Table 2. Summary statistics for the association between consumption of pickled vegetables and oesophageal cancer.

	Number of studies	Q-statistica	P-valueb	I2 (%)c	Random effects OR (95% CI)
All studies	34	282.18	<0.001	88	2.08 (1.66–2.60)
After excluding the three most influential studies	31	120.45	<0.001	75	2.32 (1.92–2.81)
Prospective studies	3	10.26	0.006	81	1.52 (0.82–1.63)
Case-control studies	31	179.25	<0.001	83	2.18 (1.75–2.73)
Large studies (standard error ⩽0.5)	30	259.42	<0.001	89	2.09 (1.67–2.63)
Studies with population-based controls	18	185.63	<0.001	91	2.11 (1.56–2.85)
Studies with age-matched controls	27	171.69	<0.001	85	2.14 (1.66–2.76)
Adjusted results	24	196.16	<0.001	88	2.15 (1.64–2.81)
Studies reporting consumption rates among cases and controls	18	57.61	<0.001	70	2.55 (2.01–3.24)
Studies from Linxian, China	4	18.69	<0.001	84	1.51 (0.98–2.33)
Studies not from Linxian, China	30	220.89	<0.001	87	2.17 (1.70–2.78)
Studies initiated in 1990 or earlierd	13	160.43	<0.001	93	1.73 (1.14–2.62)
Studies initiated after 1990d	20	51.69	0.002	63	2.28 (1.89–2.76)
Studies from China (Mainland)	29	246.64	<0.001	89	2.10 (1.63–2.69)
Studies not from China (Mainland)	5	25.31	<0.001	84	1.96 (1.12–3.43)
Studies published in English literature	15	198.34	<0.001	93	2.10 (1.47–3.00)
Studies published in Chinese literature	19	60.06	<0.001	70	2.10 (1.65–2.67)

Abbreviations: CI=confidence interval; OR=odds ratio.

^aChi-square Q-statistic for homogeneity.

^bP-value for the Q-statistic.

^cHiggins I² statistic for heterogeneity.

^dOne study was excluded from these analyses because the starting year was not reported.