Table 2. Studies performed to validate the ReHiT prediction rule for nonseminomatous testicular cancer.

	Hospitals/groups	Years	n	Prognosis	Percentage benign masses	Calibration	Discrimination	Clinical relevance
Development	Six study groups from	1979–92	544	Good/int/poor	45	OK	AUC=0.83	142/544 (26%) cons nec
	Europe and US							116/142 (82%) correct
Validation 1	Five study groups	1980–96	172	Good/int/poor	45	OK	AUC=0.82	52/172 (30%) cons nec
	from Europe							38/52 (73%) correct
Validation 2a	Indiana University	1985–99	276	Good/int/poor	28	Recalibration necessary	AUC=0.79	24/276 (9%) cons nec
								17/24 (71%) correct
Validation 3	EORTC/MRC trial	1995–98	105	Good	26	OK for predictions >5%	AUC=0.76	4/105 (4%) cons
(present study)								3/4 (75%) correct

Calibration=agreement between predicted probabilities and observed frequencies, Discrimination=ability to distinguish a benign mass from tumour, AUC=area under the curve, cons nec=masses considered as benign (predicted probability >70%), correct=masses correctly considered as benign (predicted probability >70% and histology benign).

^aModified prediction rule.