Skip to main content
. 2018 Aug 14;8:12089. doi: 10.1038/s41598-018-30568-6

Table 2.

Results of regression models for cognitive performance in UK Biobank participants at baseline (2010), using centrally-linked pollutant data.

Exposure (year) Cognitive score Unadjusted Adjusteda
n Estimate (95% CI) p (uncorr) p (FDR)b +/− n Estimate (95% CI) p (uncorr) p (FDR)b +/−
PM10 (2007) Reasoningc 83,238 −0.0139 (−0.0187, −0.0091) 1.558e-08 2.050e-08 − 72,004 0.0111 (0.0054, 0.0169) 0.0001 0.0025 +
Reaction timed 85,651 1.0019 (1.0015, 1.0023) 3.245e-18 5.795e-18 − 73,581 0.9995e (0.9986, 1.0004) 0.2799 0.3888 +
Numeric memoryf 1,456 −0.0301 (−0.0762, 0.0160) 0.2003 0.2177 − 1,258 −0.0285 (−0.0882, 0.0312) 0.3492 0.4595 −
Pairs matchingg 83,806 1.0083 (1.0065, 1.0101) 2.295e-20 4.781e-20 − 72,338 1.0019 (0.9997, 1.0042) 0.0907 0.1512 −
Prospective memoryh 86,198 0.9649 (0.9601, 0.9696) 2.986e-46 1.493e-45 − 73,974 1.0067 (0.9994, 1.0141) 0.0738 0.1419 +
PM2.5 to 10 (2010) Reasoningc 83,331 −0.0915 (−0.1092, −0.0737) 4.334e-24 1.083e-23 − 72,079 −0.0103 (−0.0334, 0.0128) 0.3830 0.4788 −
Reaction timed 85,746 1.0049 (1.0033, 1.0065) 1.996e-09 2.772e-09 − 73,656 1.0019 (0.9997, 1.0042) 0.0887 0.1512 −
Numeric memoryf 1,458 −0.0533 (−0.1237, 0.0171) 0.1376 0.1564 − 1,260 −0.2051i (−0.3706, −0.0395) 0.0152 0.0556 −
Pairs matchingg 83,901 1.0159 (1.0094, 1.0225) 1.644e-06 2.055e-06 − 72,413 0.9966 (0.9874, 1.0059) 0.4727 0.5372 +
Prospective memoryh 86,293 0.9217 (0.9053, 0.9384) 4.855e-19 9.337e-19 − 74,049 0.9963 (0.9674, 1.0261) 0.8061 0.8397 −
PM2.5 (2010) Reasoningc 83,331 −0.1196 (−0.1361, −0.1031) 1.212e-45 5.050e-45 − 72,079 −0.0141 (−0.0347, 0.0065) 0.1802 0.2816 −
Reaction timed 85,746 1.0079 (1.0064, 1.0093) 4.543e-26 1.420e-25 − 73,656 1.0032 (1.0012, 1.0053) 0.0019 0.0158 −
Numeric memoryf 1,458 −0.0382 (−0.1018, 0.0255) 0.2396 0.2496 − 1,260 −0.0876 (−0.1797, 0.0045) 0.0624 0.1300 −
Pairs matchingg 83,901 1.0211 (1.0149, 1.0274) 2.653e-11 3.901e-11 − 72,413 1.0022 (0.9938, 1.0106) 0.6133 0.6666 −
Prospective memoryh 86,293 0.8741 (0.8592, 0.8893) 1.032e-52 8.600e-52 − 74,049 0.9890 (0.9629, 1.0157) 0.4156 0.4948 −
NO2 (2005) Reasoningc 83,364 −0.0056 (−0.0070, −0.0041) 1.388e-14 2.169e-14 − 72,106 0.0032 (0.0013, 0.0050) 0.0007 0.0088 +
Reaction timed 85,779 1.0007 (1.0006, 1.0008) 1.833e-29 6.546e-29 − 73,683 0.9999j (0.9997, 1.0003) 0.9088 0.9088 +
Numeric memoryf 1,458 −0.0060 (−0.0192, 0.0072) 0.3718 0.3718 − 1,260 −0.0109 (−0.0302, 0.0085) 0.2706 0.3888 −
Pairs matchingg 83,934 1.0027 (1.0022, 1.0032) 2.417e-25 6.714e-25 − 72,440 1.0010 (1.0003, 1.0018) 0.0058 0.0290 −
Prospective memoryh 86,326 0.9887 (0.9872, 0.9901) 2.074e-53 2.592e-52 − 74,076 1.0023 (1.0000, 1.0046) 0.0458 0.1145 +
NOx (2010) Reasoningc 83,364 −0.0076 (−0.0086, −0.0066) 1.895e-49 1.184e-48 − 72,106 −0.0015 (−0.0027, −0.0002) 0.0183 0.0556 −
Reaction timed 85,779 1.0005 (1.0004, 1.0006) 7.818e-23 1.777e-22 − 73,683 1.0002 (1.0001, 1.0003) 0.0037 0.0231 −
Numeric memoryf 1,458 −0.0042 (−0.0092, 0.0007) 0.0944 0.1124 − 1,260 −0.0073 (−0.0135, −0.0012) 0.0200 0.0556 −
Pairs matchingg 83,934 1.0016 (1.0012, 1.0019) 2.149e-17 3.582e-17 − 72,440 1.0011k (1.0002, 1.0020) 0.0187 0.0556 −
Prospective memoryh 86,326 0.9919 (0.9909, 0.9929) 3.304e-58 8.260e-57 − 74,076 0.9985 (0.9970, 1.0000) 0.0529 0.1202 −

+/−, point estimate indicates that higher values of the pollutant are associated with better (+) or worse (−) cognitive performance; CI, confidence interval; FDR, false discovery rate; NO2, nitrogen dioxide; NOx, nitrogen oxides; PM, particulate matter.

All estimates are per unit (μg/m3) difference in the air pollutant.

aAdjusted for baseline age, gender, ethnic group, Townsend deprivation score, education, smoking status, physical activity, time outdoors, proximity to nearest major road, traffic intensity on nearest major road, and population density category. Adjusted results are reported from models without an interaction term between air pollutant and time outdoors, unless otherwise noted in the table.

bProbability adjusted using the Simes-Benjamini-Hochberg method implemented in the Stata qqvalue package.

cLinear regression; estimates reported as unstandardized coefficients; possible score range 0 to 13; lower is worse.

dLinear regression using log-transformed values; exponentiated estimates reported as rate ratios; values above 1 indicate relatively longer reaction time.

eInteraction between PM10 and time outdoors: estimates stratified by quintile of time outdoors ranged between 0.9988 (0.9973, 1.0002) in quintile 4 and 1.0011 (0.9997, 1.0026) in quintile 5.

fLinear regression; estimates reported as unstandardized coefficients; possible score range 2 to 12; lower is worse.

gNegative binomial regression; estimates reported as rate ratios; values above 1 indicate relatively more errors.

hLogistic regression; estimates reported as odds ratios; values below 1 indicate relatively lower odds of a correct response.

iInteraction between PM2.5 to 10 and time outdoors: estimates stratified by quintile of time outdoors ranged between −0.1674 (−0.4992, 0.1643) in quintile 2 and 0.1495 (−0.1705, 0.4696) in quintile 5.

jInteraction between NO2 and time outdoors: estimates stratified by quintile of time outdoors ranged between 0.9998 (0.9993, 1.0003) in quintile 4 and 1.0005 (1.0001, 1.0010) in quintile 5.

kInteraction between NOx and time outdoors: estimates stratified by quintile of time outdoors ranged between 0.999996 (0.9991, 1.0009) in quintile 1 and 1.0011 (1.0001, 1.0022) in quintile 3.