Symptom relief and healing in patients with erosive esophagitis

• Among 16 head-to-head trials, those with comparable doses did not find differences in symptom relief or healing of esophagitis.
• The only difference between proton pump inhibitors on the outcome of complete symptom relief at 4 weeks was in the comparison of esomeprazole 40 mg with omeprazole 20 mg; the pooled risk difference in 3 trials was 8% (95% CI 3 to 13), with a number needed to treat of 13.
• Time to relief of heartburn was similar for all proton pump inhibitors in head-to-head trials, but the methods used to measure and report this outcome varied in the 14 studies.
• Good evidence showed no difference between omeprazole, lansoprazole, pantoprazole, and rabeprazole for healing of esophagitis. Thirteen head-to-head trials found these 4 proton pump inhibitors to be equally effective in healing at 4 and 8 weeks.
• Pooled analysis of 4- and 8-week healing rates from 4 trials of esomeprazole 40 mg compared to omeprazole 20 mg indicated esomeprazole to be superior; risk difference 7% (95% CI 1 to 12) and a number needed to treat of 14 and 5% (95% CI 1 to 9), number needed to treat = 20, respectively.
• Three trials compared esomeprazole 40 mg with lansoprazole 30 mg. The pooled difference in healing rate was significantly greater with esomeprazole at 4 and 8 weeks, risk differences 5% (95% CI 2 to 7) and 3% (95% CI 1 to 5), respectively.
• Evidence on the comparison of esomeprazole 40 mg and pantoprazole 40 mg was mixed, with 2 studies finding esomeprazole superior and 2 finding no difference in healing rates. Pooled analysis of 3 trials with similar populations finds that esomeprazole was superior to pantoprazole at 4 weeks (risk difference 5%, 95% CI 2 to 8), but not at 8 weeks (risk difference 1%, 95% CI −3 to 5).

Healing in moderate to severe erosive esophagitis

• Esomeprazole 40 mg was more effective at healing esophagitis at 4 and 8 weeks than omeprazole 20 mg and lansoprazole 30 mg.
• The pooled risk difference in 3 studies comparing omeprazole 20 mg with esomeprazole 40 mg was 16% at 4 weeks and 13% at 8 weeks (number needed to treat = 6 at 4 weeks, 8 at 8 weeks).
• The pooled risk difference in 2 studies comparing lansoprazole 30 mg with esomeprazole 40 mg was 8% at 4 weeks and 9% at 8 weeks (number needed to treat = 13 at 4 weeks, 11 at 8 weeks).
• Evidence was mixed on differences between esomeprazole 40 mg and pantoprazole 40 mg.
  • At 4 weeks, esomeprazole 40 mg had a higher healing rate than pantoprazole 40 mg; pooled risk difference (2 studies) 14% (95% CI 7 to 20)
  • At 8 weeks, no difference was found in a single small study of patients with mild to moderate esophagitis.
• Lansoprazole 30 mg (2 studies) and esomeprazole 20 mg (1 study) were no different to omeprazole 20 mg at 4 or 8 weeks.

Prevention of relapse in patients with erosive esophagitis

• For maintenance of healed esophagitis, there was good evidence that no difference exists between omeprazole, lansoprazole, and rabeprazole. The longest study (over 5 years) compared omeprazole with rabeprazole.
• Two 6-month studies found lower relapse rates for esomeprazole 20 mg than for lansoprazole 15 mg or pantoprazole 20 mg.
• No difference was found between esomeprazole 20 mg and pantoprazole 20 mg in combined symptomatic and endoscopic remission rates after 6 months.

Symptom relief in patients with nonerosive gastroesophageal reflux disease or presumptively treated symptoms of gastroesophageal reflux disease

• Three head-to-head trials in patients with gastroesophageal reflux disease but without erosive esophagitis on endoscopy found no difference between esomeprazole 20 mg and omeprazole 20 mg, pantoprazole 20 mg, or rabeprazole 10 mg. These studies used different outcome measures.
• Limited indirect evidence from placebo-controlled and active-control trials suggested similar efficacy for heartburn resolution and complete symptom relief for all 5 proton pump inhibitors.

Evidence in children

• There were no direct comparisons of proton pump inhibitors for reflux esophagitis in children. A fair-quality placebo-controlled trial in infants did not find omeprazole to be superior to placebo in controlling symptoms or acid-exposure time.

Erosive esophagitis

We identified 31 randomized controlled trials comparing 2 or more proton pump inhibitors in patients with gastroesophageal reflux disease with endoscopically-proven erosive esophagitis (Evidence Table 1).^{4–6, 10–38} Two publications are supplemented with additional data provided by the manufacturer.^{4, 5} Most studies used omeprazole. No study of omeprazole in combination with sodium bicarbonate met inclusion criteria. The scales used to grade esophagitis in these studies are described in Appendix E.

In most studies of proton pump inhibitors, patients who have esophagitis before treatment undergo another endoscopy for assessment of healing 4 or 8 weeks after starting treatment. There is no evidence that rate of esophageal healing after 4 or 8 weeks of treatment is associated with risk of stricture or esophageal cancer in the long run. As distinct from symptom relief, the benefit of quicker esophageal healing is also uncertain.

The clinical importance of small differences in healing rates at 4 or 8 weeks is not known. In addition, patients who have clinically significant improvements but who are not completely healed (for example, patients whose esophagitis improves from LA classification scale grade D to grade B) are considered unhealed. Studies do not report the esophagitis grade for patients “not healed” at follow-up.

Resolution of symptoms

Five head-to-head comparisons of proton pump inhibitors measured symptom relief as a primary outcome, ^{10, 11, 13, 16, 37} and 14 reported symptoms as a secondary outcome.^{4, 5, 12, 14, 15, 17, 21–26, 30, 32, 35} Symptoms in these studies were assessed through patient diaries, investigator-elicited reports, or both.

Sixteen head-to-head trials reported the proportion of patients with resolution of symptoms at 4 weeks.^{4, 5, 10, 12–14, 16, 17, 20, 23, 24, 26, 27, 29, 33, 36} We performed a random-effects meta-analysis of data from these studies to determine an estimate of the proportion who were symptom-free at 4 weeks for each drug. Results are shown in Table 2. Proportions ranged from 65% to 77%, and 95% confidence intervals overlapped, indicating the drugs are similarly efficacious for resolution of symptoms at 4 weeks.

Table 2

Symptom resolution in head-to-head trials in patients with erosive gastroesophageal reflux disease

A systematic review of most of these trials, with search dates through 2004, evaluated the proton pump inhibitors as a group and compared to one another.³⁹ This meta-analysis found omeprazole 20 mg daily to be inferior to esomeprazole 40 mg or lansoprazole 30 mg daily in heartburn relief at day 1, with relative risks of 0.78 (95% CI 0.71 to 0.85) and 0.82 (95% CI 0.75 to 0.88), respectively. Lansoprazole and esomeprazole were not found statistically different (relative risk 1.03; 95% CI 0.87 to 1.22). Our analysis includes more recently published trials.

Figure 2 shows risk differences in rates of symptom resolution at 4 weeks in these trials.^{4, 5, 10, 12–14, 16, 17, 20, 23, 24, 26, 27, 29, 33, 36} In Table 3 we report the difference in symptom resolution for esomeprazole compared with other proton pump inhibitors. The pooled data on the comparison of esomeprazole 40 mg with omeprazole 20 mg significantly favored esomeprazole; for every 13 persons treated with esomeprazole 40 mg instead of omeprazole 20 mg, 1 additional patient would be symptom-free at 4 weeks in the esomeprazole group. The pooled data for comparison of esomeprazole 40 mg with either lansoprazole 30 mg or pantoprazole 40 mg did not indicate a significant difference between drugs.

Figure 2

Resolution of symptoms at 4 weeks in head-to-head trials of proton pump inhibitors

Table 3

Symptom resolution at 4 weeks in trials of esomeprazole compared with another proton pump inhibitor in erosive gastroesophageal reflux disease

A single study reported resolution of symptoms after 1 week of therapy,³² finding rabeprazole 20 mg daily superior to omeprazole 20 mg daily (resolution in 27.9% of patients compared with 16.6%, P=0.0013 as calculated from number randomized and using chi square analysis).

A head-to-head trial of pantoprazole 40 mg compared with esomeprazole 40 mg used the ReQuest Score to assess symptoms.³⁵ ReQuest is a validated self-assessment scale used to measure symptoms in erosive and nonerosive gastroesophageal reflux disease. Measured on the last 3 days of a 4-week treatment period, the median ReQuest-GI score in patients taking pantoprazole was found to be non-inferior to the median score in patients taking esomeprazole.

Time to relief of symptoms

Fourteen studies reported the time to resolution of symptoms (no heartburn). This outcome usually was reported as the percentage of patients with symptom resolution by a given time point, such as 1 day or 7 days), the median number of days to resolution, or both. One study reported this outcome as the number of days needed for 50% and 75% of patients to achieve resolution of symptoms.¹⁰

Another measure was the time to sustained resolution of heartburn, defined as the first of 7 consecutive days without heartburn. This outcome was used only in studies funded by the maker of esomeprazole, so it is not possible to compare this outcome with studies funded by others.

Esomeprazole compared with omeprazole

In 4 studies that compared esomeprazole 40 mg with omeprazole 20 mg, the median number of days to the first resolution of symptoms was similar; however, the median number of days to sustained resolution of symptoms favored esomeprazole in the 2 studies reporting this measure (Table 4).^{5, 12, 16} More patients taking esomeprazole 40 mg reached first of resolution of symptoms by day 1 and day 7 in absolute proportions than patients taking omeprazole 20 mg. These findings were statistically significant in 1 study,¹² nonsignificant in 2 others,^{16, 31} and not assessed in the fourth.⁵ The time to sustained resolution of heartburn was statistically superior with esomeprazole 40 mg compared to omeprazole 20 mg at day 14 in 2 studies.^{12, 16} The differences at other time points were mixed or not statistically assessed. One of these studies used a tablet formulation of esomeprazole that is not available in the US or Canada.³¹

Table 4

Time to symptom relief in trials comparing esomeprazole with omeprazole in erosive gastroesophageal reflux disease

In a comparison of esomeprazole 20 mg with omeprazole 20 mg,⁵ a higher proportion of omeprazole patients started 7 consecutive days without heartburn at day 1; esomeprazole had a higher proportion of patients with sustained relief by day 28. Neither comparison was statistically significant. The median number of days to sustained resolution was similar. This pattern was also seen in the time to first resolution of symptoms.

Esomeprazole compared with lansoprazole

In 3 studies comparing esomeprazole 40 mg with lansoprazole 30 mg, results were mixed and outcomes were reported differently (Table 4). Overall, results did not favor one drug over another.

Esomeprazole compared with pantoprazole

The 2 trials comparing esomeprazole with pantoprazole reported time to symptom resolution differently and found conflicting results. In 1 trial comparing esomeprazole 40 mg with pantoprazole 40 mg, 4% more esomeprazole patients began sustained resolution of heartburn (7 consecutive days) after 1 day of treatment: 24% compared with 20% (P value not reported).³⁰ The median time to sustained resolution was 6 days with esomeprazole, compared with 8 days (P<0.001). Based on this same life-table analysis, the Cumulative proportion of patients reporting sustained resolution of heartburn was 78% with esomeprazole and 77% with pantoprazole, again a small difference but found to be statistically significant (P<0.001). A second trial comparing esomeprazole 40 mg with pantoprazole 40 mg looked at the number of days required for relief of heartburn in 50% and 75% of patients.¹⁰ In both groups, 50% of patients had no heartburn after 2 days. But it took 3 days for 75% of the pantoprazole group to be relieved of symptoms and 8 days for the esomeprazole group. Confidence intervals overlapped, (95% CI for pantoprazole, 2 to 7 days; for esomeprazole, 3 to14 days) suggesting a significant difference between the drugs is unlikely but not proven.

Lansoprazole compared with omeprazole

Three studies reported time to relief of heartburn with lansoprazole compared with omeprazole.^{14, 15, 25} Although lansoprazole improved some symptoms more quickly, there was no strong or consistent pattern suggesting that lansoprazole provides faster symptom relief than omeprazole. Time to sustained resolution of heartburn (defined as 3 consecutive days without heartburn) was measured in 1 study and was similar for the drugs (median 3 days for both drugs, P=0.285).¹⁴ In another study, daytime and nighttime heartburn were reported separately.²⁵ After 1 day of treatment, more lansoprazole patients were free of day heartburn (48.7% compared with 37.6%, P<0.05) and night heartburn (62% compared with 52%, P<0.05). The third comparison of these drugs used a visual analogue scale to measure heartburn and reported the time to relief only for daytime heartburn.¹⁵ After 3 days, there was a significant decrease in symptom score in lansoprazole patients (−20.2 compared with −15.3, P=0.05); the difference was not significant after 7 days (scores not reported).

Rabeprazole compared with omeprazole

One study reported similar mean time to complete relief of heartburn for rabeprazole and omeprazole 20 mg daily (7 and 8 days, respectively).³² A second study reported median time to achieve heartburn control, defined as the first day heartburn score was below 3 on a 5-point Likert scale.³⁴ The median time to heartburn control was 1.5 days for both rabeprazole and omeprazole (P<0.43).

Healing of esophagitis

All the proton pump inhibitors allowed esophagitis to heal. Healing rates at 4 weeks ranged from 49% to 91% and at 8 weeks ranged from 71% to 99% (see Evidence Table 1). One small, fair-quality study conducted at a single center in China had a lower 8-week healing rate than other studies (64% for esomeprazole 40 mg, 45.5% for omeprazole 20 mg).³¹

To estimate healing rates for each drug, we pooled data from head-to-head trials, using a random-effects model to control for the effect of the study. Table 5 shows results of this analysis. (Note that data for lansoprazole 15 mg, pantoprazole 20 mg, and rabeprazole 10 mg are available from only 1 study). Healing rates were similar and confidence intervals overlapped, indicating no significant differences between proton pump inhibitors.

Table 5

Pooled estimates of healing rates for esophagitis in head-to-head trials of proton pump inhibitors

We also calculated the risk difference for healing in head-to-head comparisons. Figures 3 and 4 show the differences in healing rates at 4 and/or 8 weeks for the 23 trials that provided the number healed/total patients.^{4–6, 12, 14, 15, 17, 18, 20–23, 25–27, 29–33, 36, 37} Seven head-to-head trials are not represented in Figures 3 and 4: Three studies (2 comparing rabeprazole with omeprazole, 1 comparing omeprazole with both lansoprazole and rabeprazole)^{19, 28, 41} did not provide number healed/total and 4 trials^{10, 11, 13, 16} reported only symptom relief, not esophagitis healing. For 1 trial comparing rabeprazole 20 mg with omeprazole 20 mg the figures show calculated intention-to-treat numbers, rather than those from the article, which are not intention-to-treat.³²

Figure 3

Esophagitis healing at 4 weeks in head-to-head trials of proton pump inhibitors

Figure 4

Esophagitis healing at 8 weeks in head-to-head trials of proton pump inhibitors

Although some published studies present results according to life-table analysis, only crude rates are included in Figure 3. For published studies that do not provide crude rates, we requested and received these data from the manufacturer. Results of life-table analyses cannot be directly compared with crude rates reported in other studies, and using life-table analysis may overestimate results by excluding patients who are lost to follow-up or have withdrawn from the study.

Omeprazole 20 mg, the first proton pump inhibitor to be marketed, was the proton pump inhibitor used most often in head-to-head trials. Table 6 summarizes the risk differences in healing rate in 9 trials^{12, 15, 21, 22, 25–27, 31, 36} that compared daily omeprazole 20 mg with another proton pump inhibitor. Risk difference at 4 and 8 weeks was significant in only 1 comparison, esomeprazole 40 mg compared with omeprazole 20 mg. The pooled risk difference for 3 studies at 4 weeks was 8% and for 4 studies at 8 weeks was 6%. These risk differences translate to numbers needed to treat to heal 1 additional patient of 13 at 4 weeks and 17 at 8 weeks.

Table 6

Risk differences in healing of esophagitis in trials of omeprazole 20 mg compared with another proton pump inhibitor

Two published trials comparing esomeprazole 40 mg with omeprazole 20 mg found a statistically significantly higher healing rate in the esomeprazole group.^{5, 12} Two others^{36, 38} found no difference between groups at 4 and 8 weeks. A small study (N=48) not included in Table 5 found a higher healing rate for esomeprazole at 8 weeks (64% compared with 46%), but the difference was not statistically significant.³¹ The study may not have had sufficient power to detect a difference between treatment groups; no power calculation was reported. This study did not measure 4-week healing rates.

The pooled risk difference for 4 studies at 4 weeks was 7% and for 5 studies at 8 weeks was 5%, favoring esomeprazole (see Table 6). This translates to a number needed to treat with esomeprazole to heal 1 additional patient at 4 weeks of 14, and a number needed to treat at 8 weeks of 20.

Three studies compared esomeprazole 40 mg with lansoprazole 30 mg.^{4, 18, 29} In a large, good-quality trial in 5241 patients at multiple centers in the United States,⁴ healing rates were higher in the esomeprazole group at 4 and 8 weeks. A smaller, fair-quality trial¹⁸ in patients with mostly mild to moderate esophagitis found the drugs to have equivalent healing rates at 8 weeks; results at 4 weeks were also similar between drugs. The third study, rated good quality,²⁹ was conducted in patients with moderate to severe esophagitis. At 4 weeks, the esomeprazole group had a higher healing rate, but at 8 weeks the difference was not significant.

Pooled estimates show that with esomeprazole, healing rate is higher by 5% at 4 weeks and by 3% at 8 weeks (Table 6). With a random-effects analysis the difference at 8 weeks is not significant, but in fixed-effects analysis, the difference is significant (see table 6). The fixed-effect estimates of risk difference correspond to a number of patients needed to treat with esomeprazole instead of lansoprazole to heal 1 additional patient at 4 weeks equal to 20 and at 8 weeks equal to 33.

Four trials compared esomeprazole 40 mg with pantoprazole 40 mg.^{20, 30, 33, 37} Two studies find esomeprazole superior, while 2 do not. One³⁰ large study (N=3171) found that healing at 4 weeks was 6% higher in the esomeprazole group (95% CI 3 to 9). At 8 weeks, the difference was smaller but statistically significant (risk difference, 3%; 95% CI 1 to 5). We rated this study fair quality. A much smaller study (N=180) was also rated fair to poor because numbers of patients enrolled and analyzed in tables did not match the numbers discussed in the text (apparently a typographical error was made in Table 1), the study was apparently open-label, and no details on randomization or allocation concealment procedures were given.³⁷ This study also found esomeprazole 40 mg to have significantly higher rates of healing at 4 weeks (78% compared with 72%; P<0.05). Rates at 8 weeks were not statistically significantly different (92% compared with 91%). No patients with Grade D esophagitis were enrolled in this study, although they were not excluded

Two studies (N=227 and 581) found no differences in healing rates between the drugs^{20, 33} at early time points (4 to 6 weeks in 1 study, 4 weeks in the other) or later time points (8 to 10 weeks in 1 study, 8 and 12 weeks in the other). The smaller of these studies included only patients with Grade B or C esophagitis.²⁰

The 2 largest of these studies also examined the impact of Helicobacter pylori status on healing rates with the 2 drugs. One found that healing rates with esomeprazole were not different based on Helicobacter pylori status, but that Helicobacter pylori negative patients had lower healing rates with pantoprazole compared to those who were Helicobacter pylori positive.³⁰ The other study, however did not find any associated differences in healing rate, symptom relief or ‘complete remission’ when using intention to treat analyses.³³

The largest study³⁰ reports only life-table analysis results, while the other studies report raw rates of number of patients healed. However, data on the crude rates of healing were provided by AstraZeneca through public comment on this report; the data used in the analysis below for this study are not published data. Using these data to conduct a pooled analysis of the 3 studies that included patients with all grades of esophagitis indicates that esomeprazole 40 mg is superior to pantoprazole 40 mg in rates of patients with healed erosions at 4 weeks, but not at 8 weeks (Table 8 below). Sensitivity analysis including the fourth study which included only patients with Grades B and C esophagitis,²⁰ or only the 2 highest quality studies, did not change these results.^{30, 33}

Table 8

Risk differences in healing of esophagitis in trials of esomeprazole 40 mg compared with another proton pump inhibitor

Analysis of healing rates by baseline severity of esophagitis

Nineteen head-to-head trials reported information about esophagitis healing rates by baseline severity of esophagitis.^{4–6, 12–15, 18–23, 25, 27, 29, 30, 37, 38} These results are shown in Evidence Table 1. Ten trials stratify by baseline severity the ratio of number of patients with healed esophagitis to total number patients (Figures 5 and 6).^{4–6, 12, 15, 25, 29, 30, 37, 38} To estimate healing rate for each drug at 4 and 8 weeks for patients with moderate to severe esophagitis (that is, grades C-D or 3–4; see Appendix E for grading scales), we conducted a random-effects meta-analysis of data from these 9 studies^{4–6, 12, 15, 25, 29, 30, 38} (Table 9). An additional study¹⁸ reports a combined outcome of improved by 2 grades or healed; those data were not included in the meta-analysis.

Figure 5

Healing of moderate to severe esophagitis at 4 weeks in head-to-head trials of proton pump inhibitors

Figure 6

Healing of moderate to severe esophagitis at 8 weeks in head-to-head trials of proton pump inhibitors

Table 9

Estimated healing rates in patients with moderate to severe esophagitis at baseline

Esomeprazole compared with omeprazole

Four studies comparing daily esomeprazole 40 mg with omeprazole 20 mg^{5, 12, 36, 38} reported healing rate in patients with moderate to severe esophagitis at baseline (Figures 5 and 6). The pooled risk difference at 4 weeks was 16% (95% CI 11 to 22) and at 8 weeks was 13% (95% CI 9 to 17).

In 2 studies comparing esomeprazole 20 mg with omeprazole 20 mg^{5, 6} there was no difference in healing rate at 4 weeks (pooled risk difference 2%; 95% CI −5 to 10) or 8 weeks (pooled risk difference 4%; 95% CI −3 to 10). Estimates of healing rates with esomeprazole 20 mg were similar to omeprazole 20 mg (see Table 7). There were no comparisons of esomeprazole (any dose) with omeprazole 40 mg.

Table 7

Risk differences in healing of esophagitis in head-to-head trials of esomeprazole 40 mg compared with lansoprazole 30 mg

Esomeprazole compared with lansoprazole

Two studies comparing esomeprazole 40 mg with lansoprazole 30 mg reported healing rates in patients with moderate to severe esophagitis at baseline.^{4, 29} The pooled risk difference at 4 weeks was 8% (95%, CI 4 to12) and at 8 weeks was 9% (95% CI 5 to 12). These correspond to a number needed to treat of 13 at 4 weeks and 11 at 8 weeks.

A third study, published by the maker of lansoprazole, reported only the combined outcome of healing or improvement of at least 2 grades in the subgroup of patients with moderate to severe esophagitis.¹⁸ In this small (N=109) subanalysis lansoprazole had a statistically nonsignificant higher rate of healing/improvement at 8 weeks (10%; 95% CI −2 to 22); results at 4 weeks were not reported.

Esomeprazole compared with pantoprazole

In 1 study patients with moderate (Grade C) esophagitis at baseline who were taking pantoprazole 40 mg had a higher healing rate at “later” time points (8 to 10 weeks) than patients on esomeprazole 40 mg (67% compared with 45%).²⁰ Esophagitis in 100% of patients with Grade C esophagitis on pantoprazole and 91% of patients on esomeprazole improved by 1 or 2 grades (to Grade B or A) by the final visit (10 weeks). Rates at 4 weeks are not reported, and no patients with Grade D esophagitis were enrolled.

In 2 trials of esomeprazole 40 mg compared with pantoprazole 40 mg in patients with moderate to severe esophagitis, there was a 14% risk difference favoring esomeprazole after 4 weeks (95% CI 7 to 21).^{30, 37} At 8 weeks, there was no difference between the drugs in healing rate, although the 1 study that reported this outcome was small (N=29). ³⁷

Lansoprazole compared with omeprazole

Three studies comparing lansoprazole with omeprazole reported healing rate in patients with moderate to severe (Grades 3 and 4) esophagitis.^{14, 15, 25} Two of these compared lansoprazole 30 mg with omeprazole 20 mg.^{15, 25} There was no difference in healing rate at 4 weeks (pooled risk difference 1%; 95% CI −13 to 16) or 8 weeks (pooled risk difference 3%; 95% CI −4 to 10). The third study compared lansoprazole 30 mg with omeprazole 40 mg and reported healing rates as percentages only.¹⁴ There was no significant difference between groups at 4 or 8 weeks. The distribution of the severity of esophagitis among patients in this study is not reported.

Indirect evidence

Comparisons of proton pump inhibitors across studies are difficult because patient populations and healing rates in control groupswere dissimilar.

Esophagitis healing

In the systematic review mentioned above,⁴⁵ 4 proton pump inhibitors were better than ranitidine at healing esophagitis, but there were no differences among them. No study of esomeprazole was included.⁴⁵

We reviewed 22 randomized controlled trials published through 2001 that compared a proton pump inhibitor with an H2 receptor antagonist for esophagitis healing. Figure 7 shows the rates of esophagitis healing at 8 weeks. These trials compared an H2 receptor antagonist with omeprazole (11 studies),^49–59 lansoprazole (5 studies),^60–64 pantoprazole (5 studies),^65–69 and rabeprazole (1 study).⁷⁰

Figure 7

Esophagitis healing at 8 weeks in 22 randomized controlled trials comparing proton pump inhibitor with H2 receptor antagonist

We did not create evidence tables of these studies or rate their quality, because after graphing their results we found no indication that the proton pump inhibitors differed. If an obvious difference in healing rates were seen in an individual study or studies, investigation of study quality would have been undertaken. In our meta-analysis, proton pump inhibitors were more effective at healing than H2 receptor antagonists, but there was no difference in healing rate among the proton pump inhibitors for any comparison. Healing rate ranged from 71.2% to 85.6%.

Nonerosive and endoscopically unexamined gastroesophageal reflux disease

We identified 3 fair-quality head-to-head trials of proton pump inhibitors in short-term treatment of patients with nonerosive or empirically treated reflux disease. They compared esomeprazole with omeprazole,⁸⁰ rabeprazole,⁸¹ or pantoprazole.⁸² The 3 studies used different outcome measures, but all found esomeprazole to be similar in efficacy to the compared drug (Evidence Table 3). A fourth head-to-head trial (lansoprazole compared with omeprazole) included patients with erosive and nonerosive gastroesophageal reflux disease but did not separate results by these patient populations.⁸³ Three identically designed 4-week trials comparing omeprazole 20 mg and esomeprazole 20 mg and 40 mg were conducted simultaneously and were described in 1 publication.⁸⁰ There was no difference in the resolution of heartburn at 14 days (secondary outcome) or 28 days (primary outcome) between patients taking omeprazole 20 mg or esomeprazole 20 mg or 40 mg. At 2 weeks, proportions of patients with resolution ranged from 35% to 44%, and at 4 weeks ranged from 57% to 70%. Results for adequate control of symptoms were similar, with no significant differences between drugs.

A head-to-head trial comparing pantoprazole 20 mg with esomeprazole 20 mg measured time to first and sustained relief of symptoms.⁸⁴ This trial was designed to test for noninferiority of pantoprazole compared with esomeprazole. The noninferiority margin was set at −2 days for the primary outcome of time to first symptom relief (that is, a lower boundary of the 95% confidence interval greater than 2 days would indicate noninferiority). Symptom assessment was based on patient report using a validated questionnaire (ReQuest). The questionnaire includes items on the 7 dimensions of gastroesophageal reflux disease symptoms (general well-being, acid complaints, upper abdominal/stomach complaints, lower abdominal/digestive complaints, nausea, sleep disturbances, and other complaints). Results showed that pantoprazole was not inferior to esomeprazole for first and sustained relief of symptoms.

A 4-week trial comparing rabeprazole 10 mg with esomeprazole 20 mg was conducted in 134 patients in Singapore.⁸¹ The primary outcome was time to first 24-hour period without symptoms of heartburn or regurgitation. There was no difference between groups on this endpoint (for heartburn, 8.5 days for rabeprazole compared with 9.0 days for esomeprazole; for regurgitation, 6.0 days for rabeprazole compared with 7.5 days for esomeprazole; P=NS). There was also no significant difference between groups on secondary outcomes, including complete and satisfactory relief of heartburn symptoms at weeks 1 and 4, and symptom severity score in the first 5 days.

A good-quality Cochrane systematic review of literature through 2003 addressed the efficacy of proton pump inhibitors, H2 receptor antagonists, and prokinetics in adults with endoscopically verified nonerosive or empirically treated symptoms of reflux disease.⁸⁵This review was not designed to compare the efficacy of different proton pump inhibitors. The primary efficacy outcome of the review was heartburn remission, defined as mild heartburn on no more than 1 day per week. Proton pump inhibitors were superior to placebo for heartburn remission and overall symptom improvement. Proton pump inhibitors also were more effective than H2 receptor antagonists for heartburn remission in empirically treated patients (pooled relative risk 0.69; 95% CI 0.61 to 0.77), but not in patients with nonerosive gastroesophageal reflux disease (pooled relative risk 0.74; 95% CI 0.53 to 1.03). However, only 3 trials compared proton pump inhibitors with H2 receptor antagonists in nonerosive gastroesophageal reflux disease.

Another systematic review evaluated the efficacy of proton pump inhibitors for resolution of heartburn in patients with nonerosive gastroesophageal reflux disease.⁸⁶ This review searched literature through 2002, including the US Food and Drug Administration website. 7 placebo-controlled trials (3 published and 4 unpublished) were included: 2 rabeprazole, 2 esomeprazole, and 3 omeprazole. In patients with nonerosive gastroesophageal reflux disease, the risk difference in comparisons with placebo for resolution of heartburn at 4 weeks was 25% (95% CI 18 to 31). The review does not provide evidence about comparative efficacy of different proton pump inhibitors in patients with nonerosive gastroesophageal reflux disease.

Table 10 shows rates of heartburn remission rates and complete symptom relief calculated from data provided in the Cochrane review.⁸⁵ Similar proportions of patients experienced heartburn resolution or complete symptom relief across the drugs.

Table 10

Percent patients with resolution of heartburn at 4 weeks from Cochrane review

We identified 1 additional placebo-controlled⁸⁷ and 1 active-control (ranitidine) trial⁸⁸ published since this review (Evidence Table 3). In a fair-quality trial of empiric treatment of patients with symptoms of gastroesophageal reflux disease, more patients taking pantoprazole 20 mg than ranitidine 300 mg were free of gastroesophageal reflux disease symptoms (heartburn, acid eructation, and pain on swallowing) at 4 weeks (68% compared with 43%).⁸⁸ In a fair- to poor-quality, 8-week, placebo-controlled trial of patients with endoscopically verified nonerosive gastroesophageal reflux disease whose primary symptom was upper abdominal discomfort, patients taking lansoprazole 15 mg had fewer days with upper abdominal discomfort and reduced severity of average daily pain.⁸⁷ Patients whose predominant symptom was heartburn were not included. It is not clear what proportion of patients was analyzed; patients were excluded from analysis for a specific endpoint if there were no data available for that endpoint.

Prevention of relapse

We identified only 1 head-to-head trial of maintenance treatment in patients with nonerosive gastroesophageal reflux disease.⁸⁹ We also included 2 placebo-controlled trials of on-demand rabeprazole⁹⁰ and esomeprazole⁹¹ and a placebo-controlled trial of scheduled of omeprazole.⁹² Details of these trials are shown in Evidence Table 5. Three other trials included patients with endoscopically verified nonerosive and erosive gastroesophageal reflux disease, but did not report results separately by group.^{40, 93, 94}

A head-to-head trial compared on-demand esomeprazole 20 mg with scheduled lansoprazole 15 mg for 6 months in patients with endoscopically verified nonerosive gastroesophageal reflux disease who had experienced complete relief of heartburn with esomeprazole 20 mg during an acute treatment phase (2 to 4 weeks).⁸⁹ Patients were not blinded to treatment and the primary outcome measure was time to discontinuation from the maintenance phase due to unwillingness to continue. Patients also recorded heartburn and other symptoms on diary cards and were asked about their satisfaction with treatment during scheduled clinic visits. By 6 months, significantly more patients receiving lansoprazole 15 mg were unwilling to continue than patients receiving esomeprazole 20 mg on demand (13% compared with 6%, P=0.001). More patients in the lansoprazole group said they discontinued because of adverse events (7.4% compared with 2.3%, P=0.0028), but discontinuations because of heartburn were not significantly different between treatment groups (4.8% for lansoprazole and 2.9% for esomeprazole, P value reported as NS). At 1 month, more esomeprazole patients were satisfied with their treatment, but at 3 and 6 months there was no difference between treatment groups on this measure. During the maintenance phase, the mean frequency of heartburn symptoms was higher in the on-demand esomeprazole group than the scheduled lansoprazole group.

Two 6-month placebo-controlled studies reported efficacy of on-demand therapy with rabeprazole 10 mg⁹⁰ or esomeprazole 20 mg⁹¹ in patients with endoscopically verified nonerosive gastroesophageal reflux disease. In both studies, only patients who experienced complete symptom relief during an acute treatment phase were enrolled in the maintenance phase. In the study of rabeprazole 10 mg, rate of discontinuation due to inadequate heartburn control was 20% for placebo and 6% for rabeprazole (P<0.00001). Although mean length of heartburn-free periods was similar between groups, the time required for resolution of an episode of heartburn was significantly shorter with rabeprazole than placebo. In the study of esomeprazole 20 mg, 14% of patients taking esomeprazole discontinued the study drug compared with 51% taking placebo. Discontinuation was mainly due to inadequate control of heartburn (P<0.0001).⁹¹

In a placebo-controlled trial of daily omeprazole 10 mg, 27% of patients taking omeprazole discontinued the drug due to inadequate control of heartburn over 6 months compared with 52% of patients taking placebo.⁹²

Children

There were no head-to-head trials of proton pump inhibitors in children. Placebo-controlled and active-control trials in children are shown in Evidence Table 6.

A fair-quality placebo-controlled trial of omeprazole (10 to 20 mg daily) in infants (3 to 12 months old) with gastroesophageal reflux defined as a gastric pH <4 for 5% of the monitoring time (unspecified) and/or abnormal esophageal histology found no difference in the cry/fuss time or visual analog scale scores of parent-assessed irritability between placebo and omeprazole.⁹⁵ Histologic and pH measures improved significantly with omeprazole but not placebo.

A poor-quality trial comparing omeprazole (40 mg daily per 1.73 square meters body surface area) with high-dose ranitidine (20 mg/kg daily) in children with reflux refractory to standard-dose ranitidine found both drugs to be effective; but, high dropout rate (19%), lack of intention-to-treat analysis, and inadequate reporting of baseline characteristics make these results unreliable.⁹⁶

Duodenal ulcer

• The data on comparative effectiveness of various proton pump inhibitors for treating duodenal ulcer were strong, with 10 head-to-head trials. Omeprazole 20 mg daily was typically the comparison.
• The evidence was strong for omeprazole and lansoprazole having similar effectiveness in both symptom relief and endoscopically verified healing. The pooled risk difference for 5 trials comparing daily lansoprazole 30 mg with omeprazole 20 mg was −0.2 (95% CI −3.0 to +2.6).
• The evidence for pantoprazole, rabeprazole, and esomeprazole was less strong because there are only single studies for each newer drug compared with omeprazole and no comparisons to other proton pump inhibitors.
• No evidence of a difference in healing rate among proton pump inhibitors.
• Symptom relief was an important measure in ulcer disease and did not always correlate with healing confirmed by endoscopy. Method of assessing symptom relief varied across studies and reporting of findings was often limited to early time points and few outcome measures (of many measured). Few studies found a difference in any of the many measures of symptom relief and the lack of reported data from later time points may indicate that symptom relief at those time points was equivalent for different proton pump inhibitors.

Gastric ulcer

• Comparative data about proton pump inhibitors for the treatment of gastric ulcer was very limited, with 3 studies comparing rabeprazole with omeprazole. No significant difference in healing rates was found.
• Symptom relief was better with rabeprazole 20 mg than omeprazole 20 mg in 3 of 12 measures at 3 weeks and in 2 measures at 6 weeks but no difference in symptom relief was found between rabeprazole 10 mg and omeprazole 20 mg daily.

Nonsteroidal anti-inflammatory drug-induced ulcer

• There were no head-to-head trials.
• Only 4 trials compared a proton pump inhibitor with another drug: 2 with omeprazole, 1 with esomeprazole, and 1 with lansoprazole. No differences between proton pump inhibitors could be discerned from these studies; confidence intervals for healing rates overlapped.

Direct evidence

Duodenal ulcer

Ten randomized controlled trials compared one proton pump inhibitor with an equipotent dose of another.^{41, 97–105} The details of these studies are summarized in Evidence Table 7. Six of these trials compared lansoprazole 30 mg with omeprazole 20 mg.^{97–101, 104} One study each compared pantoprazole 40 mg and rabeprazole 20 mg with omeprazole 20 mg,^{41, 102} 1 study compared esomeprazole 40 mg with omeprazole 40 mg daily (20 mg twice daily for each),¹⁰³ and 1 small study compared omeprazole enteric coated capsules with omeprazole magnesium.¹⁰⁵

The studies were fair quality. They were generally similar with respect to design and demographics, with the following exceptions: One study was unusual in that as a part of an Helicobacter pylori eradication regimen, patients with active duodenal ulcer were given esomeprazole plus antibiotics for only 1 week while omeprazole patients received antibiotics plus omeprazole for 1 week then continued omeprazole (only) for another 3 weeks.¹⁰³

As shown in Figure 8, omeprazole 20 mg, lansoprazole 30 mg, rabeprazole 20 mg, and pantoprazole 40 mg did not differ in the percentage of patients with duodenal ulcer that was healed by 4 weeks compared with omeprazole 20 mg daily. The pooled risk difference for daily lansoprazole 30 mg compared with omeprazole 20 mg was −0.2 (95% CI −3.0 to +2.6). The risk differences found between esomeprazole 40 mg, pantoprazole 40 mg, and rabeprazole 20 mg and omeprazole were approximately −0.97%, 6%, and 5%, respectively; however, these estimates were based on single studies and were not statistically significant. Similarly, no difference in healing rate was found between omeprazole enteric coated capsules and omeprazole magnesium both at 40 mg daily with all 57 patients being healed at 4 weeks.¹⁰⁵ Results from a large multicenter trial comparing esomeprazole 20 mg twice daily with omeprazole 20 mg twice daily also showed no difference in healing rate.¹⁰³ The results for healing at 2 weeks were similar for all comparisons.

Figure 8

Duodenal ulcer healing at 4 weeks in trials comparing proton pump inhibitors

Symptoms (pain, nausea, vomiting, antacid use, and overall well-being) were assessed by investigators at visits and through patient diaries in 8 studies. Only 1 study found a significant difference between proton pump inhibitors.⁴¹ Daytime pain was “improved” in 92% of the rabeprazole group and 83% of the omeprazole group at 4 weeks (P=0.038), however, no difference was found in nighttime pain or in the number of patients without pain. Antacid use, gastrointestinal symptoms, and overall well-being were not different in any of the studies.

Only 1 head-to-head study addressed maintenance, comparing lansoprazole 15 mg, lansoprazole 30 mg, and omeprazole 20 mg for up to 12 months (see Evidence Table 8).¹⁰⁰ At 6 months after healing, recurrence rates were 4.5%, 0%, and 6.3%, respectively. At 12 months the recurrence rates were 3.3%, 0%, and 3.5%, respectively. These differences were not statistically significant.

Indirect evidence

Duodenal ulcer

Twenty-five randomized controlled trials compared a proton pump inhibitor with an H2 receptor antagonist. Of these, 22 papers were reviewed.^110–135 Since these studies could only be used to make indirect comparisons of the effectiveness of the various proton pump inhibitors, we presented a limited analysis. The most common H2 receptor antagonist used in comparisons was ranitidine 300 mg daily, with 10 studies comparing omeprazole 20 mg. There were no studies comparing esomeprazole with an H2 receptor antagonist.

Figure 9 shows rates of healing at 4 weeks in 21 studies comparing a proton pump inhibitor with an H2 receptor antagonist for treatment of duodenal ulcer. Proton pump inhibitors were more effective than H2 receptor antagonists, but there was no significant difference in healing rate among the proton pump inhibitors. With omeprazole and lansoprazole, healing rate was correlated with H2 receptor antagonists’ healing. That is, as the healing rate in the H2 receptor antagonist group increased, proton pump inhibitor healing rate increased. One comparison showed pantoprazole to have a significantly higher healing rate than rabeprazole (risk difference 11.3%), but this comparison was made in only 1 study, and the confidence interval is large (95% CI 2.4 to 23.2).

Figure 9

Duodenal ulcer healing at 4 weeks in comparisons of proton pump inhibitor with H2 receptor antagonist

Another study¹³² examined the added benefit of continuing omeprazole 20 mg for 3 additional weeks after 1 week of eradication therapy with omeprazole 20 mg plus amoxicillin 1000 mg and clarithromycin 500 mg. At 4 weeks, there was no difference in healing rates in patients assigned to omeprazole (89%) and placebo (87%). Another reported symptom relief only.¹³¹

Gastric ulcer

Fifteen studies compared a proton pump inhibitor with an H2 receptor antagonist for treatment of gastric ulcer (Evidence Table 9).^{101, 110, 136–148} Two looked at maintenance therapy^149–151 and 1 was a follow-up of healed patients from another study.⁹⁸ One of the maintenance studies included patients with either gastric or duodenal ulcer, all of which were resistant to H2 receptor antagonist therapy.¹⁵¹ The other evaluated healing of gastric ulcer with esomeprazole compared with ranitidine in patients who continued to take a nonsteroidal anti-inflammatory drug.¹³⁶ This study is examined under Key Question 3. No study compared rabeprazole with a H2 receptor antagonist. Of the 15 trials, 5 compared omeprazole with ranitidine; 3 compared lansoprazole with ranitidine; 1 compared pantoprazole with ranitidine; 2 compared lansoprazole with famotidine; 3 compared omeprazole with cimetidine, and 1compared lansoprazole with cimetidine.

The total follow-up times varied, but healing rates at 4 weeks were available from all studies. Differences in the percentages of patients healed with different proton pump inhibitors at 4 weeks are plotted in Figure 10. The pooled risk differences range from 1.1% to 62.5%, with the smallest studies showing larger effects. The confidence intervals of the risk differences for healing with proton pump inhibitors compared with H2 receptor antagonists all overlap.

Figure 10

Gastric ulcer healing at 4 weeks in comparisons of proton pump inhibitor with H2 receptor antagonist

Symptoms were assessed by investigators at visits and through patient diaries in 13 studies. One did not report symptoms.¹³⁹ Pain was the most commonly assessed symptom. The pain scales differed among studies (0 to 3 in some, 0 to 4 in others) and sometimes were not described. Most studies found that the proton pump inhibitor relieved symptoms somewhat more quickly, with no difference later on between groups in percentage of patients without pain. However, only 3 studies found statistically significant differences on symptom measures, and then only in some of the many measures assessed.

One study¹⁵² reported maintenance therapy for the comparison of lansoprazole 15 mg or 30 mg with placebo. Lansoprazole was effective for preventing endoscopically verified recurrence, eliminating symptoms, and reducing antacid use. A 6-month open study reported that omeprazole 20 mg daily as more effective than ranitidine in preventing relapse in patients with refractory ulcer (unhealed after 8 weeks of treatment with an H2 receptor antagonist).¹⁵¹ In this study only 12 patients of 102 enrolled were assigned to ranitidine, and patients with either gastric or duodenal ulcer were included. A 6-month follow-up study without treatment¹⁴⁹ looked at patients who had healed with 6 weeks of treatment with omeprazole or cimetidine;¹³⁸ no significant difference was found in relapse rate. All of these studies had high or differential dropout rates.

Direct evidence

In a study of 595 patients with arthritic diseases, continuously taking an nonsteroidal anti-inflammatory drug (not including COX-2 Inhibitors), and considered at high risk for gastrointestinal injury (previous ulcer or taking anticoagulants), patients were randomized to 6 months of pantoprazole 20 mg or 40 mg or omeprazole 20 mg daily.¹⁵⁸ Using life-table analysis methods, remission rates were compared across and between groups. The primary outcome, therapeutic failure, was defined as peptic ulcer, >10 erosions, reflux esophagitis, and discontinuations of study drug due to an adverse event or severe gastrointestinal symptoms. Examination of baseline risk characteristics revealed that the pantoprazole 40 mg group had fewer patients taking anticoagulants (1% compared with 4%), experiencing a change in nonsteroidal anti-inflammatory drug in the last month (6% compared with 9% or 10%), and fewer with a history of endoscopically proven peptic ulcer (20% compared with 24% or 25%). These differences are small but may have biased the risk level in favor of the pantoprazole 40 mg group. Patients were censored from the analyses (considered lost to follow up) if they had low adherence to the nonsteroidal anti-inflammatory drug regimen, found to not meet inclusion after randomization, failed to adhere to the protocol, or withdrew from the study due to an adverse event not considered related to study drugs or due to “refusal to continue”. The numbers of patients censored for these reasons were greater in the omeprazole group (N=42) and lowest in the pantoprazole 40 mg group (N=29). With these issues in mind, we rate this trial as fair quality (rather than poor quality, as it does meet other aspects of internal validity) and suggest caution in interpreting the results. There was no statistically significant difference between the groups in remission rates based on either therapeutic failure or failure limited to endoscopic findings, with more than 90% of patients remaining in remission in all groups at 3 and 6 months.

Indirect evidence

One good-quality systematic review addressed the question of proton pump inhibitors for treatment of nonsteroidal anti-inflammatory drug-induced ulcer.¹⁵⁹ Its search for literature covered 1966 to 2000 (MEDLINE search from 1966 to January 2000, Current Contents for 6 months prior to January 2000, EMBASE to February 1999, and a search of the Cochrane Controlled Trials Register from 1973 to 1999). The review found 5 randomized trials that assessed omeprazole 20 mg with 40 mg in prevention of nonsteroidal anti-inflammatory drug-induced gastroduodenal toxicity. None of the studies were designed to evaluate the effectiveness of proton pump inhibitors in preventing serious complications of ulcers (hemorrhage, perforation, or death). The review showed that omeprazole is superior to the H2 receptor antagonists but provided no data on any other proton pump inhibitor.

Eight trials published more recently^{154, 155, 160–165} than the omeprazole review are presented in Evidence Table 11. None of these studies was a head-to-head comparison and there were important differences in treatment regimens and follow-up, making comparisons across studies impossible. All the trials enrolled patients who were regular users of nonsteroidal anti-inflammatory drugs, with 1 including COX-2 Inhibitors.¹⁶⁵ Symptom assessment and reporting varied among these studies.

One study¹⁶⁰ included only patients without Helicobacter pylori who were randomized to received placebo, misoprostol 800 μg, lansoprazole 15 mg, or 30 mg with follow-up at 1, 2, and 3 months, while another study¹⁶² included patients with Helicobacter pylori who developed ulcer complications (bleeding, perforation, or obstruction) after a month of daily low-dose aspirin. After ulcers were healed and Helicobacter pylori were eradicated, patients continued with aspirin 100 mg and were randomized to lansoprazole 30 mg or placebo. In the last study of Helicobacter pylori and prevention of nonsteroidal anti-inflammatory drug-induced ulcers,¹⁶³ patients with Helicobacter pylori but without past or current ulcer were assigned to 1 of 4 treatment groups: omeprazole 20 mg plus clarithromycin 500 mg and amoxicillin 1 gram for 1 week; placebo or omeprazole 20 mg daily for 4 weeks; omeprazole 20 mg once daily for 5 weeks; or placebo for 5 weeks.

In the study of Helicobacter pylori negative patients,¹⁶⁰ lansoprazole was inferior to misoprostol in preventing gastric ulcers. At 3 months, the gastric ulcer rate (failure rate) was 7% for misoprostol, 20% for lansoprazole 15 mg, and 18% for lansoprazole 30 mg, with no significant difference between lansoprazole doses. However, when adverse effects were included as failures, the failure rate for all 3 treatment groups was 31%. A post hoc subgroup analysis of patients taking nonsteroidal anti-inflammatory drugs and low dose aspirin found no significant difference among treatments at 12 weeks.¹⁶⁶ Symptoms (antacid use and day and nighttime abdominal pain) were assessed by patient diary and were found to be significantly better in the lansoprazole groups than the misoprostol group, but comparisons between the 2 lansoprazole doses were not made.¹⁶⁰

In the study of Helicobacter pylori positive patients with ulcer complications (bleeding, perforation, or obstruction),¹⁶² the primary endpoint was prevention of ulcer complications and the secondary endpoint was recurrence. The rate of recurrence of ulcer complications at a median follow-up of 12 months was 1.6% in the lansoprazole group compared with 14.8% in the placebo group. Two patients in the placebo group (N=61) were also taking nonsteroidal anti-inflammatory drugs.

In patients with Helicobacter pylori but no history of ulcer, all 3 active treatment regimens were better than placebo in reducing the occurrence of ulcer and dyspeptic symptoms requiring therapy. There were no significant differences between the treatment groups.

A study comparing pantoprazole with placebo¹⁶¹ presented a life-table analysis rather than simple proportions of patients without ulcer, making comparison with other placebo-controlled studies of proton pump inhibitors difficult. The pantoprazole group had 17% fewer ulcers at 4 weeks and 27% at 12 weeks. Patients who dropped out due to adverse events were included in the 4 week data as treatment failures. The methods or scales used to assess symptoms were not described but reported just “symptoms.”¹⁶¹ Presence of gastrointestinal symptoms differed at baseline in the 2 groups: They were present in 43% of the pantoprazole group and 18% of the placebo group. At 4 and 12 weeks presence of gastrointestinal symptoms improved in the pantoprazole group (to 17% and 20%, respectively), while in the placebo group remained stable (20% and 19%, respectively).

The only evidence on prevention of ulcers related to COX-2 inhibitors came from a combined report of 2 similar fair quality trials that enrolled patients who were regularly taking a nonselective nonsteroidal anti-inflammatory drug or a COX-2 inhibitor and were at risk of peptic ulcer (age > 60 years or documented peptic ulcer in last 5 years).¹⁶⁵ Combined, the studies randomized 1429 patients to esomeprazole 20 mg, esomeprazole 40 mg, or placebo daily for 6 months. Using pooled and separate life-table analyses, the overall analysis indicated that both esomeprazole groups prevented peptic ulcer statistically significantly more often than placebo for all nonsteroidal anti-inflammatory drugs. While no statistical analyses were undertaken comparing the 2 doses of esomeprazole, the rates of ulcer development were very similar (5.2% with 20 mg and 4.6% with 40 mg). The rates of ulcer development among the subgroup taking a COX-2 inhibitor were also statistically significantly lower with either dose of esomeprazole compared to placebo (16.5% with placebo compared with 0.9% and 4.1% with 20 mg and 40 mg of esomeprazole, respectively). The separate study analyses of the subgroup taking nonselective nonsteroidal anti-inflammatory drugs indicated that esomeprazole was superior to placebo in one trial (N=844) but not in the other (N=585), while the pooled analysis indicated statistically significant benefit with either dose of esomeprazole compared to placebo.

Direct evidence

Five systematic reviews have evaluated the efficacy of proton pump inhibitors in eradication of Helicobacter pylori, however because these reviews focused on comparisons to H2 receptor antagonists, were out of date (literature searches conducted prior to 2001), or included non-randomized studies or studies published only as abstracts, they were not sufficient to evaluate this question.^{47, 167–171}

Twenty-nine studies directly compared one proton pump inhibitor with another, in combination with the same antibiotic(s), and reported Helicobacter pylori eradication rates.^{98, 104, 172–198} They were fair-quality with the exception of 5 poor-quality studies that were not blinded or provided inadequate data to compare eradication rates directly.^{183, 187, 192, 197, 199} Several studies included antibiotic regimens that are no longer standard.^{187, 104, 178, 183, 190, 198, 200}

Of these, 23 trials compared proton pump inhibitors using identical regimens of antibiotics (within study) and reported eradication rates in a way that allowed statistical pooling (Table 11). ^{68, 104, 173, 175–180, 183–187, 189–191, 193–197, 201} All of these trials included treatment with 2 antibiotics and assessed Helicobacter pylori eradication at 4 to 6 weeks after treatment. While most of these trials used a 7 day proton pump inhibitor regimen in combination with 2 antibiotics, 3 trials used longer proton pump inhibitor regimens, a 14 day^{190, 202} and a 30 day regimen.¹⁷⁵ Interestingly, these regimens resulted in lower eradication rates at follow-up. Overall, eradication rates ranged from 67% in a trial of lansoprazole 60 mg daily for 30 days to 100% in a trial of pantoprazole 40 mg daily for 7 days. Pooled rates of eradication from these trials vary (see Table 11 below), but pooled relative risks of these rates did not identify statistically significant differences between groups when stratified by number of days of treatment and dose comparison (Table 11 below). Several trials examined different dose levels (high compared to usual and low compared to usual) of proton pump inhibitors without finding statistically significant differences in eradication rates.

Table 11

Rates of eradication of Helicobacter pylori

In 1 additional study, patients who had failed a 1 week regimen of amoxicillin, clarithromycin, and a proton pump inhibitor were randomized to rabeprazole 20 mg, lansoprazole 60 mg, or omeprazole 40 mg daily each plus amoxicillin and metronidazole.¹⁸⁸ No differences were found among the proton pump inhibitors in eradication rates, with 91% eradication in each group.

Two systematic reviews addressed similar questions, 1 focusing on the comparison of esomeprazole to other proton pump inhibitors²⁰³ and the other directly comparing proton pump inhibitors.²⁰⁴ Both of these reviews were fair to poor quality because they pooled studies with differing proton pump inhibitor regimens (for example drug A given for 7 days compared to drug B given for 14 days) with those comparing similar regimens or provided inadequate details of studies included to determine the comparisons being made.

Indirect evidence in children

Two trials evaluated lansoprazole in eradication of Helicobacter pylori in children.^{205, 206} Both studies used antibiotic regimens of amoxicillin and tinidazole, given for 6 or 7 days, in combination with lansoprazole or placebo. The 2 protocols were very similar, but not identical; in 1 the dose of lansoprazole was 30 mg daily with children 10 to 21 years eligible for enrollment,²⁰⁵ while in the other dosing was based on weight (<20 kg, 15 mg daily; ≥20 kg, 30 mg daily)²⁰⁶ and the age range was 8 to 14 years. However, the mean age for participants in both trials was 11 years. Neither trial resulted in significantly different eradication rates between placebo (58%²⁰⁶ and 71%²⁰⁵) and lansoprazole (67%²⁰⁶ and 68%²⁰⁵).

Standard dose compared with low-dose proton pump inhibitor

• Time in remission was longer for higher doses compared with lower doses for omeprazole and rabeprazole, but the same for higher and lower doses of lansoprazole. Evidence on esomeprazole was inconclusive.
• Rates of endoscopically verified remission at study end were greater with the higher dose of rabeprazole compared with the lower dose, but were not different between dose strategies for omeprazole and lansoprazole.
• Rates of relapse of symptoms were generally higher with lower doses of omeprazole, lansoprazole, and rabeprazole.

Standard dose compared with intermittent or on-demand proton pump inhibitor

• For patients with healed erosive esophagitis, a regimen of daily proton pump inhibitor was superior in preventing relapse of esophagitis or recurrence of symptoms compared with 3 days a week or on-demand regimens at 6 months.
• For patients with nonerosive esophagitis, assessments of symptom severity or relapse of symptoms was not different between daily and on-demand regimens. Patient satisfaction and quality of life ratings at study end were also not different, although the mean change in quality of life score from baseline was better with daily therapy.
• For patients presenting with symptoms of gastroesophageal reflux disease, but without endoscopic assessment, evidence is mixed.

Proton pump inhibitor compared with H2 receptor antagonist

• Daily proton pump inhibitor therapy was found superior to daily H2 antagonist therapy in preventing relapse of erosive esophagitis, or symptoms of gastroesophageal reflux disease.

Standard-dose proton pump inhibitor compared with low-dose proton pump inhibitor

Eleven trials compared a standard dose of a proton pump inhibitor with a lower dose of the same proton pump inhibitor for longer-term treatment of gastroesophageal reflux disease (Evidence Table 13). Five trials compared lansoprazole 30 mg with lansoprazole 15 mg;^{21, 207–210} 2 compared omeprazole 20 mg with omeprazole 10 mg;^{211, 212}1 compared pantoprazole 40 mg with pantoprazole 20 mg;²¹³ 2 compared rabeprazole 20 mg with rabeprazole 10 mg; ^{214, 215} and 2 compared esomeprazole 40 mg with esomeprazole 20 mg and esomeprazole 10 mg.^{216, 217} Eight trials also included a placebo arm. In most of the trials, the drug and dose used for acute treatment before maintenance treatment began was the same as the higher dose used in the maintenance phase. The studies’ follow-up periods were 6 months in 4 trials, 12 months in 6 trials, and 5 years in 1 trial.²¹⁴ Of these, 2 were poor quality. One had significant differences in prognostic factors at baseline combined with other flaws relating to assignment of group.²⁰⁹ In the other, patients with adverse events thought to possibly be or probably be related to the study drug were counted as having a relapse, the margin allowed for noninferiority was very large (20%), and there were flaws related to assignment of group.²¹³ These studies are not discussed below, and the remainder were fair quality.

All trials reported recurrence rate of endoscopically verified disease (either as relapse rates or remission rates) and the time in remission. Remission was considered grade 0 on any esophagitis scale in most studies, although some allowed grade 1 as well. All but 1 trial²¹² also reported recurrence rate of symptoms or the number of patients with mild or no symptoms at study end. Study characteristics are summarized in Table 12 and results are shown in Table 13.

Table 12

Proton pump inhibitors and treatment durations in longer-term studies of gastroesophageal reflux disease: Comparisons of standard doses with lower doses

Table 13

Remission of gastroesophageal reflux disease erosions and symptoms in longer-term studies of proton pump inhibitors: Comparisons of standard doses with lower doses

Standard-dose proton pump inhibitor compared with intermittent or ‘on-demand’ proton pump inhibitor

We identified 2 systematic reviews that compared intermittent or on-demand treatment to daily treatment for patients with gastroesophageal reflux disease.^{218, 219} These reviews included studies of H2 receptor antagonists, studies comparing different doses of a proton pump inhibitor to one another, and different proton pump inhibitors with differing regimens (e.g. omeprazole given daily compared with esomeprazole given intermittently). Additionally, quality assessments of included studies were not undertaken. Most of the studies included in these reviews did not make a comparison between continuous (daily) proton pump inhibitor therapy and intermittent (3 times a week) or on-demand (taken daily when symptoms occur, discontinue when symptoms resolve) and were not included here. We have used these reviews only to identify additional studies not found in our literature searches.

Eight trials compared daily treatment with a proton pump inhibitor with intermittent or on-demand treatment of the same proton pump inhibitor.^{208, 220–225, 226} One study followed patients for 1 year,²²⁵ the rest followed patients for 6 months. Details of the study patients and treatment strategies are presented in Table 14. In patients with healed endoscopically proven gastroesophageal reflux disease (Table 15), a regimen of daily proton pump inhibitor was superior to either 3 days a week or on-demand proton pump inhibitors of the same daily dose in preventing recurrence of erosive esophagitis based on endoscopy.^{208, 221, 225} A 3-day-a-week regimen was also inferior to a daily regimen in preventing relapse of overall symptoms but no difference was found between daily treatment and on-demand treatment, although 1 study found that severity of heartburn was lower with the daily regimen.²²¹

Table 14

Proton pump inhibitors and treatment durations in longer-term studies of gastroesophageal reflux disease: Comparisons of daily treatment with intermittent or on-demand treatment

Table 15

Remission of gastroesophageal reflux disease erosions and symptoms in longer-term studies of proton pump inhibitors: Comparisons of daily treatment with intermittent or on-demand treatment (more...)

In 3 studies of patients with endoscopically proven non-erosive esophagitis, no significant difference was found between daily and on-demand treatment with proton pump inhibitors rabeprazole or lansoprazole, in the severity of symptoms as measured on a visual analog scale at 3, 6, and 12 months,²²² or in the proportion with symptom relapse at 6 months.²²⁴ Assessment of overall satisfaction with treatment was not different between regimens in 1 study²²² and final quality of life scores were also not different between groups in the other study.²²⁴ However, the mean change in quality of life scores from baseline to 6 months was significantly better in the daily treatment group compared to the on-demand group (P=0.03).²²⁴ The third study of pantoprazole 20 mg found the on-demand regimen to be noninferior to the daily regimen, based on the rates of ‘treatment failure’ defined as moderate symptoms for 3 or more days, use of > 1 dose of study medication on > 3 consecutive days, or withdrawal from study due to lack of efficacy.²²⁶

In patients presenting with symptoms of gastroesophageal reflux disease (but with no endoscopic examination), 2 studies found mixed results.^{220, 227} In a study of on-demand esomeprazole, the results differ by which symptom-based outcome measure is used.²²⁸ Statistical analyses of the results were not undertaken in the study, but here we have used a Yates corrected chi-square test. Using an outcome of “no heartburn” at 6 months, daily therapy is superior to on-demand treatment with 72% compared with 62% (P=0.002 using Yates corrected chi-square). However, the percentage of patients with no regurgitation at 6 months was 78% with daily therapy and 91% with on-demand treatment (P<0.001002 using Yates corrected chi-square). The other study found that daily treatment with rabeprazole resulted in statistically significantly more heartburn-free days (90%) compared with on-demand treatment (65%; P<0.0001) and fewer heartburn episodes (N=7 and 26, respectively; P<0.0001).These 2 studies also report different findings in quality of life. Again, the study of esomeprazole found the daily regimen superior to the on-demand regimen in both change from baseline in quality of life and patient satisfaction. However, data (including P values) supporting these claims are not clearly presented.²²⁸ The other study found that on-demand treatment with rabeprazole resulted in greater improvement in quality of life at 6 months compared to the daily regimen.²²⁰

Standard-dose proton pump inhibitor compared with H2 receptor antagonist or placebo

Four studies found proton pump inhibitors to be superior to ranitidine 150 mg twice daily, 3 for prevention of relapse of healed esophagitis and 1 for prevention of recurrence of symptoms of gastroesophageal reflux disease.^{229,230,223, 225} After 12 months, proton pump inhibitor therapy (pantoprazole 10 mg, 20 mg, or 40 mg and omeprazole 20 mg daily) resulted in lower relapse rates compared with ranitidine therapy. In 2 studies, more patients remained healed on pantoprazole at all doses than on ranitidine, and the rate of relapse was related to the dose of pantoprazole: Relapse occurred in 60%, 32%, and 18% of the 10 mg, 20 mg, and 40 mg groups, respectively. A second study of the same doses of pantoprazole and ranitidine found similar results.²³⁰ During the first 12 months of maintenance treatment, healing was maintained in 78% of patients treated with pantoprazole 40 mg, 55% of patients treated with pantoprazole 20 mg, 46% of patients treated with pantoprazole 10 mg, and 21% of those treated with ranitidine. This study is planned for 3 years, but only the first 12 months have been reported so far. With omeprazole, at 12 months 89% remained in remission compared with 25% on ranitidine (P<0.001).²²⁵ In those with symptoms suggestive of gastroesophageal reflux disease, 72% had relief of symptoms after 6 months of esomeprazole 20 mg daily compared with 33% taking ranitidine (statistical analysis not presented).

Additionally, a study of famotidine 20 mg twice daily compared with lansoprazole 15 mg daily, both as step down therapy from lansoprazole 30 mg daily for treatment of erosive esophagitis,found the proton pump i nhibitor to be superior in preventing recurrence of regurgitation and heartburn, but not dysphagia or assessment of esophagitis grade after 8 weeks of maintenance treatment.²³¹ Fifty percent of patients taking famotidine experienced recurrence of heartburn, and 79% experienced recurrence of regurgitation compared to 0% and 7%, respectively, with lansoprazole 15 mg daily.

Comparison of esomeprazole administered orally compared to esomeprazole administered intravenously

A trial conducted in 246 ambulatory patients compared esophagitis healing rates at 4 weeks in patients given esomeprazole 40 mg either orally, via intravenous injection, or via intravenous infusion.²³² Patients were randomized to either 1 week of intravenous esomeprazole (injection or infusion) followed by 3 weeks of oral esomeprazole or 4 weeks of oral esomeprazole. The study was blinded using multiple placebos. After 4 weeks, there was no difference in healing rates among the 3 treatment groups (approximately 80%). The frequency and type of adverse events were also similar among the treatment groups.

Comparison of a reduced-dose proton pump inhibitor with an H2 receptor antagonist in children

One fair-quality randomized trial compared reduced-dose omeprazole with ranitidine for longer-term treatment of erosive gastroesophageal reflux disease in children (Evidence Table 6).²³³ Children who had been treated with omeprazole and shown by endoscopy to be healed after 3 months began treatment with omeprazole 0.7 mg/kg daily (half the starting dose for the healing phase), ranitidine 10 mg/kg daily, or nothing for 6 months. Although no statistically significant difference was found among the groups at baseline, children in the group receiving no treatment had slightly less severe esophagitis and slightly lower symptom scores than children in the other groups. They were also slightly older at enrollment and at age of symptom onset. Follow-up endoscopy at 3 months after the end of maintenance treatment was blinded. No statistically significant differences were seen in endoscopic or histologic grade or in symptom scores. One patient in the no treatment group had a relapse of erosive esophagitis (Hetzel and Dent grade 3). Twelve (25%) had mild symptoms at study endpoint and remained untreated, 6 in the no treatment group, 1 in the ranitidine group, and 5 in the omeprazole group. While the differences at baseline between the no treatment group and the drug groups may result in confounding results of those comparisons, there is no apparent difference between the drug groups in maintenance of remission of esophagitis and symptoms and in the number of patients requiring no further treatment.

Taper off proton pump inhibitor

A group of 97 patients with at least 8 weeks of daily use of a proton pump inhibitor with no history of peptic ulcer or esophagitis, and no evidence of esophagitis on endoscopy were enrolled in a 3 week trial of tapering the proton pump inhibitor dose prior to discontinuation or abrupt discontinuation.²³⁴ In this study, patients were assigned to take omeprazole 20 mg daily for 3 weeks or to a blinded taper of omeprazole 20 mg daily for 1 week, 10 mg daily for 1 week, and 10 mg every other day for 1 week. Symptoms were assessed 1 week after discontinuation of drug, and the rate of resumption of proton pump inhibitor therapy was measured after 1 year. The mean duration of proton pump inhibitor treatment at study entry was 48 months. No statistically significant differences were found between tapering and non-tapering groups on symptom scores at 4 weeks, or the rate of resumption of treatment at 1 year.

Diarrhea

A nested case-control study of 10 008 lansoprazole users followed for 4 years found a dose-related trend for diarrhea (5%, 4%, and 3% of patients using 60 mg or more, 30 mg, and 15 mg or less, respectively; P=0.08).²⁴⁸ In 42% of patients reporting diarrhea the lansoprazole dosage was reduced or discontinued as a response. Cases had a higher current use of oral antibiotics than controls with no diarrhea (adjusted odds ratio, 2.7; 95% CI 1.0 to 6.9).

Two case control studies examined the relationship between clostridium difficile associated diarrhea and acid suppression, including proton pump inhibitors.^{253, 254} The first, based on 1672 cases and 16720 controls, found a significantly increased risk of community acquired clostridium difficile diarrhea in patients who were currently using a proton pump inhibitor (relative risk 2.9; 95% CI 2.4 to 3.4).²⁵³ However, the second, based on 1389 cases and 12303 controls, did not find a significant association between hospitalization due to clostridium difficile diarrhea and exposure to a proton pump inhibitor within 90 days (odds ratio 0.0.9; 95% CI 0.8 to 1.1).²⁵⁴ Neither study examined differences between proton pump inhibitors.

Bone fractures

Four nested case control studies examined the association between exposure to proton pump inhibitors and risk of fracture.^255–258 Three of the studies found statistically significant increased risk of fracture associated with proton pump inhibitor use, although they differed in the duration of exposure that was found significantly associated with increased risk. The largest included 124 655 cases and 373 962 controls drawn from Danish registers of National Board of Health, the Danish Medicines Agency, and the National Bureau of Statistics.²⁵⁶ Cases included any patient with a fracture in the year 2000. An increased risk of any fracture was associated with last use of a proton pump inhibitor within 1 year of the index date (adjusted odds ratio 1.18; 95% CI 1.12 to 1.43). Exposure that ended more than 1 year prior to the fracture was not significantly associated, and a dose-response effect was not found. Cumulative dose was used as a proxy for duration of exposure, and the increased risk was found to be similar across exposure groups (≤ 25, 26–99 and ≥ 100 defined daily dosages). Similar results were found for specific fracture sites (hip, forearm and spine). This study controlled for exposure to multiple drug classes, but was not able to control for calcium or vitamin D and did not differentiate types of fracture.

In contrast, 2 studies involving 13 566 and 15 792 cases found increased risk based on duration and dose of proton pump inhibitor use in patients 50 years and older.^{255, 257} One identified patients older than 50 years, who had been exposed to a proton pump inhibitor for at least 1 year prior to the index date (date of hip fracture). After 1 year of use, an increased risk was found; adjusted odds ratio of 1.44 (95% CI 1.30 to 1.59), increasing by year to 1.59 (95% CI 1.39 to 1.80) at 4 years of use.²⁵⁵ The risk increased again with higher daily dosages of proton pump inhibitor, with adjusted odds ratios of 1.40 (95% CI 1.26 to 1.54) for those using < 1.75 average daily doses, and 2.65 (95% CI 1.80 to 3.90) for those using > 1.75 average daily doses. Multiple potential confounding factors were controlled for; including several groups of drugs know to influence bone metabolism, including calcium or vitamin D. The second study included patients with vertebral, wrist or hip fractures, again controlling for multiple potential confounders, including drugs (but not calcium or vitamin D).²⁵⁷ No increase in risk was found with durations of exposure up to 6 years. The risk for any osteoporotic fracture was increased only with 7 or more years of exposure (adjusted odds ratio 1.92; 95% CI 1.16 to 3.18). The risk of hip fracture alone was increased after 5 years of exposure (adjusted odds ratio 1.62; 95% CI 1.02 to 2.58) although the magnitude of risk increased again with 6 and 7 years of exposure.

The fourth study limited the population of cases and controls to those with no major risk for hip fracture.²⁵⁸ With 1098 cases and 10 923 controls, this was the smallest study. No association was found between proton pump inhibitors and incidence of hip fractures. The estimated relative risk of hip fracture for those who received one or more proton pump inhibitor prescriptions before the index date was 0.9 (95% CI 0.7 to 1.1), compared with those who received no prescriptions. This study also evaluated individual proton pump inhibitors and found similar results for each drug. The discordant results of this study compared to the other 3 may be due to smaller numbers and a differing selection process in that patients with as little as 1 prescription for a proton pump inhibitor were included and further stratification of exposure or dose were not undertaken.

Community acquired pneumonia

Two studies examined the association between proton pump inhibitor use and community acquired pneumonia, coming to somewhat different conclusions.^{259, 260} A large, good-quality nested case-control study identified 80 066 cases and 799 881 controls drawn from a cohort of patients from a general practice research database.²⁶⁰ The adjusted odds ration for the association of recent (within 30 days) use of a proton pump inhibitor and a diagnosis of community acquired pneumonia was 1.02 (0.97–1.08). This study did find, however, that the risk was significantly increased if the patient had started the proton pump inhibitor within 2 or 14 days, but not with longer durations of therapy. The other study, fair quality, identified 475 cases and 4960 controls from a cohort of patients who had all been exposed to an acid reducing drug during the study period.²⁵⁹ The exposure was then stratified into recent (within 30 days) or past (>30 days since exposure). This study found an increased risk among current users of a proton pump inhibitor, with an adjusted relative risk of 1.89 (95% CI 1.36 to 2.62) compared to those who had stopped taking a proton pump inhibitor 30 or more days ago.²⁵⁹ This study did report an analysis of each proton pump inhibitor with enough cases to conduct an analysis, finding an increased risk with omeprazole and pantoprazole, adjusted odds ratios 1.74 (95% CI 1.28 to 2.35) and 2.29 (95% CI 1.43 to 3.68), respectively, but not with lansoprazole 0.91 (95% CI 0.35 to 2.34). However, because there were few cases for each drug, these results should be interpreted with caution.

A study combining data from all Phase II-IV trials of esomeprazole examined the risk of respiratory tract infections in 16 583 patients assigned to esomeprazole and 12 044 assigned to placebo or other acid suppressing drugs.²⁶¹ Compared to placebo, this analysis did not find a difference in risk of any respiratory tract infection (relative risk 0.93; 99% CI 0.78 to 1.11); lower respiratory tract infection (relative risk 0.92; 99% CI 0.59 to 1.42); or pneumonia (relative risk 0.94; 99% CI 0.29 to 3.07). Analyses of the relative risk with esomeprazole compared with omeprazole, lansoprazole, or ranitidine did not indicate statistically significant differences. Because this is a pooled analysis of selected studies without a systematic review, the quality of this study is undetermined and the results should be interpreted with caution.

Colorectal cancer

A nested case control study of 4432 cases and 44 292 controls from the General Practice Research Database (UK) evaluated the association between duration of proton pump inhibitor use and incidence of colorectal cancer.²⁶² While multiple durations of exposure were examined, the one showing a statistically significant increased risk was diagnosis of colorectal cancer with less than 1 year exposure to a proton pump inhibitor with an adjusted odds ratio of 2.6 (95% CI 2.3 to 2.9). Less than 1 year of exposure, more than 12 months prior to the index date, and 1 to 2, 2 to 3, 3 to 4, 4 to 5, or >5 years of proton pump inhibitor use were not statistically significantly associated with colorectal cancer. The adjusted odds ratio for ≥5 years of proton pump inhibitor exposure was 1.1 (95% CI 0.7 to 1.9). Among high-dose proton pump inhibitor users (≥1.5 defined daily doses/day), there was a nonstatistically significant trend toward an increased risk with increasing duration of use (test for trend, P=0.2).

Serum gastrin levels

Serum gastrin level were monitored in several studies and found to be significantly elevated above baseline although the magnitude of increase was small and generally not considered clinically significant. A dose-related difference was found in some studies, but there were no differences between different proton pump inhibitors. Likewise, when studied, the effect of different proton pump inhibitors on Helicobacter pylori-related gastritis was similar, worsening gastritis in the corpus and improving gastritis in the antrum.²⁶³

Adverse events in children

Reporting of adverse events in children was limited to short-term trials and 2 open-label uncontrolled studies with longer follow-up.^{264, 87, 88, 188, 189, 209–212 265} In short-term trials of omeprazole no serious adverse events were reported.^{87, 209, 214}

Lansoprazole was studied in infants and neonates in 2 similar trials of children with symptoms of gastroesophageal reflux disease. ²⁶⁵ The infants, age > 28 days by but < 1 year, were given a suspension of lansoprazole dosed at 1 or 2 mg/kg/day and the neonates (up to 28 days after birth) were given 0.5 to 1.0 mg/kg/day for 5 days. Twenty-four neonates and 24 infants were enrolled. Mean age in the infant group was 24 weeks, and 3.7 weeks in the neonate group. While most neonates were white, 50% of the infants were black. While a large number of adverse events were reported (58%) 4 in the neonates (8%), and 1 in the infant group (4%) were considered related to the drug. In neonates, the adverse events were anemia, flushing (2 patients), and elevated aspartate aminotransferase level and were considered mild or moderate in severity. One infant also had elevated an aspartate aminotransferase level. The increases in aspartate aminotransferase occurred in the higher dose groups for each age group (0.5 and 2.0 mg/kg/day, respectively).

A retrospective chart review of 113 children identified from a registry-type database examined children with erosive esophagitis who received a proton pump inhibitor for at least 1 year.²⁶⁴ The majority (66%) was taking lansoprazole, followed by omeprazole (22%), and few were taking pantoprazole, rabeprazole, or esomeprazole. Overall, 88% of the children had no adverse event while taking a proton pump inhibitor, with a range of 80% to 100% for specific proton pump inhibitors. The most frequent adverse events recorded in patients’ charts were constipation (4%) and diarrhea (5%). Serum gastrin level was elevated (>90 pg/mL) in 73% of children, with no statistically significant differences by specific proton pump inhibitor, dose, dosing frequency, or treatment duration. No elevation in liver enzymes was reported.

In a before-after study of omeprazole for esophageal reflux, 15 children were followed for a mean of 12 months. Seven (47%) had elevation of liver enzymes. Eleven (73%) had hypergastrinemia.²¹³ A short-term before-after study of pantoprazole reported elevated liver enzymes in 1 of 18 children exposed for 28 days and 5 of 18 (28%) had hypergastrinemia.²¹¹ In a 2-week study of lansoprazole in children (mean age 11 years) only mild gastric adverse events were reported.²¹²

Two short-term trials compared lower dose and higher dose esomeprazole in children with gastroesophageal reflux disease. These trials made no comparison to placebo or other drugs. In 148 adolescents aged 12 to 17 years assigned 20 or 40 mg esomeprazole daily for 8 weeks, 15% experienced an adverse event considered related to esomeprazole; headache (8%), abdominal pain (3%), nausea (2%), and diarrhea (2%).²⁶⁶ In 108 younger children, aged 1 to 11 years, who were assigned to 5 or 10 mg esomeprazole if weight < 20 Kg or 10 or 20 mg if weight > 20 Kg, 9% reported an adverse event considered related to esomeprazole; diarrhea (2.8%), headache (1.9%), and somnolence (1.9%).²⁶⁷ Serious adverse events thought to be related to esomeprazole were not reported in either study.

Age and sex

In included head-to-head studies, the enrolled patients were middle aged, with mean ages ranging from 43²⁶⁸ to 70¹⁶² years. From 38% to 89% of the patients were male. The ethnicity of participants was stated in only 5 trials.^{4, 25, 75, 108, 268} The majority of studies included mostly white populations. In those studies with greater variation subgroups were too small for meaningful analyses by racial or ethnic group.

An open-label, single-center trial conducted in 320 patients over age 65 compared 4 proton pump inhibitors for healing and symptom resolution in erosive esophagitis.²⁶⁹ This was the only head-to-head trial conducted exclusively in elderly patients. The mean age of the group was 77.4 years (standard deviation 7.9 years, range 65–93 years). Nineteen patients withdrew from the study (5.9%), 2 due to adverse events. Patients were randomized to omeprazole 20 mg, lansoprazole 30 mg, pantoprazole 40 mg, or rabeprazole 20 mg. After 8 weeks of treatment, the healing rate in the overall group was 85% (intention-to-treat). Healing rates in the pantoprazole (90%) and rabeprazole (89%) groups were significantly higher than the omeprazole group (75%; P=0.022 and P=0.040, respectively). No difference was found between omeprazole and lansoprazole (75% and 85%; P=NS). Pantoprazole and rabeprazole were also superior to omeprazole and to lansoprazole for resolution of heartburn (rates 100% for pantoprazole and rabeprazole, 87% for omeprazole, and 82% for lansoprazole). The frequency of adverse events was low (4 patients; 1.3%), and there were no differences between treatment groups in the prevalence of adverse events.

There was 1 small, 12-month, placebo-controlled trial in which pantoprazole 20 mg was effective for maintenance treatment of gastroesophageal reflux disease in patients age 65 or older.²⁷⁰ An age-based analysis of healing or prevention was not possible in most head-to-head trials, due to the small numbers of older patients. However, 2 trials did assess the impact of age, gender, and race on the incidence of adverse effects.^{5, 106} There were no differences between proton pump inhibitors (omeprazole, rabeprazole, esomeprazole) on the basis of these characteristics. The effect of age on eradication rate was also evaluated.¹⁹⁸ This study found higher eradication rates among patients older than 50 years than patients younger than 50, but proton pump inhibitors were not compared.

In trials comparing a proton pump inhibitor with another drug, the same general statements can be made, but a few findings deserve comment. Studies looking at healing or prevention of nonsteroidal anti-inflammatory drug-induced ulcer included more women than men, with the proportion of women ranging from 62% to 67% and 64% to 83% in the respective types of study. This is most likely due to the greater prevalence among women of diseases requiring long-term nonsteroidal anti-inflammatory drug treatment. However, no gender-based analyses were presented.

Genotype

The proton pump inhibitors are all metabolized, largely by the CYP2C19 and CYP3A4 liver enzymes. Theses enzymes are estimated to be deficient in 3% of white and African Americans and 17% to 25% of Asians. The deficiency results in a significantly longer half-life of proton pump inhibitors, although clinically significant accumulation of these drugs has not been shown. While dose adjustments are not required, and adverse effect profiles of the drugs do not differ, there is some evidence that lower doses may be effective in these populations^{184, 271} and that rapid metabolizers may have a higher rate of failure to eradicate Helicobacter pylori^{176, 177, 187} and to heal esophagitis.²⁷² Subgroup analysis found no effect by race in 1 study of esomeprazole and lansoprazole in healing of erosive esophagitis.⁴ A small study (N=80) found no statistically significant difference at 8 weeks in rate of ulcer healing between rabeprazole 10 mg daily and omeprazole 20 mg daily among patients with differing CYP2C19 genotype.¹⁰⁸ The few adverse events were not analyzed by genotype. A trial of omeprazole in Japanese patients with recurrent esophagitis found no difference in efficacy or safety by genotype.²⁷³

Older patients also metabolize proton pump inhibitors more slowly, resulting in significantly higher drug levels and half-lifes. However, accumulation has not been shown, and dose adjustments are not recommended. One reanalysis of data from 2 trials comparing omeprazole with either ranitidine or cimetidine for reflux esophagitis compared effect in patients age 65 or older with those under age 65.²⁷⁴ In this analysis there was no difference in healing rate or symptom resolution at 4 or 8 weeks, with a slightly higher proportion of older patients both healed and symptom-free. Withdrawals due to adverse event were higher in the older group, 7.6% compared with 2.5%. Similar data are not available for other proton pump inhibitors.

Comorbidity

In an uncontrolled, non-randomized open-label study, patients with peptic ulcer and comorbid liver disease were given 6 to 8 weeks of rabeprazole 10 mg to 20 mg.²⁷⁵ Eleven of 108 patients (10%) reported 21 adverse drug events, resulting in 5 withdrawals (5%) and an additional 5 patients with an adverse event were lost to follow up. Two patients (2%) had adverse events that were rated as serious, 1 had an elevated bilirubin level, and the other had hepatic encephalopathy. Analysis by dose was not conducted.

Concomitant medications

Two good quality observational studies assessed the impact of a potential drug interaction between proton pump inhibitors and clopidogrel following an acute coronary syndrome (ACS). ^{276, 277} Clopidogrel is activated by a liver enzyme system known as P450 C19, and proton pump inhibitors can inhibit this system. A cohort study of 8205 patients who were discharged after an ACS and were prescribed clopidogrel between October 2003 and January 2006 were examined to determine if the rate of death or rehospitalization for ACS was affected by concomitant use of a proton pump inhibitor.²⁷⁶ Of these patients, 64% were prescribed a proton pump inhibitor. Multivariable analysis found that there was an increased risk of death or rehospitalization for ACS in those patients taking both clopidogrel and a proton pump inhibitor; adjusted odds ratio 1.25 (95% CI 1.11 to 1.41). The analysis controlled for multiple variables, included demographic characteristics, comorbidities, previous cardiac history, and other medications. In patients who had period with and without a proton pump inhibitor, but continued clopidogrel use, the risk of the primary outcome was also increased during the proton pump inhibitor periods; adjusted odds ratio 1.27 (95% CI 1.10 to 1.46). In addition to the primary outcome, secondary outcomes were evaluated. The risk of rehospitalization for ACS was increased (adjusted odds ratio 1.86 95% CI 1.57 to 2.20); the risk of a revascularization procedure was increased (adjusted odds ratio 1.49 95% CI 1.30 to 1.71); however risk of all-cause mortality was not significantly increased (adjusted odds ratio 0.91 95% CI 0.80 to 1.05). The authors also conducted a nested case-control analysis with these data in an attempt to confirm their findings, resulting in an adjusted odds ratio of 1.32 (95% CI 1.1.4 to 1.54) for the risk of death or rehospitalization for ACS. Multiple sensitivity analyses were conducted, with no meaningful change to the results. This study was conducted using data from the Veteran’s Affairs hospitals, and no patients were taking esomeprazole. Too few patients were taking pantoprazole or lansoprazole to be able to conduct individual analyses, but omeprazole and rabeprazole resulted in increased adjusted odds ratios for the primary outcome (adjusted odds ratios: 1.24 95% CI 1.08 to 1.41 for omeprazole, N=3132; 2.83 95% CI 1.96 to 4.04, N=151 for rabeprazole). Analysis by dose of proton pump inhibitor indicated did not indicate a dose-response relationship.

A population-based nested case-control study examined data from all patients in Ontario, Canada who were prescribed clopidogrel after hospital discharge following a myocardial infarction between April 2002 and December 2007.²⁷⁷ In this study, 13 636 patients were identified. Among this group, cases were identified as patients who were rehospitalized for myocardial infarction within 90 days of discharge (N=734), while controls were those who were not. Controls were identified in a 3:1 ratio to cases and matched on age, percutaneous coronary intervention, and a validated risk score (N=2057). Proton pump inhibitor exposure was defined as current (within 30 days of rehospitalization), previous (31 to 90 days) or remote (91 to 180 days). The logistic regression analysis controlled for demographic variables, socioeconomic status, Charlson comorbidity index, length of stay during initial admission for myocardial infarction, and 9 comorbid conditions (for example diabetes). A similar adjusted odds ratio was found in this study as the cohort study; 1.27 (95% CI 1.03 to 1.57) for current users of a proton pump inhibitor. Previous or remote use was not associated with an increased risk of recurrent myocardial infarction. All-cause mortality was again not affected statistically significantly (adjusted odds ratio 0.82 95% CI 0.57 to 1.18). Analysis of recurrent myocardial infarction within 1 year of initial discharge also indicated an increased risk with current proton pump inhibitor use; odds ratio 1.23 (95% CI 1.01 to 1.49). Because pantoprazole does not inhibit the P450 2C19 enzyme system responsible for activation of clopidogrel, it has been suggested that it may not result in a clinically-relevant drug interaction. An analysis of pantoprazole alone (N=cases 46, controls 125) found no statistically significant increase in risk (adjusted odds ratio 1.02 95% CI 0.70 to 1.47). Analysis of all other proton pump inhibitors (which inhibit the P450 2C19 enzyme system to varying degrees) together resulted in increased risk; adjusted odds ratio 1.40 (95% CI 1.10 to 1.77; N=cases 148, controls 299). Analysis stratified further by individual proton pump inhibitor was not undertaken; insufficient data may have prevented such analysis. Because these are post-hoc sub-group analyses of small groups, further research is needed to confirm these findings.

Pregnancy

A multicenter, prospective cohort study enrolled 410 pregnant women who had sought counseling after exposure to omeprazole (N=295), lansoprazole (N=62), or pantoprazole (N=53) between 1992 and 2001.²⁷⁸ Details of exposure were collected during pregnancy before pregnancy outcome was known, and follow-up was performed in the neonatal period. A control group of 868 women who had been counseled during pregnancy about exposures known to be nonteratogenic served as a control group. There were some differences between control and treatment groups at baseline (for example, number of children was larger in then treatment than the control group), and confounders were not controlled for in the analysis. There was a higher rate of elective termination of pregnancy in the omeprazole and lansoprazole groups than the control group. Two of these terminations in the omeprazole group, 1 in the lansoprazole group, 0 in the pantoprazole group, and 5 in the control group were because of prenatal diagnosis of anomalies. There was no difference in the rate of major anomaly between each of the 3 groups and the control group; the relative risk was 0.95 (95% CI 0.46 to 1.98) for omeprazole, 1.04 (95% CI 0.25 to 4.21) for lansoprazole, and 0.55 (95% CI 0.08 to 3.95) for pantoprazole. Median birth weight was lower by 60 grams in the omeprazole group than the control group, but no difference was seen between groups for median gestational age at delivery or rates of preterm birth, miscarriage, ectopic pregnancy, or stillbirth.

Applicability

Applicability of most trials to community practice was difficult to determine. These studies generally excluded patients who had serious medical conditions. In addition, although most treatment and control groups received standard doses of anti-ulcer drug, there were instances where doses were higher or lower than typical. In trials comparing maintenance treatment or different strategies for longer-term treatment of gastroesophageal reflux disease, patients were enrolled on the basis of a successful response to acute treatment. This preselection may have resulted in a group of patients who were adherent to treatment, who were able to tolerate any side effects, and whose disease was less severe in comparison with patients who were not enrolled. Another concern is that of studies that stated their funding source, most were funded by the pharmaceutical industry, and industry employees often served as co-authors.