Clinical Question

What is the most effective method for deprescribing chronic use of proton pump inhibitor (PPI) medications?

Evidence-Based Answer

Several deprescribing methods may provide limited success. Some patients may have a return of symptoms with abrupt discontinuation. (Strength of Recommendation [SOR]: B, placebo arms in randomized controlled trials [RCTs].) Changing the PPI prescription to as-needed dosing results in fewer pills being used (by about four pills per week), but with lower patient satisfaction and less symptom control. (SOR: A, meta-analysis.) Taking a PPI as needed decreases use in two-thirds of patients. (SOR: B, RCTs.) Adding a rescue histamine H₂ blocker may reduce the risk of restarting a PPI by about 80%. (SOR: B, secondary outcome in small RCTs.)

Evidence Summary

Deprescribing is recommended for patients with gastroesophageal reflux disease (GERD) who are low risk and are symptom-free after four weeks of PPI therapy. A systematic review and meta-analysis identified six RCTs (n = 1,758) of adults 48 years and older with nonerosive GERD or milder grades of reflux esophagitis (Los Angeles classification system for GERD grades A and B) comparing symptom control, pill burden, and patient satisfaction in patients deprescribed PPIs.¹ Researchers compared abrupt withdrawal and changing to as-needed treatment with continuing daily therapy. Researchers measured outcomes using symptom surveys. Compared with continuous PPI use, the as-needed PPI regimen had a higher risk of poorly controlled symptoms (relative risk [RR] = 1.71; 95% CI, 1.31 to 2.21) and reduced patient satisfaction (RR = 1.82; 95% CI, 1.26 to 2.65), although with a lower pill burden (mean difference = −3.79 pills per week; 95% CI, −4.73 to −2.84). One study (n = 105) examined abrupt discontinuation in an older study group (mean age = 73 years) who all had endoscopic evidence of erosive esophagitis. Symptoms returned in two-thirds of patients who abruptly stopped their PPI (67.9% vs. 22.4% for continuous use; RR = 3.02; 95% CI, 1.74 to 5.24). Trials were inconsistent in how they reported symptom control, the studies were not always blinded, and many trials had small sample sizes.