Table 2. Quantitative criteria for the strength of the evidence for air pollution and risk of bipolar disorder and schizophrenia.

Criteria	Bipolar Disorder	Schizophrenia
Large amount of evidence
Data on >1,000 disease diagnoses	+	+
Strong statistical support
P-value < 10−3 (random effects)	+	–
P-value < 10−6 (random effects)	–	–
No large heterogeneity
Heterogeneity I2 < 50%	+	–
95% PI excludes the null	N/A	N/A
Largest study shows an effect
P < 0.05 in the largest study	+	–
No obvious hints of selective reporting
No small study effects detected	N/A	N/A
No excess significance detected	N/A	N/A

N/A: cannot calculate given the small number of studies; only the two estimates from US and Denmark from Khan and colleagues are considered here. PI calculation would require at least 3 studies with independent effect estimates to be available. Small study effects and excess significance tests would provide hints (not proof) for possible selection and other publication biases, but these tests require many studies with published estimates to be available in order to be assessed with any reliability. It is unknown how many other investigators may have tried to evaluate the association of air pollution and these mental health phenotypes. For reviews of previous studies of environmental exposures, see [14,16]. Abbreviations: N/A, not available; PI, prediction interval.