To investigate the clinical response of gastro-esophageal reflux disease (GERD) symptoms to exclusion diets based on food intolerance tests. A double blind, randomized, controlled pilot trial was performed in 38 GERD patients partially or completely non-responders to proton pump inhibitors (PPI) treatment. Fasting blood samples from each patients were obtained; leukocytotoxic test was performed by incubating the blood with a panel of 60 food items to be tested. The reaction of leukocytes (rounding, vacuolization, lack of movement, flattening, fragmentation or disintegration of cell wall) was then evaluated by optical microscopy and rated as follows: level 0 = negative, level 1 = slightly positive, level 2 = moderately positive, and level 3 = highly positive. A "true" diet excluding food items inducing moderate-severe reactions, and a "control" diet including them was developed for each patient. Then, twenty patients received the "true" diet and 18 the "control" diet; after one month (T1) symptoms severity was scored by the GERD impact scale (GIS). Hence, patients in the "control" group were switched to the "true" diet, and symptom severity was re-assessed after three months (T2). At baseline (T0) the mean GIS global score was 6.68 (range: 5-12) with no difference between "true" and control group (6.6 ± 1.19 vs 6.7 ± 1.7). All patients reacted moderately/severely to at least 1 food (range: 5-19), with a significantly greater number of food substances inducing reaction in controls compared with the "true" diet group (11.6 vs 7.0, P < 0.001). Food items more frequently involved were milk, lettuce, brewer's yeast, pork, coffee, rice, sole asparagus, and tuna, followed by eggs, tomato, grain, shrimps, and chemical yeast. At T1 both groups displayed a reduction of GIS score ("true" group 3.3 ± 1.7, -50%, P = 0.001; control group 4.9 ± 2.8, -26.9%, P = 0.02), although the GIS score was significantly lower in "true" vs "control" group (P = 0.04). At T2, after the diet switch, the "control" group showed a further reduction in GIS score (2.7 ± 1.9, -44.9%, P = 0.01), while the "true" group did not (2.6 ± 1.8, -21.3%, P = 0.19), so that the GIS scores didn't differ between the two groups. Our results suggest that food intolerance may play a role in GERD symptoms development, and leucocytotoxic test-based exclusion diets may be a possible therapeutic approach when PPI are not effective or indicated.