The difficulty with any study of low-level radiation doses on exposed organisms is that ecosystems are messy and complicated, and the actual low-level biological mechanics for low-dose exposures are entirely conjectural, so observed effects have a very, very low signal to noise ratio.
Setting up controlled studies with a large enough scale to make statistically significant judgements greatly exceeds available resources for researchers in the field, leaving statistical analysis of effectively uncontrolled real-world populations as the only option.
So, the conclusions are only as reliable as the observations and the statistical analysis. This becomes educated guesswork, and for the most part educated guesswork based on the theoretical model that is being tested.
It's all very well and good to throw out statistics like:
Many other cell types and tissues have been shown to be affected by Chernobyl contaminants. Møller, Bonisoli-Alquati, et al. (2013) demonstrated that the frequency of visible tumors on birds was significantly higher in radioactive areas, presumably reflecting elevated mutation rates in somatic tissues. Visible tumor rates in birds from Chernobyl were in excess of 15/1000 birds while tumors have never been observed in Danish populations despite extensive surveys (0/35000 birds observed) (Møller, Bonisoli-Alquati, et al. 2013). ...unfortunately, the former USSR as a whole exhibits levels of persistent organic pollutants several times greater than observations in even industrialized areas of Western Europe. POPs are rather easier to study, and are definitively linked to tumor formation.
Attempting to control for various effects in real-world populations is a black art, often practiced and seldom practiced effectively. Often, before you can even start to evaluate the reliability of an article, you'll need to jump several citation links back just to see what assumptions a study is based upon. ...and that is why we still lack conclusive evidence about any long-term negative effects of low-dose radiation exposure.