Skip to main content
. 2021 Jul 13;7:92. doi: 10.1038/s41523-021-00296-8

Table 2.

Odds ratios of the constrained partial proportional odds model.

Health status thresholds OR (95% HPDI) Posterior probability of harm (OR < 1)
EORTC QLQ-C30 sum score, 60 vs 85 ≥50 0.54 (0.27–1.12) 95%
≥66 0.34 (0.20–0.62) 100%
≥75 0.27 (0.15–0.47) 100%
≥83 0.21 (0.11–0.38) 100%
Perceived risk of recurrence, very high vs medium-low Any 0.47 (0.23–1.04) 97%
Perceived risk of recurrence, very high vs high Any 0.84 (0.49–1.46) 73%
Perceived risk of recurrence, high vs medium-low Any 0.56 (0.27–1.11) 56%
No ALND Any 0.77 (0.40–1.52) 79%
Age, 40 vs 55 years ≥50 0.26 (0.08–0.82) 99%
≥66 0.37 (0.15–0.92) 98%
≥75 0.45 (0.19–10.6) 97%
≥83 0.53 (0.21–1.28) 92%
TNM stage, 3 vs 2 Any 0.67 (0.27–1.68) 81%
TNM stage, 2 vs 1 Any 1.11 (0.63–2.01) 35%
Nontaxane regimen Any 1.04 (0.55–1.84) 44%
Mastectomy ≥50 0.30 (0.11–0.84) 99%
≥66 0.42 (0.20–0.91) 99%
≥75 0.49 (0.24–0.99) 98%
≥83 0.57 (0.28–1.23) 93%

We can see that the CPPO model assumes proportional odds for all variables, except for baseline QoL and age (that interacts with surgery). The model imposes a linear restriction on the coefficients of these variables that do not meet the proportional odds assumption with respect to cutoffs. In contrast, the variables that do meet the proportional odds assumption have a single odds ratio that is valid for the entire range of possible cutoffs. The EORTC QLQ-C30 sum score was measured before adjuvant chemotherapy.

ALND axillary lymph node dissection, CI confidence interval, TNM tumor–node–metastases, OR odds ratio, HPDI highest posterior density interval.