Selection bias

An article in the American Journal of Epidemiology reported the results of a case-control study designed to examine the association between exposure to various forms of nickel and development of lung cancer. The study group included men who had been employed at a nickel refinery in Norway for at least 12 months. The observed data pertaining to metallic nickel exposure are summarized in the table below:

	Lung cancer case	Control	Total
Exposed to nickel	203	468	671
Unexposed to nickel	10	57	67
Total	213	525	738

1. What is the appropriate measure of association that describes the relationship between any nickel exposure and lung cancer for this study? Use the data in the table to calculate this ratio measure. Provide an interpretation.

Assume that all eligible cases participated in the study, but among controls, 90% of the exposed and 41% of the unexposed agreed to participate.

Observed:

	Lung cancer case	Control	Total
Exposed to nickel	203	468	671
Unexposed to nickel	10	57	67
Total	213	525	738

If we were able to include the 10% of exposed controls and 59% of unexposed controls that did not participate in our study we get the 2x2 table below (the truth).

	Lung cancer case	Control	Total
Exposed to nickel	203	468/0.9 = 520	723
Unexposed to nickel	10	57/0.41 = 139	149
Total	213	659	872

1. What is the unbiased OR (the OR if we had no selection bias) for the association between nickel exposure and lung cancer?