We agree with Dr. Shojania that our meta-analysis can be reasonably interpreted as reflecting the limitations of the currently accepted reference tests for GERD, a point we made in the introduction and in the Methods and Discussion sections. Considerable uncertainty clearly surrounds the optimal definition and diagnosis of GERD (1). Our meta-analysis provided a detailed understanding of the relationship between these various tests and the PPI test. We were able to better define sensitivity, specificity, and predictive values of the PPI test for GERD while delineating factors that affect test accuracy.

We disagree with the suggestion that our inclusion and exclusion criteria might have reinforced misclassification. We targeted a population in which physicians wanting to confirm the diagnosis of GERD would consider the use of short-term PPI treatment. This does not include patients with alarm symptoms (because they should be referred for upper endoscopy) or patients with possible coronary artery disease, who should undergo cardiac evaluation.

As we noted in the Discussion section, the high false-positive rate of the PPI test may indeed reflect a placebo response, suggesting that we may be using an expensive placebo in many patients. Whether the accuracy of the PPI test is any better when assessed at later time points is uncertain. Also unclear is what the implications would be on practice; faced with a patient who does not experience sustained benefit, would most clinicians stop the drug, order additional testing, or increase the dosage?

We did not recommend changing practice. On the other hand, our conclusion, that successful short-term treatment with a PPI does not confidently establish the diagnosis of GERD, has important implications. When examining a patient with possible GERD, the clinician must consider several potential courses of action on the basis of the clinical setting. These include an empirical trial of a PPI (or other drug) or specific testing. Part of the decision to use a PPI as initial empirical therapy may be based on the presumption that it will help to establish the diagnosis of GERD. Our study suggests that this presumption may not be as grounded in evidence as is commonly believed.

Thus, while empirical therapy with a PPI is reasonable in patients without alarm symptoms, a response (or lack thereof) does not necessarily prove or disprove a GERD diagnosis. This suggests that it may be reasonable to lower the threshold for considering alternative diagnoses or medical treatments before committing patients empirically to long-term PPI therapy.