Citrus and upper digestive cancers: the subsite changes the story
A prospective-study review found different estimates for esophageal, gastric cardia, and gastric non-cardia cancers.

Illustrative food image, not a photograph of the study intervention.
Why this review focused on prospective studies
This systematic review examined prospective studies of citrus fruit intake and esophageal and gastric cancer risk. The PubMed search included studies published through March 1, 2016, and the investigators used random-effects models to calculate summary relative risks and confidence intervals. The abstract explains that much earlier evidence had come from case-control studies, which may be affected by recall and selection biases. Moving to prospective studies changes the evidence base but does not turn a dietary association into a randomized experiment. The useful question here is specific: whether citrus intake was associated with later cancer risk, and whether the answer differed by cancer type.
Vingeliene S, Chan DS, Aune D, Vieira AR et al.. Cancer causes & control : CCC, 2016. PMID 27153845 · DOI
The esophageal dose-response finding was borderline
For each 100 g/day increase in citrus fruit intake, the summary relative risk of esophageal cancer was 0.86, with a 95% confidence interval of 0.74 to 1.00. This analysis included 1,057 cases across six studies, and the authors described the inverse association as marginally significant. The histologic subtype estimates were less precise: 0.87 for squamous cell carcinoma, with a confidence interval of 0.69 to 1.08, and 0.93 for esophageal adenocarcinoma, with a confidence interval of 0.78 to 1.11. Each subtype analysis contained three studies. Those intervals do not support describing every esophageal cancer subtype as clearly associated with lower risk.
Vingeliene S, Chan DS, Aune D, Vieira AR et al.. Cancer causes & control : CCC, 2016. PMID 27153845 · DOI
Gastric cardia and non-cardia were different
For gastric cancer, the abstract separates the cardia and non-cardia results rather than treating the stomach as one undifferentiated endpoint. The dose-response estimate for gastric cardia cancer was 0.75, with a 95% confidence interval of 0.55 to 1.01, across three studies. The authors describe that inverse association as nonsignificant. Gastric non-cardia cancer had a relative risk of 1.02, with a confidence interval of 0.90 to 1.16, across four studies. An overall claim that citrus clearly lowers all gastric cancer risk would conceal this contrast. The direction and uncertainty of the findings depended on which gastric subsite was being examined.
Vingeliene S, Chan DS, Aune D, Vieira AR et al.. Cancer causes & control : CCC, 2016. PMID 27153845 · DOI
Category comparisons are not the same as a serving target
When the researchers compared highest with lowest citrus intake, they found statistically significant inverse associations for esophageal cancer and gastric cardia cancer. The summary relative risks were 0.77, with a 95% confidence interval of 0.64 to 0.91, across seven studies for esophageal cancer, and 0.62, with a confidence interval of 0.39 to 0.99, across three studies for gastric cardia cancer. These category comparisons are distinct from the per-100-g/day analyses. Their different precision is a reason to retain the original comparison labels rather than convert the review into a serving prescription. The authors concluded that further studies were needed.
Vingeliene S, Chan DS, Aune D, Vieira AR et al.. Cancer causes & control : CCC, 2016. PMID 27153845 · DOI