Skip to main content
. 2011 May 10;104(11):1747–1754. doi: 10.1038/bjc.2011.162

Table 4. Cox's proportional hazard model estimating the prognostic impact of PROX1 expression and of established prognostic factors for LGG on survival (n=116).

    Full model
Stepwise model
Prognostic factor Risk ratio Confidence limits P-value Risk ratio Confidence limits P-value
Clinical variables
 Age at onset <40 vs ⩾40 years 1.24 0.87–187 0.3167
 Performance status KPS >80 vs ⩽80 0.95 0.58–1.60 0.8380
 Tumour size ⩾6 vs <6 cm 1.37 0.83–2.21 0.2184
 Tumour crossing midline Yes vs no 1.71 1.09–2.67 0.0211 1.71 1.11–2.61 0.0161
 Contrast enhancement Yes vs no 1.62 1.06–2.47 0.0248 1.52 1.01–2.28 0.0446
 Histology A vs OA/O 1.22 0.99–1.51 0.0624 1.22 1.00–1.50 0.0557
 Extent of resection GTR vs not 0.75 0.43–1.27 0.2951
               
Molecular variables
 Ki-67 expression <4% vs ≅4% 0.93 0.51–1.65 0.8143
 IDH1 mutation Yes vs no 0.57 0.35–0.95 0.0313 0.61 0.39–1.02 0.0575
 PROX1-positive cells ⩾10% vs <10% 1.61 1.04–2.47 0.0310 1.63 1.07–2.45 0.0237

Abbreviations: KPS=Karnofsky performance status; A=astrocytoma; OA=oligoastrocytoma; O=oligodendroglioma; GTR=gross total resection; IDH=isocitrate dehydrogenase; LGG=low-grade gliomas.

Factors removed from the model using the backwards exclusion method (p-to-remove >0.10): Performance status (at step 1, P=0.8380), Ki-67 expression (at step 2, P=0.8627), age at onset (at step 3, P=0.3101), extent of resection (at step 4, P=0.2742), tumour size (at step 5, P=0.1057).