Predictors of treatment response in patients with non-erosive reflux disease


Dr N. J. Talley, Professor of Medicine, Mayo Clinic College of Medicine, 200 First Street SW, PL-6-56, Rochester, MN 55905, USA.




Up to 40% of patients with non-erosive reflux disease (NERD) fail to respond to proton pump inhibitor therapy.


To determine useful prognostic factors for response to and suppression in NERD.


A pooled analysis from three multicentre, double-blind trials of patients with a normal endoscopy and heartburn for 4 days or more during the 7 days prior to the start of each treatment. Patients received omeprazole 20 mg, esomeprazole 20 mg or esomeprazole 40 mg/day for 4 weeks. Complete resolution of heartburn was defined as no heartburn during the last week.


Of 2458 patients included, complete heartburn resolution was achieved in 63% at the end of 4 weeks treatment. Response on days 5–7 provided an 85% probability of complete resolution of heartburn at 4 weeks; the probability of complete heartburn resolution at 4 weeks in those with moderate to severe symptoms on days 5–7 was 22%. Sensitivity and specificity of no heartburn on days 5–7 was 55% and 83% respectively. Patient demographics, duration of symptoms, medications used, other symptoms and body mass index were not predictors.


Assessment of heartburn resolution during the first week of therapy was the best predictor of treatment success at 4 weeks in non-erosive reflux disease, but was suboptimal as a test.


Symptoms of gastro-oesophageal reflux disease are remarkably common, and the majority of patients with these symptoms have non-erosive reflux disease (NERD) when investigated.1, 2 Non-erosive reflux disease can be defined as the presence of typical symptoms of gastro-oesophageal reflux caused by intraoesophageal acid exposure in the absence of visible endoscopic mucosal injury.3 The hallmark symptom of NERD is heartburn although other symptoms may also be present.3

Systematic review data have suggested that patients with NERD respond less well to acid suppression therapy than patients with documented oesophagitis.4 A subsequent clinical trial programme that compared esomeprazole with omeprazole reported the response rate in NERD, defined as complete resolution of heartburn, occurred in 57–70% of the patients after 4 weeks of treatment.5 This indicates that there is a considerable group of patients, approximately 40%, who may not have true acid-induced reflux symptoms in this condition.

Proton pump inhibitors (PPI) are widely prescribed for patients with symptoms of gastro-oesophageal reflux.6 It would be clinically useful and potentially cost-effective to be able to predict who would be highly likely to respond, and not respond, to treatment with acid suppression in NERD, and thereby avoid prolonged use of medications that will not ultimately be beneficial. However, there remains a lack of published data on predictors of treatment response in patients with NERD.

We aimed to identify predictors of treatment outcomes in patients with documented NERD following 4 weeks treatment with either omeprazole or esomeprazole in a large randomized-clinical trial programme comprising three separate studies of very similar design.5


Multicentre randomized-controlled trials

The analysis was performed on pooled data from three multicentre, double-blind, randomized-clinical trials; the main results have been reported by us elsewhere.5

The studies (trials A, B and C) were conducted to compare the effect of once daily oral dosing of esomeprazole 40 mg, esomeprazole 20 mg or omeprazole 20 mg in symptomatic patients with NERD. All medication was taken once daily in the morning. Each study was conducted in accordance with the Declaration of Helsinki and was consistent with Good Clinical Practice. Signed informed consent was obtained from all patients before conducting any specific procedures in all of the studies.

The studies were of identical design and each provided a 4-week treatment period.5 Patients in study A received either esomeprazole 40 mg, esomeprazole 20 mg or omeprazole 20 mg. Patients in study B received esomeprazole 40 mg or omeprazole 20 mg and patients in study C received esomeprazole 20 mg or omeprazole 20 mg.

Included were all adult patients who had experienced heartburn (defined as a burning feeling, rising from the stomach or lower part of the chest up towards the neck) as their main symptom for 6 months or longer, and for 4 days or more during the last week before the start of each study, and who had a normal endoscopy, defined as no oesophageal mucosal breaks or other oesophageal pathology and no more than five gastric or duodenal erosions. Treatment began on day 0 (visit 1), with a baseline endoscopy within 14 days prior to visit 1, and ended on day 28 ± 4 (visit 3). Prior acid suppression treatment was not an exclusion criterion.

Testing for the presence of Helicobacter pylori with a 13C-urea breath test (UBT) was performed in all patients before the start of treatment; patients and investigators remained unaware of the results of the UBT until after completion of the study.

Outcome assessments

Complete resolution of heartburn was defined as no heartburn during the week prior to a clinic visit and was assessed by the blinded investigator in each of the trials. At each visit, the frequency and severity of heartburn were assessed by the investigator. Frequency was recorded as the number of days with heartburn during the last 7 days (scale of 0–7 days). Symptom severity was classified, on a validated7 4-point scale, as none, mild (awareness of symptoms, but easily tolerated), moderate (discomfort sufficient to cause interference with normal activities), or severe (incapacitating with inability to perform normal activities).

Patients also recorded their daily symptoms of heartburn on diary cards during the 4 weeks of treatment. The symptom severity was classified on the same 4-point scale as the one used for the investigator assessment.

The severity of other reflux symptoms (regurgitation, dysphagia) and gastrointestinal symptoms (epigastric pain, nausea) during the last 7 days was assessed by standard questioning by the investigator and classified in the same way as heartburn.

Statistical analysis

The primary efficacy end point was complete resolution of heartburn, assessed retrospectively by the investigator, and defined as no days with heartburn episodes during the last 7 days before visit 3. A priori, a number of potential explanatory variables were extracted from the available clinical trial data. These included trial (A, B or C), treatment, patient demographics [gender, age, employed (yes/no)], duration of reflux disease (<5/5 years or more), other medication used prior to enrolment, other gastrointestinal symptoms [epigastric pain, regurgitation, nausea, dysphagia, vomiting (all yes/no)], body mass index, endoscopic findings [hiatus hernia (yes/no), erosions (yes/no)], H. pylori status (positive/negative), and heartburn severity assessed on daily diary cards during the first 2–3 days and the first 5–7 days of therapy and classified as 0–3 [0 = none, 1 = at most mild, 2 = at most moderate or severe some day(s), 3 = moderate or severe all days].

Logistic regression analysis was used to identify predictors of response at 4 weeks (complete resolution of heartburn as assessed by the investigator). A forward elimination process was also carried out using a conservative significance level of 0.01. Sensitivity, specificity, positive predictive value and negative predictive value of the predictors were calculated and a receiver operating characteristic (ROC) curve was constructed.


In all, 2458 of the 2645 patients who were randomized and included in an intention-to-treat (ITT) analysis, had data on all variables for inclusion in the logistic regression analysis. For the separate analyses of response on days 2–3 and on days 5–7, 2530 and 2487 patients, respectively, were included. The mean age was 49 years and the female to male ratio was 1.2:1, with females slightly outnumbering males (Table 1). The majority of patients had moderate heartburn severity (62%); over half of the patients had daily symptoms at entry (59%).

Table 1.   Demographic characteristics (n = 2645)
 n (%)
Age (years)
 0–30294 (11)
 31–45822 (31)
 46–60899 (34)
 >60630 (24)
 Male1206 (46)
 Female1439 (54)
Heartburn severity
 Unknown1 (0)
 Mild471 (18)
 Moderate1656 (63)
 Severe517 (20)
Frequency of heartburn (days)
 <42 (0)
 4468 (18)
 5346 (13)
 6265 (10)
 71564 (59)
Duration of heartburn (years)
 <1397 (15)
 1–5981 (37)
 >51267 (48)
Other GI symptoms
 Regurgitation1748 (66)
 Dysphagia544 (21)
 Epigastric pain1557 (59)
 Nausea914 (35)
Other medication use prior to entry
 Depression (ATC codes N05A/B, N06A)212 (8)
 Asthma (ATC codes R03)175 (7)
Body mass index
 Overweight (25–30)1116 (42)
 Obese (>30)546 (21)
Helicobacter pylori-positive855 (32)

The primary end point, complete heartburn resolution based on investigator assessment, was achieved in 63% of the pooled patient population at the end of 4 weeks treatment with either omeprazole or esomeprazole. There were no significant differences in terms of complete response of heartburn to the PPIs compared with any of the three trials included in this study.5 Among the 2216 patients with complete diary cards for the last 7 days of therapy, complete resolution of heartburn was reported by 65% according to the investigator and by 58% according to the diary cards.

Logistic regression analysis identified heartburn response during days 5–7 of treatment as the most discriminating predictor of treatment outcome (P < 0.0001). Figure 1 depicts the percentage of patients with a complete resolution of heartburn after 4 weeks based on their symptom response during the first week of treatment. On days 5–7, 1016 of the 2487 patients (41%) with evaluable data reported no symptoms of heartburn, and of these, 85% continued to be symptom-free at 4 weeks. The only other significant predictive factors identified in the model were study (one of the three studies had a significantly higher 4-week response rate) and being heartburn-free during days 2–3 of treatment (Figure 2; P < 0.0001). None of the other explanatory variables investigated reached statistical significance at the 0.01 level in the full logistic regression, and none of them was selected by the forward selection procedure.

Figure 1.

 Complete resolution of heartburn (%) after 4 weeks by symptom response during days 5–7 of the first week of treatment.

Figure 2.

 Complete resolution of heartburn (%) after 4 weeks by symptom response during days 2–3 of the first week of treatment.

The sensitivity, specificity, predictive values and likelihood ratios of heartburn response in the first week (days 5–7) as a predictor of complete symptom resolution are presented in Table 2 for three possible tests. The area under the ROC curve (0.73) demonstrated that heartburn response in the first week was not an adequate diagnostic test in this clinical trial setting. Similarly, the sensitivity and specificity, and area under the curve of the heartburn response ‘no heartburn at days 2–3’ was suboptimal (36%, 88% and 0.68, respectively).

Table 2.   Diagnostic parameters for the 1 week test (no heartburn on days 5–7 as classified in the table) predicting complete resolution at 4 weeks
Positive test defined as:SensitivitySpecificityPositive predictive valueNegative predictive valuePositive likelihood ratioNegative likelihood ratio
  1. Values are expressed as estimates (95% confidence intervals).

none0.55 (0.52–0.58)0.83 (0.81–0.85)0.85 (0.83–0.87)0.52 (0.49–0.55)3.30.54
at most mild0.84 (0.82–0.86)0.43 (0.40–0.46)0.71 (0.69–0.73)0.61 (0.57–0.65)1.50.38
at most moderate– severe some days0.98 (0.97–0.99)0.12 (0.10–0.14)0.65 (0.63–0.67)0.78 (0.71–0.85)1.10.16


In this study, we have shown that, in patients with endoscopically confirmed NERD, an assessment of heartburn response, defined as no heartburn symptoms from day 5 to day 7 of therapy with omeprazole or esomeprazole was the best predictor of treatment success at 4 weeks amongst the items assessed. However, heartburn relief for any period during the first week of acid suppression therapy remained inadequate to make a decision on whether to continue or cease therapy after 1 week.

Non-erosive reflux disease remains a potentially difficult management problem because it is likely a heterogeneous condition. Pathophysiological differences between patients with NERD and functional heartburn have been reported in limited studies. For example, Frazzoni et al. showed that patients with NERD were more likely to have a hiatal hernia, a lower mean oesophageal sphincter tone and greater upright diurnal acid reflux.8 They suggested that functional heartburn patients would therefore be unlikely to respond to acid suppression therapy in comparison with true NERD patients. However, the exact proportion of patients with non-erosive disease who have functional heartburn, and whether this defines non-response, remains to be well established. Based on the pooled data from the current trials, approximately 40% of patients failed to have complete heartburn resolution on PPI therapy at 4 weeks.5 Furthermore, a higher dose of PPI therapy (namely esomeprazole 40 mg) did not appear to provide additional benefit in terms of symptom response despite the fact that better acid suppression should be obtained with such dosing.5 Underlying pathophysiological disturbances causing heartburn, such as oesophageal hypersensitivity may account for non-response in part although a full explanation is currently lacking.8 The present study confirms, however, that a short course of acid suppression therapy (1 week) has inadequate predictive value for defining any benefit or lack of benefit at the end of 4 weeks of treatment. Very few other data are available on predictive factors for NERD. In an out-patient study in Germany, Labenz et al. reported that they could not adequately distinguish erosive oesophagitis from NERD, but they did not analyse predictors of treatment response in the study.9

In reflux oesophagitis, it needs to be noted that in clinical trial programmes, complete resolution of heartburn occurs in only 40–68% by 4 weeks;10–12 thus, failure to respond to acid suppression therapy at 4 weeks in patients with reflux symptoms and a normal endoscopy does not exclude a diagnosis of reflux-related disease. Whether a longer course of acid suppression therapy in NERD would yield higher rates of symptom relief and better predictive value of a 1-week therapeutic course remains unclear.

The present pooled analysis has a number of strengths. This represents some of the first research looking at predictors of treatment response in NERD. The pooled studies were nearly identical in design and this, therefore, minimizes the potential for heterogeneity that may confound meta-analyses of studies using disparate study designs and outcome measures. The analysis included a large number of patients, comparable, for example, with the total number of patients included in a meta-analysis of PPI diagnostic test characteristics;13 as a result, the present analysis also had adequate precision based on the narrow confidence limits obtained (Table 2).

There are also potential limitations. Investigator assessment of symptom resolution, although blinded and validated, may be less appropriate than a patient-reported outcome approach. Recall bias for symptoms over a week may also be a concern, but because they had kept a daily diary, there is a reasonable chance that they would recall if they had experienced symptoms over the previous 7 days. As previous acid suppression therapy was not an exclusion criterion, misclassification of NERD is also possible although any error should have been non-differential with respect to different treatment groups. The trials evaluated were not designed to specifically assess prognostic features and therefore the number of potential explanatory variables was relatively limited, although those available seem to represent a reasonable group of factors in terms of face validity. Another potential weakness of these types of studies is that modelling can potentially exaggerate any benefit of predictors, which then fail to hold up in subsequent studies. Hence, the predictive value of the variables identified may be biased in the positive direction. However, in this analysis, based on a large number of patients, very few predictors were identified, so the process has not in this case identified spurious predictors whose relevance has to be confirmed in subsequent studies. Furthermore, our results are concordant with a meta-analysis of the PPI acid suppression test for diagnosing gastro-oesophageal reflux disease; compared with a structured questionnaire as the ‘gold standard’ comparator, the sensitivity and specificity of high-dose PPI therapy for 1–2 weeks ranged very widely, from 0.38 to 0.90 and 0.09 to 0.64, respectively.13

We found as expected that a less robust predictor of outcome, albeit significant, was heartburn severity at 2–3 days on acid suppression therapy. McQuaid and Laine have reported in a pooled analysis that, while PPIs can provide benefit early on, two-thirds of patients fail to obtain adequate symptom relief within 2 days.14 Hence, it would not be expected that a very early response to acid suppression would be a robust predictor of subsequent success. We did not choose to look at the first day of therapy as a predictive response based on this analysis but rather at days 2–3;14 the sensitivity and specificity of this variable were lower than those of the variable acid suppression at 5–7 days.

In conclusion, patients with NERD who had complete resolution of their heartburn by days 5–7 of use had an 85% probability of complete resolution of heartburn after 4 weeks. In contrast, those with persistent heartburn on days 5–7 of acid suppression therapy still had a 22% probability of complete resolution of heartburn after 4 weeks. Further study is needed to identify accurate predictors of treatment response in this condition, in order to more appropriately target acid suppression therapy and to determine whether a high-dose PPI acid suppression test has a role for differentiating between ‘true’ NERD and functional heartburn in patients who have typical reflux symptoms and a normal upper endoscopy.


No external funding was received for this study.