Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial

E. M. Holmes, I. Bradbury, L. S. Williams, L. Korde, E. De Azambuja, D. Fumagalli, A. Moreno-Aspitia, J. Baselga, M. Piccart-Gebhart, A. C. Dueck, R. D. Gelber

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Background: Design, conduct, and analysis of randomized clinical trials (RCTs) with time to event end points rely on a variety of assumptions regarding event rates (hazard rates), proportionality of treatment effects (proportional hazards), and differences in intensity and type of events over time and between subgroups. Design and methods: In this article, we use the experience of the recently reported Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization (ALTTO) RCT, which enrolled 8381 patients with human epidermal growth factor 2-positive early breast cancer between June 2007 and July 2011, to highlight how routinely applied statistical assumptions can impact RCT result reporting. Results and conclusions: We conclude that (i) futility stopping rules are important to protect patient safety, but stopping early for efficacy can be misleading as short-term results may not imply long-term efficacy, (ii) biologically important differences between subgroups may drive clinically different treatment effects and should be taken into account, e.g. by pre-specifying primary subgroup analyses and restricting end points to events which are known to be affected by the targeted therapies, (iii) the usual focus on the Cox model may be misleading if we do not carefully consider non-proportionality of the hazards. The results of the accelerated failure time model illustrate that giving more weight to later events (as in the log rank test) can affect conclusions, (iv) the assumption that accruing additional events will always ensure gain in power needs to be challenged. Changes in hazard rates and hazard ratios over time should be considered, and (v) required family-wise control of type 1 error ≤ 5% in clinical trials with multiple experimental arms discourages investigations designed to answer more than one question. Trial Registration: clinicaltrials.gov Identifier NCT00490139.

Original languageEnglish (US)
Pages (from-to)1507-1513
Number of pages7
JournalAnnals of Oncology
Volume30
Issue number9
DOIs
StatePublished - Sep 1 2019

Fingerprint

Randomized Controlled Trials
Medical Futility
Therapeutics
Patient Safety
Proportional Hazards Models
Epidermal Growth Factor
Clinical Trials
Breast Neoplasms
Weights and Measures
lapatinib
Trastuzumab
Power (Psychology)
Drive

Keywords

  • accelerated failure time models
  • early breast cancer
  • family-wise type 1 error
  • power
  • proportional hazards
  • stopping boundaries

ASJC Scopus subject areas

  • Hematology
  • Oncology

Cite this

Holmes, E. M., Bradbury, I., Williams, L. S., Korde, L., De Azambuja, E., Fumagalli, D., ... Gelber, R. D. (2019). Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial. Annals of Oncology, 30(9), 1507-1513. https://doi.org/10.1093/annonc/mdz195

Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial. / Holmes, E. M.; Bradbury, I.; Williams, L. S.; Korde, L.; De Azambuja, E.; Fumagalli, D.; Moreno-Aspitia, A.; Baselga, J.; Piccart-Gebhart, M.; Dueck, A. C.; Gelber, R. D.

In: Annals of Oncology, Vol. 30, No. 9, 01.09.2019, p. 1507-1513.

Research output: Contribution to journalArticle

Holmes, EM, Bradbury, I, Williams, LS, Korde, L, De Azambuja, E, Fumagalli, D, Moreno-Aspitia, A, Baselga, J, Piccart-Gebhart, M, Dueck, AC & Gelber, RD 2019, 'Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial', Annals of Oncology, vol. 30, no. 9, pp. 1507-1513. https://doi.org/10.1093/annonc/mdz195
Holmes EM, Bradbury I, Williams LS, Korde L, De Azambuja E, Fumagalli D et al. Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial. Annals of Oncology. 2019 Sep 1;30(9):1507-1513. https://doi.org/10.1093/annonc/mdz195
Holmes, E. M. ; Bradbury, I. ; Williams, L. S. ; Korde, L. ; De Azambuja, E. ; Fumagalli, D. ; Moreno-Aspitia, A. ; Baselga, J. ; Piccart-Gebhart, M. ; Dueck, A. C. ; Gelber, R. D. / Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial. In: Annals of Oncology. 2019 ; Vol. 30, No. 9. pp. 1507-1513.
@article{af0545c43eee4e908ae451162ee907ab,
title = "Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial",
abstract = "Background: Design, conduct, and analysis of randomized clinical trials (RCTs) with time to event end points rely on a variety of assumptions regarding event rates (hazard rates), proportionality of treatment effects (proportional hazards), and differences in intensity and type of events over time and between subgroups. Design and methods: In this article, we use the experience of the recently reported Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization (ALTTO) RCT, which enrolled 8381 patients with human epidermal growth factor 2-positive early breast cancer between June 2007 and July 2011, to highlight how routinely applied statistical assumptions can impact RCT result reporting. Results and conclusions: We conclude that (i) futility stopping rules are important to protect patient safety, but stopping early for efficacy can be misleading as short-term results may not imply long-term efficacy, (ii) biologically important differences between subgroups may drive clinically different treatment effects and should be taken into account, e.g. by pre-specifying primary subgroup analyses and restricting end points to events which are known to be affected by the targeted therapies, (iii) the usual focus on the Cox model may be misleading if we do not carefully consider non-proportionality of the hazards. The results of the accelerated failure time model illustrate that giving more weight to later events (as in the log rank test) can affect conclusions, (iv) the assumption that accruing additional events will always ensure gain in power needs to be challenged. Changes in hazard rates and hazard ratios over time should be considered, and (v) required family-wise control of type 1 error ≤ 5{\%} in clinical trials with multiple experimental arms discourages investigations designed to answer more than one question. Trial Registration: clinicaltrials.gov Identifier NCT00490139.",
keywords = "accelerated failure time models, early breast cancer, family-wise type 1 error, power, proportional hazards, stopping boundaries",
author = "Holmes, {E. M.} and I. Bradbury and Williams, {L. S.} and L. Korde and {De Azambuja}, E. and D. Fumagalli and A. Moreno-Aspitia and J. Baselga and M. Piccart-Gebhart and Dueck, {A. C.} and Gelber, {R. D.}",
year = "2019",
month = "9",
day = "1",
doi = "10.1093/annonc/mdz195",
language = "English (US)",
volume = "30",
pages = "1507--1513",
journal = "Annals of Oncology",
issn = "0923-7534",
publisher = "Oxford University Press",
number = "9",

}

TY - JOUR

T1 - Are we assuming too much with our statistical assumptions? Lessons learned from the ALTTO trial

AU - Holmes, E. M.

AU - Bradbury, I.

AU - Williams, L. S.

AU - Korde, L.

AU - De Azambuja, E.

AU - Fumagalli, D.

AU - Moreno-Aspitia, A.

AU - Baselga, J.

AU - Piccart-Gebhart, M.

AU - Dueck, A. C.

AU - Gelber, R. D.

PY - 2019/9/1

Y1 - 2019/9/1

N2 - Background: Design, conduct, and analysis of randomized clinical trials (RCTs) with time to event end points rely on a variety of assumptions regarding event rates (hazard rates), proportionality of treatment effects (proportional hazards), and differences in intensity and type of events over time and between subgroups. Design and methods: In this article, we use the experience of the recently reported Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization (ALTTO) RCT, which enrolled 8381 patients with human epidermal growth factor 2-positive early breast cancer between June 2007 and July 2011, to highlight how routinely applied statistical assumptions can impact RCT result reporting. Results and conclusions: We conclude that (i) futility stopping rules are important to protect patient safety, but stopping early for efficacy can be misleading as short-term results may not imply long-term efficacy, (ii) biologically important differences between subgroups may drive clinically different treatment effects and should be taken into account, e.g. by pre-specifying primary subgroup analyses and restricting end points to events which are known to be affected by the targeted therapies, (iii) the usual focus on the Cox model may be misleading if we do not carefully consider non-proportionality of the hazards. The results of the accelerated failure time model illustrate that giving more weight to later events (as in the log rank test) can affect conclusions, (iv) the assumption that accruing additional events will always ensure gain in power needs to be challenged. Changes in hazard rates and hazard ratios over time should be considered, and (v) required family-wise control of type 1 error ≤ 5% in clinical trials with multiple experimental arms discourages investigations designed to answer more than one question. Trial Registration: clinicaltrials.gov Identifier NCT00490139.

AB - Background: Design, conduct, and analysis of randomized clinical trials (RCTs) with time to event end points rely on a variety of assumptions regarding event rates (hazard rates), proportionality of treatment effects (proportional hazards), and differences in intensity and type of events over time and between subgroups. Design and methods: In this article, we use the experience of the recently reported Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization (ALTTO) RCT, which enrolled 8381 patients with human epidermal growth factor 2-positive early breast cancer between June 2007 and July 2011, to highlight how routinely applied statistical assumptions can impact RCT result reporting. Results and conclusions: We conclude that (i) futility stopping rules are important to protect patient safety, but stopping early for efficacy can be misleading as short-term results may not imply long-term efficacy, (ii) biologically important differences between subgroups may drive clinically different treatment effects and should be taken into account, e.g. by pre-specifying primary subgroup analyses and restricting end points to events which are known to be affected by the targeted therapies, (iii) the usual focus on the Cox model may be misleading if we do not carefully consider non-proportionality of the hazards. The results of the accelerated failure time model illustrate that giving more weight to later events (as in the log rank test) can affect conclusions, (iv) the assumption that accruing additional events will always ensure gain in power needs to be challenged. Changes in hazard rates and hazard ratios over time should be considered, and (v) required family-wise control of type 1 error ≤ 5% in clinical trials with multiple experimental arms discourages investigations designed to answer more than one question. Trial Registration: clinicaltrials.gov Identifier NCT00490139.

KW - accelerated failure time models

KW - early breast cancer

KW - family-wise type 1 error

KW - power

KW - proportional hazards

KW - stopping boundaries

UR - http://www.scopus.com/inward/record.url?scp=85072747026&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072747026&partnerID=8YFLogxK

U2 - 10.1093/annonc/mdz195

DO - 10.1093/annonc/mdz195

M3 - Article

C2 - 31240310

AN - SCOPUS:85072747026

VL - 30

SP - 1507

EP - 1513

JO - Annals of Oncology

JF - Annals of Oncology

SN - 0923-7534

IS - 9

ER -