Proton pump inhibitors are not associated with fundic gland polyps - a systematic review that takes into consideration all known confounders

Abstract

Sporadic fundic gland polyps (FGPs) progress, albeit rarely, to dysplasia and cancer. Two meta-analyses, including 8 and 11 studies, concluded that proton pump inhibitors (PPIs) were associated with FGPs. Intervention is considered unnecessary when FGPs have a background of PPIs use. Both meta-analyses, however, disregarded known confounders: age, sex, endoscopy indications, study design (prospective or retrospective), duration of PPI use, and H. pylori infection. Confounders are known to invalidate meta-analyses. We followed PRIXMA guidelines and searched the literature for studies on FGPs in PPI-users and PPI-nonusers. In the 22 studies searched, we compared FGPs in PPI-users (n = 6534) and PPI-nonusers (n = 41 115). Heterogeneity was significant (Cochran Q = 277.8, P < 0.0001; I2 = 92.8%), annulling meta-analysis performed by blanket tallying. To offset the above confounders, we matched PPI-users and PPI-nonusers by (a) age and sex (n = 4300 and 29 307, respectively) and (b) their propensity scores derived from the confounders (n = 2950 and 4729, respectively). After both matching, FGPs were not significantly different between PPI-users and PPI-nonusers [odds ratio (OR) = 1.1, P = 0.3078; OR = 0.9, P = 0.3258, respectively]. Furthermore, FGP frequency did not correlate with increasing duration of PPI use (Pearson and Spearman correlation coefficients = 0.1162, 0.0386, P < 0.6064, 0.8646, respectively); it was not significantly different between any of the duration periods of observation, namely, <10, 10-20, 20-40, >40 months, nor was it significantly different between PPI-users and PPI-nonusers within each duration period (P > 0.05). We conclude that PPIs are not associated with FGPs, implying that a background history of PPI use is not a justification for nonintervention in the management of FGPs.

Fig. 1.

Flowchart of studies searched and identified according to PRISMA 2020 flow diagram of selection process [74].

Fig. 2.

(a) Forest plot and (b) funnel plot. Both plots show the presence of significant heterogeneity among the 22 studies identified in the literature reporting on odds ratio for fundic gland polyps in patients treated versus patients not treated with proton pump inhibitors. In the forest plot, the area of each square is proportional to the study’s size in the meta-analysis; the fixed effects and random effects (more representative of the true effect [92] and preferred [93]) with weight % are shown. The funnel plot also indicates presence of publication bias.

Table 2a.

Balance of known confounders by matching of propensity scores

Fig. 3.

Duration of observation and percent (mean ± SE) of patients with fundic gland polyps in patients treated and not treated with PPI. Data were derived from reported studies as follows – duration <10 months: Jeong et al. [35], Ally et al. [18], Cao et al. [32], Vieth and Stolte [28], Peura et al. and Metz et al. [91], Garcia-Alonso [50], Lee et al. [24]; duration 10–20 months: Ally et al. [18], Jeong et al. [35], Huang et al. [22], Cao et al. [32], Camilo et al. [33], Velazquez-Dohorn et al. [36], Graham [14], Reekmans et al. [16], Jalving et al. [17], Yoshida et al. [20], Zelter et al. [21], Loffeld [23], Shinozaki et al. [94]; duration 20–40 months: Ally et al. [18], Jeong et al. [35], Hsu et al. [19], Choudhry et al. [15]; duration >40 months: Lundell et al. [30]/Fiocca et al. [31], Ally et al. [18], Gao et al. [25], Kroupa et al. [90]. Data for duration <10 months were available in studies by Ally et al., Jeong et al., and Cao et al.; data for duration 10–20 and 20–40 months were available for Ally et al. and Jeong et al. Duration of PPI use did not correlate with the percent of patients with FGPs (Pearson correlation coefficient, r = −0.1162, P < 0.6064; Spearman’s rank correlation coefficient, r_s = −0.0386, P < 0.8646). FGPs, fundic gland polyps; PPI, proton pump inhibitors.