Table 3.

Association (z-value; p trend) of FEV1 %predicted with CRP quartile classes^§

		z-value#	ptrend (across crp quartiles)
Non-smoking men
FEV1 (% predicted)*		-3.91	<0.000
FEV1 (% predicted)**		-4.53	<0.000
Non-smoking women ¶
FEV1 (% predicted)*		-2.32	0.02
FEV1 (% predicted)**		-2.15	0.03
Smoking men
FEV1 (% predicted)*		-2.44	0.015
FEV1 (% predicted)**		-4.82	<0.000
Smoking women ¶
FEV1 (% predicted)*		-2.44	0.015
FEV1 (% predicted)**		-3.14	0.002

*controlled for BMI, systolic blood pressure, ETS (in non smokers), packyears (in smokers), alcohol intake, study area; ** crude associations; ¶additionally adjusted for menopausal status, and intake of female hormones; subjects with crp > 10 mg/l were excluded; §CRP quartiles (Q) (mg/l): Q1 < 0.6, Q2 < 1.1, Q3 < 2.2, Q4 > 2.2; # Regression models without CRP quartiles as predictor variable were computed first. A Cuzick's trend test was then used to test whether regression residuals showed a monotonous association with CRP quartiles. Under the null hypothesis of no trend, the z-value of Cuzick's trend test statistic approximately follows a standard normal distribution, enabling the computation of approximate p-values; a negative sign indicates an inverse association between CRP and FEV1%predicted [15].