Table 4.

Sample Size^a and Study Duration Estimates for Binary Hypoglycemia Outcomes in a Nocturnal Hypoglycemia Prevention Trial^b

Treatment effect (intervention versus control)	Nights per subject	Unadjustedc
		Total subjects	Total nights
Outcome threshold 70 mg/dl
25% versus 12.5%	14	26	364
	28	13	364
	42	9	378
	56	7	392
25% versus 8.3%	14	13	182
	28	7	196
	42	5	210
	56	4	224
Outcome threshold 60 mg/dl
15% versus 7.5%	14	47	658
	28	24	672
	42	16	672
	56	12	672
15% versus 5%	14	24	336
	28	12	336
	42	8	336
	56	6	336

^aThe sample size estimates apply to a study design using any type of CGM device with false positive and false negative rates comparable to those described in the text.

^bAssuming two-tailed test with alpha = 0.05 and 90% power.

^cBinomial variance decreased by 12% to account for subject serving as their own control in the crossover design. Correlation from repeated nights from the same subject using sensor data were estimated from data obtained in the JDRF CGM RCT.⁶ As described in the text, for an outcome threshold of 70 mg/dl, the number of nights should be increased by a factor of 3 to account for misclassification and increased by a factor of 5 for a threshold of 60 mg/dl.