Table 6.

Net benefits and interventions avoided for the models assessed through decision‐curve analysis for the detection of any cancer in the follow‐up of 352 patients with non‐muscle‐invasive bladder cancer

Threshold probability (%)	Net benefits					Interventions avoided per 100 patients
	Treat all	Treat none	Treat based on Bladder EpiScore	Treat based on current predictors	Advantage	Bladder EpiScore	Current predictors
5	0.09	0	0.09	0.089	0.001	2.6	0
10	0.04	0	0.08	0.05	0.03	35.2	3.98
15	−0.01	0	0.07	0.003	0.067	49.5	8.8
20	−0.08	0	0.07	0	0.14	57.1	30.4
25	−0.01	0	0.06	0	0.21	61.9	44.3
30	−0.02	0	0.05	0	0.28	65.2	53.6
35	−0.03	0	0.04	0	0.37	68.3	60.2
40	−0.04	0	0.04	0	0.47	71	65.2
45	−0.06	0	0.03	0	0.59	72	69.1
50	−0.07	0	0.03	0	0.75	75	72.1
55	−0.09	0	0.02	0	0.93	76.5	74.7
60	−1.1	0	0.01	0	1.16	77.6	76.8
65	−1.5	0	0	0	1.46	78.6	78.6
70	−1.9	0	0	0	1.87	80	80

In decision‐curve analysis prediction models are compared to two default strategies: (i) assume that all patients are test positive and therefore treat everyone, or (ii) assume that all patients are test negative and offer treatment to no one. The table shows the net benefits for a strategy of performing a cystoscopy in every patient (treat all), in no one (treat none), based on Bladder EpiCheck and on current predictors (i.e. last stage and last grade). For example, given a personal threshold probability of 15% (i.e. one would undergo a cystoscopy if the probability of cancer is >15%) the value of 0.07 can be interpreted as: ‘Compared to performing no cystoscopy, performing a cystoscopy on the basis of the Bladder EpiCheck is the equivalent of a strategy that found seven cancers per 100 patients without conducting any unnecessary cystoscopy’. Moreover, at this threshold probability every decision based on Bladder EpiCheck would avoid 49.5% of unnecessary cystoscopies without missing any cancer.