Table 3.

Associations between higher BMI and three negative control variables, using regression and instrumental variable analysis and based on the unrelated individuals used in the one sample MR. Our depression variable is included for reference

		Observationalb		Geneticc		Geneticd
Negative control variable	N	Beta or LN(OR)e representing change in negative control variable or depression per SD higher BMI	P	Beta or LN(OR)e representing change in negative control variable per SD higher genetically instrumented BMI	P	Beta or LN(OR)e representing change in negative control variable per SD higher genetically instrumented BMI	P
Regular sun protection usea	375 720	−0.041 (−0.047, −0.036)	<1 x 10−15	−0.052 (−0.098, −0.006)	0.028	−0.040 (−0.087, 0.006)	0.09
Nitrogen dioxide pollution level	372 791	0.023 (0.020, 0.026)	<1 x 10−15	0.025 (0.000, 0.050)	0.050	−0.006 (−0.030, 0.018)	0.63
Rural dwelling	386 131	−0.067 (−0.076, −0.058)	<1 x 10−15	−0.023 (−0.094, 0.048)	0.52	0.006 (−0.065, 0.077)	0.88
Depression	41 397 (246 106)	0.150 (0.139, 0.160)	<1 x 10−15	0.166 (0.084, 0.247)	7 x 10−5	0.153 (0.07, 0.235)	0.0002

^aCoded as never, sometimes, most of the time, always. Analysed using ordinal logistic regression.

^bRegression analysis adjusting for age and sex.

^cOne-sample instrumental variable analysis with a BMI GRS.

^dOne-sample instrumental variable analysis with a BMI GRS and accounting for socioeconomic position using the Townsend deprivation index.

^eOn natural log scale, LN(OR).