Table 3 Summary of the major studies related to the prediction of OA radiographic incidence

Shamir et al. 2009 [35]

BLSA

(Unknown, 123)

240

KL = 0

ΔKL ≥ 2

Baseline TBTA, using a Compound Hierarchy of Algorithms Representing Morphology (WND-CHARM) algorithm, of the region adjacent to the tibial spines was predictive of knee Kellgren-Lawrence (KL) incidence (best accuracy = 0.72, KL from 0 to 3)

Woloszynski et al. 2012 [8]

Lund University

(CR, 105)

48

KL = 0

ΔJSN_M ≥ 1

Baseline TBTA, measured by the signature dissimilarity measure method (SDM), of the tibial plateau was predictive of medial knee JSN incidence (AUC = 0.75)

Podsiadlo et al. 2016 [9]

MOST

(CR&RG, 1433)

48–60

KL ≤ 1

ΔJSN_M ≥ 1

Baseline TBTA, measured by the variance orientation transform (VOT) method, was associated with incident radiographic OA, independently of risk factors for knee OA. Most of the OA incidence occurred in medial compartments

Janvier et al. 2017 [10]

OAI

(CR, 319)

48

KL = 0

ΔJSN_M ≥ 1

Baseline TBTA of the tibial plateau at baseline was predictive of medial knee JSN progression. The best model included TBTA, measured by the quadratic Variations estimator (Var) (AUC = 0.73)