Prognostic Probability Functions from Clinical-Trial Data

  • Olli S. Miettinen
  • Johann Steurer
  • Albert Hofman


In the eminent intervention-prognostic cohort study (experimental) addressed in Chap.  19 above, the data were synthesized in a manner that represents the prevailing state-of-the-art in clinical trials at large; for, the core results on the various outcomes were derived by means of Cox regression, and they were, thus, treatments-comparing “hazard ratios” – incidence-density ratios, that is – for the various entities of the investigators’ concern. Very notably, prognostic probabilities were not addressed in that particular study, nor are they being addressed in other clinical trials, as the Cox model (for “proportional hazards”) does not provide for this. But, as we show in this Chapter, prognostic probabilities, including their functional dependence on prognostic time (jointly with treatment and prognostic indicators), actually can be studied in another theoretical framework for clinical trials (and their quasi-experimental counterparts). In this approach to clinical-trial data, theory of the other species causal gnostic studies – etiognostic studies (singular in their true essence; Sect.  15.6.2) – is brought to bear on intervention-prognostic studies. This brings logistic-regression models into preeminence in the data-synthesis of intervention-prognostic cohort studies (Sect.  15.6) and in prognostic non-cohort studies (non-experimental; Chap.  21).


Cox regression Alternative sampling Incidence density proper Logistic regression Examples Basic models Augmented models Model fitting 


  1. 1.
    The Women’s Health Initiative steering committee (2004) Effects of conjugated equine estrogen in postmenopausal women with hysterectomy. The Women’s Health Initiative randomized controlled trial. JAMA 292:1701–1712Google Scholar
  2. 2.
    Miettinen OS (1976) Estimability and estimation in case-referent studies. Am J Epidemiol 103:226–235CrossRefGoogle Scholar
  3. 3.
    Bhatnagar S, Turgeon M, Saarela O, Hanley J (2017) Casebase: fitting flexible smooth-in-time hazards and risk functions via logistic and multinomial regression. R package version 0.1.0.
  4. 4.

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Olli S. Miettinen
    • 1
    • 2
  • Johann Steurer
    • 3
  • Albert Hofman
    • 4
  1. 1.Faculty of MedicineMcGill UniversityMontrealCanada
  2. 2.Department of MedicineWeill Cornell Medical CollegeNew YorkUSA
  3. 3.Horten Center for Patient-oriented Research and Knowledge TransferUniversity of ZürichZürichSwitzerland
  4. 4.Department of EpidemiologyHarvard T.H. Chan School of Public HealthBostonUSA

Personalised recommendations