Exposure to low doses of the radioactive gas radon, as found indoors in dwellings, has been consistently shown to be a risk factor for lung cancer (LC). The linear-no-threshold hypothesis (LNT) is often applied to estimate excess odds ratios or population attributable risks, albeit this LNT assumption remains debated. We investigate the profile of radon-induced LC-risk in a sample of 8,927 cases and 5,562 controls of the International Lung Cancer Consortium (ILCCO), contributed by studies with sufficient exposure heterogeneity. Spatial indoor-radon exposure in the residential area (sIR) obtained from the national surveys were linked to the participants’ residential geo-location. Parametric linear- and spline-functions were fitted within framework of logistic regression. We observed a U-shaped dose-risk relation, with the lowest risk exposure level (LRE) being 57.6 Bq/m³ (95%.CI: 56.1–59.2 Bq/m³). The risk of overall-LC at 25 Bq/m³ (OR = 1.31, 95%-CI: 1.01–1.59) was comparable to that at 100 Bq/m³ (OR = 1.34, 95%-CI: 1.20–1.45). Regarding histological subtypes, we observed the strongest risk for small-cell LC, and weak association for squamous-cell LC with no association below 58 Bq/m³. Our results showed a U-shaped risk-profile for radon-induced LC risk at very low exposure levels (sIR < 200 Bq/m³), lowest in areas of mean indoor radon levels of about 58 Bq/m³. Risk profiles differ between histological subtypes, and sex, age and smoking behaviour modify the lowest risk thresholds. sIR is a useful proxy for radon exposure, and the linearity-no-threshold assumption in this data seems not optimal for the dose-response relation of sIR less than 200 Bq/m³.