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Routine use of the package could shift pharmacovigilance practice toward greater reliance on temporal patterns rather than count-based methods alone.
  • The simulation tuning feature opens a route for data-driven selection of which member of the WSP family performs best in different reporting environments.
  • Future extensions could incorporate additional distributions or hybrid frequentist-Bayesian procedures without changing the package structure.

Load-bearing premise

The existing WSP test family is suitable and effective for signal detection when applied through the package.

What would settle it

A head-to-head comparison on a real pharmacovigilance database that shows the package outputs fail to detect known adverse drug reactions or produce excessive false positives relative to established methods.

Figures

Figures reproduced from arXiv: 2606.18809 by Julia Dyck, Odile Sauzet.

Figure 1
Figure 1. Figure 1: Workflow to get a signal/no signal from time-to-event data using a frequentist [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Decision rules for credibility interval (CI) + region of practical equivalence [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Graphics for exploratory analysis of the occurrence of musculoskeletal pain over [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: ROC curves representing the top five frequentist Weibull shape parameter test [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Graphics for exploratory analysis of the occurrence of musculoskeletal pain over [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Hazard functions obtained under the prior belief that the adverse event is not an [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: ROC curves representing the top five Bayesian Weibull shape parameter test [PITH_FULL_IMAGE:figures/full_fig_p027_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Outcome from eval.execution_times(pc_list2). Boxplots comparing execu￾tion times in minutes grouped by WSP model and prior distributions. All execution times are below 10 minutes. 0 10000 20000 30000 40000 dw − gg dw − ll pgw − gg pgw − ll tte.dist − prior.dist n_eff Parameter shape1 shape2 Effective sample sizes [PITH_FULL_IMAGE:figures/full_fig_p032_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Outcome from eval.eff_sample_sizes(pc_list2). Boxplots of effective sam￾ple sizes (n_eff) grouped by WSP model and prior distribution for parameters shape/shape1 (red) and shape2/powershape (blue). The recommended threshold for the effective sample size is marked with a horizontal dashed line at 10 000. 32 [PITH_FULL_IMAGE:figures/full_fig_p032_9.png] view at source ↗
read the original abstract

Post-marketing pharmacovigilance relies on statistical signal detection methods to identify potential adverse drug reactions. The Weibull shape parameter (WSP) test concept exploits temporal information (electronic health records) to assess the hazard of an adverse event over time after drug initiation. A statistically significant deviation from constancy results in a signal. The WSP framework comprises a family of tests that differ with respect to the estimation approach (frequentist or Bayesian), the chosen time-to-event distribution (Weibull, double Weibull, power generalized Weibull) for hazard modeling, and test specification parameters. To facilitate practical application and encourage consideration of the WSP signal detection test in future research, we developed the R package WSPsignal. The package consolidates all functionalities required for WSP testing into a unified, open-source interface. It enables practitioners and researchers to apply default test specifications or perform simulation-based tuning to identify the optimal test for a given data scenario. We illustrate the package functionalities in two examples to follow along. In a large-sample setting (ca. 20 000 observations), a frequentist WSP test is considered. In a small-sample setting (ca. 1 000 observations), a Bayesian WSP test is chosen. The additional test specifications are optimized through simulation-based tuning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript describes the R package WSPsignal, which consolidates functionalities for the family of Weibull Shape Parameter (WSP) tests used in pharmacovigilance signal detection. The package supports frequentist and Bayesian estimation, multiple time-to-event distributions (Weibull, double Weibull, power generalized Weibull), default or simulation-tuned test specifications, and is illustrated via two examples: a large-sample (N≈20,000) frequentist application and a small-sample (N≈1,000) Bayesian application with simulation-based optimization of specifications.

Significance. If the package is implemented and validated as claimed, it would provide a practical, open-source interface that lowers the barrier for applying and customizing WSP tests in post-marketing surveillance. This could encourage wider use of temporal hazard modeling in signal detection. The contribution is primarily software-oriented rather than methodological, with no new statistical derivations or performance claims beyond the two illustrative examples.

major comments (2)
  1. [Abstract] Abstract and overall manuscript: no implementation details, source code structure, validation results, error handling, or performance benchmarks are supplied for the claimed functionalities (unified interface, simulation tuning, frequentist/Bayesian options). This directly undermines the central claim that the package 'consolidates all functionalities required for WSP testing' and enables reliable simulation-based tuning, as these cannot be assessed from the description alone.
  2. [Examples] Examples section (large-N frequentist and small-N Bayesian cases): the manuscript provides no quantitative outputs (test statistics, p-values, posterior summaries, or tuning results), error checks, or comparison against known WSP implementations, making it impossible to verify that the package correctly reproduces the underlying WSP test family.
minor comments (1)
  1. [Abstract] Notation for sample sizes ('ca. 20 000' and 'ca. 1 000') is inconsistent in formatting; use standard scientific notation or exact values if available.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive review of our manuscript describing the WSPsignal R package. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract and overall manuscript: no implementation details, source code structure, validation results, error handling, or performance benchmarks are supplied for the claimed functionalities (unified interface, simulation tuning, frequentist/Bayesian options). This directly undermines the central claim that the package 'consolidates all functionalities required for WSP testing' and enables reliable simulation-based tuning, as these cannot be assessed from the description alone.

    Authors: We agree that the manuscript provides only a high-level overview of the package without implementation specifics or benchmarks. The focus was on introducing the consolidated interface and usage rather than serving as technical software documentation. The complete source code is available on CRAN and GitHub for direct inspection of structure, error handling, and tuning routines. In revision we will add a dedicated section outlining the package architecture, key exported functions, and summary validation steps performed during development. revision: yes

  2. Referee: [Examples] Examples section (large-N frequentist and small-N Bayesian cases): the manuscript provides no quantitative outputs (test statistics, p-values, posterior summaries, or tuning results), error checks, or comparison against known WSP implementations, making it impossible to verify that the package correctly reproduces the underlying WSP test family.

    Authors: The examples were written to illustrate workflow rather than to report full numerical results. We acknowledge that including the actual outputs would allow direct verification. In the revised manuscript we will expand both examples to report the test statistics, p-values, posterior summaries, and simulation-tuned specifications, together with any available checks against earlier WSP implementations. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The manuscript is a software-description paper that implements and documents an R package for applying an existing family of WSP tests previously defined in the literature. No new statistical derivations, predictions, or uniqueness claims are advanced. The central assertions concern package functionality and usage examples; these are direct descriptions of code behavior rather than reductions of results to fitted inputs or self-citations. External benchmarks (prior WSP papers) are referenced only as background, not as load-bearing self-referential steps. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard domain assumptions about modeling adverse event timing with Weibull-family distributions and the utility of shape-parameter deviation for signal detection; no new free parameters, axioms, or invented entities are introduced by the paper.

axioms (1)
  • domain assumption Time-to-event data from electronic health records can be modeled with Weibull or related distributions to test deviation from constant hazard for adverse event signal detection.
    Invoked as the basis for the WSP test family in the abstract description of the framework.

pith-pipeline@v0.9.1-grok · 5762 in / 1169 out tokens · 27386 ms · 2026-06-26T20:20:44.089129+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 20 canonical work pages · 1 internal anchor

  1. [1]

    2025 , note =

    WSPsignal: Bayesian Weibull Shape Parameter Tests For Signal Detection , author =. 2025 , note =

    2025

  2. [2]

    Julia Dyck and Odile Sauzet , year=. The. preprint , eprint=

  3. [3]

    Drug safety , volume=

    Principles of signal detection in pharmacovigilance , author=. Drug safety , volume=. 1997 , publisher=. doi:10.2165/00002018-199716060-00002 , note=

    work page doi:10.2165/00002018-199716060-00002 1997

  4. [4]

    Adverse drug reaction , howpublished =

  5. [5]

    Adverse event , howpublished =

  6. [6]

    Signal Detection and Monitoring Based on Longitudinal Healthcare Data , journal=

    Suling, M and Pigeot, I , year=. Signal Detection and Monitoring Based on Longitudinal Healthcare Data , journal=. doi:10.3390/pharmaceutics4040607 , note=

    work page doi:10.3390/pharmaceutics4040607

  7. [7]

    Databases for Pharmacoepidemiological Research , chapter=

    Sturkenboom, M and Schink, T , year=. Databases for Pharmacoepidemiological Research , chapter=. doi:10.1007/978-3-030-51455-6 , note=

    work page doi:10.1007/978-3-030-51455-6

  8. [8]

    Statistical Methods in Medical Research , volume=

    Disproportionality methods for pharmacovigilance in longitudinal observational databases , author=. Statistical Methods in Medical Research , volume=. 2013 , publisher=. doi:10.1177/0962280211403602 , note=

    work page doi:10.1177/0962280211403602 2013

  9. [9]

    Methods for drug safety signal detection in longitudinal observational databases:

    Schuemie, MJ , journal=. Methods for drug safety signal detection in longitudinal observational databases:. 2011 , publisher=. doi:10.1002/pds.2051 , note=

    work page doi:10.1002/pds.2051 2011

  10. [10]

    Data Mining and Knowledge Discovery , volume=

    Temporal pattern discovery in longitudinal electronic patient records , author=. Data Mining and Knowledge Discovery , volume=. 2010 , publisher=. doi:10.1007/s10618-009-0152-3 , note=

    work page doi:10.1007/s10618-009-0152-3 2010

  11. [11]

    https://www.fda.gov/safety/fdas-sentinel-initiative , howpublished =

  12. [12]

    The role of data mining in pharmacovigilance , journal =

    M Hauben and D Madigan and CM Gerrits and L Walsh and EP. The role of data mining in pharmacovigilance , journal =. 2005 , publisher =. doi:10.1517/14740338.4.5.929 , note =

    work page doi:10.1517/14740338.4.5.929 2005

  13. [13]

    Quantitative Drug Safety and Benefit Risk Evaluation: Practical and Cross-Disciplinary Approaches , year =

    Wang, W and Munsaka, M and Buchanan, J and Li, J , publisher =. Quantitative Drug Safety and Benefit Risk Evaluation: Practical and Cross-Disciplinary Approaches , year =

  14. [14]

    Drug safety , volume=

    A signal detection method to detect adverse drug reactions using a parametric time-to-event model in simulated cohort data , author=. Drug safety , volume=. 2012 , publisher=

    2012

  15. [15]

    Generalised weibull model-based approaches to detect non-constant hazard to signal adverse drug reactions in longitudinal data , year =

    O Sauzet and VR Cornelius , journal =. Generalised weibull model-based approaches to detect non-constant hazard to signal adverse drug reactions in longitudinal data , year =. doi:10.3389/fphar.2022.889088 , note =

    work page doi:10.3389/fphar.2022.889088 2022

  16. [16]

    Drug Safety , keywords =

    Sauzet, O and Dyck, J and Cornelius, VR , doi =. Drug Safety , keywords =

  17. [17]

    preprint , year=

    An Exposure Model Framework for Signal Detection based on Electronic Healthcare Data , author=. preprint , year=. 2404.14213 , archivePrefix=

    arXiv

  18. [18]

    Pharmacoepidemiology and Drug Safety , volume=

    Methods for drug safety signal detection using routinely collected observational electronic health care data: A systematic review , author=. Pharmacoepidemiology and Drug Safety , volume=. 2023 , publisher=. doi:10.1002/pds.5548 , note =

    work page doi:10.1002/pds.5548 2023

  19. [19]

    2024 , publisher=

    Signal Analysis in Pharmacovigilance , chapter=. 2024 , publisher=. doi:10.1201/9781032629940-9 , note =

    work page doi:10.1201/9781032629940-9 2024

  20. [20]

    Pharmaceutical Statistics , year =

    Prieto-Merino, D and Quartey, G and Wang, J and Kim, J , title =. Pharmaceutical Statistics , year =. doi:10.1002/pst.524 , publisher =

    work page doi:10.1002/pst.524

  21. [21]

    2025 , note =

    R: A Language and Environment for Statistical Computing , author =. 2025 , note =

    2025

  22. [22]

    2022 , note =

    HDInterval: Highest (Posterior) Density Intervals , author =. 2022 , note =

    2022

  23. [23]

    Survival analysis: A self-learning text , pages=

    Introduction to survival analysis , author=. Survival analysis: A self-learning text , pages=. 2012 , publisher=

    2012

  24. [24]

    2001 , publisher=

    Accelerated Life Models: Modeling and Statistical Analysis , author=. 2001 , publisher=. doi:10.1201/9781420035872 , note=

    work page doi:10.1201/9781420035872 2001

  25. [25]

    2016 , publisher=

    The Cox Model and Its Applications , author=. 2016 , publisher=. doi:10.1007/978-3-662-49332-8 , note=

    work page doi:10.1007/978-3-662-49332-8 2016

  26. [26]

    Drug safety , volume=

    Illustration of the weibull shape parameter signal detection tool using electronic healthcare record data , author=. Drug safety , volume=. 2013 , publisher=. doi:10.1007/s40264-013-0061-7 , note=

    work page doi:10.1007/s40264-013-0061-7 2013

  27. [27]

    Journal of Machine Learning Research , year =

    MD Hoffman and A Gelman , title =. Journal of Machine Learning Research , year =

  28. [28]

    preprint , year=

    T Fawcett , title=. preprint , year=

  29. [29]

    Lloyd, CJ , journal=. Using. 1998 , publisher=. doi:10.1080/01621459.1998.10473797 , note=

    work page doi:10.1080/01621459.1998.10473797 1998

  30. [30]

    Bioinformatics , volume =

    T Sing and O Sander and N Beerenwinkel and T Lengauer , year =. Bioinformatics , volume =. doi:10.1093/bioinformatics/bti623 , note=

    work page doi:10.1093/bioinformatics/bti623

  31. [31]

    Doing Bayesian Data Analysis , edition =

    Kruschke, J , publisher =. Doing Bayesian Data Analysis , edition =. 2015 , pages =

    2015

  32. [32]

    Advances in Methods and Practices in Psychological Science , volume =

    J Kruschke , title =. Advances in Methods and Practices in Psychological Science , volume =. 2018 , doi =

    2018

  33. [33]

    2024 , organization =

    Stan Reference Manual, Version 2.36 , author =. 2024 , organization =

    2024

  34. [34]

    WSPsignal: Weibull Shape Parameter Tests For Signal Detection , author =

  35. [35]

    Carpal Tunnel

    Carvajal, A and Martin A, Luis H and S. Carpal Tunnel. PLoS One , volume=. 2016 , publisher=. doi:10.1371/journal.pone.0146772 , note=

    work page doi:10.1371/journal.pone.0146772 2016

  36. [36]

    2008 , issn =

    Bisphosphonates: Mechanism of Action and Role in Clinical Practice , journal =. 2008 , issn =

    2008

  37. [37]

    PLoS One , publisher =

    The adverse effects of bisphosphonates in breast cancer: A systematic review and network meta-analysis , year =. PLoS One , publisher =. doi:10.1371/journal.pone.0246441 , author =

    work page doi:10.1371/journal.pone.0246441

  38. [38]

    Journal of the American Medical Informatics Association , volume =

    Walonoski, J and Kramer, M and Nichols, J and Quina, A and Moesel, C and Hall, D and Duffett, C and Dube, K and Gallagher, T and McLachlan, S , title = ". Journal of the American Medical Informatics Association , volume =. 2017 , doi =

    2017

  39. [39]

    Adverse effects of bisphosphonates: Implications for osteoporosis management

    Kennel, KA and Drake, MT. Adverse effects of bisphosphonates: Implications for osteoporosis management. Mayo Clinic proceedings. 2009. doi:10.4065/84.7.632

    work page doi:10.4065/84.7.632 2009

  40. [40]

    Anderson localization in an interacting fermionic system

    A O’Hagan , title =. The American Statistician , volume =. 2019 , publisher =. doi:10.1080/00031305.2018.1518265 , note =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1080/00031305.2018.1518265 2019