Population-based deprescribing strategy for proton pump inhibitors: Health outcomes from a case-control study

Abstract

Introduction

In 2017, 2018, and 2022, a population-based strategy was implemented to deprescribe proton pump inhibitors (PPIs) in patients who did not meet clinical indications for treatment. This study evaluates health outcomes related to the safety of deprescribing and the adverse effects associated with PPI consumption in these patients.

Patients and methods

A nested case-control study within a cohort was conducted, with follow-up through December 31, 2024. The primary endpoint was a composite of mortality or urgent hospital admission. Secondary endpoints included upper gastrointestinal bleeding, other gastrointestinal events (ulcers, gastritis, or esophagitis), hypomagnesemia, cyanocobalamin deficiency, hypocalcemia, iron deficiency, pneumonia, Clostridioides difficile infection, and bone fractures. Adjustments were made for sex, age, Charlson Comorbidity Index, prior urgent hospital admissions, number of primary care visits in the preceding year, and number of prescribed medications. Additionally, a process mining analysis was performed to describe patient trajectories.

Results

A total of 28,161 patients were included. PPI deprescribing proved to be safe regarding mortality and hospitalization (OR = 0.78; 95% CI: 0.70–0.88). A higher incidence of gastric adverse events was observed, but not of gastrointestinal bleeding. Furthermore, a lower incidence of micronutrient deficiencies, fractures, and pneumonia was found among patients who underwent PPI deprescribing. In the sex-stratified analysis, deprescribing was associated with fewer admissions or hospitalizations in both groups.

Conclusions

The PPI deprescribing strategy in patients without risk factors for gastrointestinal bleeding seems to be safe in terms of mortality and hospital admissions.

Resumen

Antecedentes y objetivo

En 2017, 2018 y 2022 se implementó una estrategia poblacional para deprescribir inhibidores de la bomba de protones (IBP) en pacientes que no cumplían indicación de tratamiento. Este trabajo evalúa los resultados en salud relacionados con la seguridad de la deprescripción y con los efectos adversos de tomar IBP en esos pacientes.

Pacientes y métodos

Estudio de casos y controles anidado en una cohorte, con seguimiento hasta el 31/12/2024. La variable principal fue la combinada de muerte o ingreso hospitalario urgente. Las variables secundarias: hemorragias digestivas altas, otros eventos digestivos (úlceras, gastritis o esofagitis), hipomagnesemia, hipovitaminosis B12, hipocalcemia, déficit de hierro, neumonía, infección por Clostridioides difficile y fracturas óseas. Se ajustó por sexo, edad, índice de Charlson, ingresos hospitalarios urgentes previos, número de visitas en atención primaria en el año anterior y número de medicamentos prescritos. Además, se realizó un análisis de minería de procesos para describir las trayectorias de los pacientes.

Resultados

Se incluyó a 28161 pacientes. La deprescripción del IBP mostró ser segura en cuanto a fallecimiento y hospitalización (OR=0,78; IC95%: 0,70-0,88). Se observó una mayor incidencia de eventos adversos gástricos, pero no de hemorragias digestivas, y una menor incidencia de déficit de micronutrientes, fracturas y neumonías en los pacientes a quienes se deprescribió el IBP.

Al realizar el análisis por sexo, la deprescripción se asoció en ambos grupos con un menor número de ingresos u hospitalizaciones.

Conclusiones

La estrategia de deprescripción de IBP en pacientes sin factores de riesgo de hemorragia digestiva parece ser segura en términos de mortalidad e ingresos hospitalarios.

Introduction

Proton pump inhibitors (PPIs) are drugs that irreversibly inhibit the H⁺/K⁺-ATPase enzyme of the parietal cells of the gastric mucosa. In this way, they reduce gastric acid secretion.¹

Their currently approved indications include the treatment of gastro-oesophageal reflux disease, eradication therapy for Helicobacter pylori, prevention of non-steroidal anti-inflammatory drug-induced gastropathy, treatment of Zollinger–Ellison syndrome, and the treatment and prevention of gastroduodenal ulcers.¹ In addition, they are recommended for other off-label indications such as Barrett’s oesophagus or eosinophilic oesophagitis. PPIs are undoubtedly effective for these indications, with a favourable risk–benefit balance, particularly when used for short periods.

At present, the high use of PPIs in the population is further increased by their prescription for non-approved indications.^2,3 For this reason, the American Gastroenterological Association (AGA) and other organisations recommend periodically reassessing the indication for PPIs and considering their deprescribing in cases where the benefit to the patient is unclear.^4,5 In addition, the safety profile of long-term treatment appears less favourable, having been associated with adverse effects such as deficiencies of certain micronutrients (magnesium, vitamin B₁₂, calcium, iron), bone fractures, pneumonia, or enteric infections, among others.^6–8

In the case of Navarra, and with this objective in mind, it was confirmed that during 2016, prior to the start of the study, one in five individuals received at least one PPI at some point, making this drug group the second most commonly prescribed in terms of number of patients treated (billing database of the Servicio Navarro de Salud-Osasunbidea, 2016). Therefore, a population-based strategy was designed to deprescribe chronic PPIs in patients who did not meet criteria for treatment indication.

Following implementation of the strategy, we considered it essential to evaluate its health outcomes. Our hypothesis was that withdrawal of PPI treatment in patients without an indication for these drugs is safe. For this reason, this retrospective observational study was conducted with the aim of evaluating health outcomes related to the safety of deprescribing and the adverse effects of PPI use in patients who did not meet treatment criteria.

Patients and methods

Study design

An analytical retrospective observational case–control study nested within a population-based cohort was conducted.

Study population and criteria

The cohort comprised all individuals included in the deprescribing strategy implemented at three different time points: 2017, 2018 and 2022. In 2017, the target population was restricted to individuals over 65 years of age, whereas in 2018 and 2022, all adult patients meeting the criteria were included, regardless of age. The cohort was followed up until 31 December 2024.

Inclusion criteria were: adult patients entitled to healthcare and pharmaceutical provision within the Servicio Navarro de Salud-Osasunbidea, receiving treatment with PPIs (omeprazole, esomeprazole, lansoprazole, pantoprazole, rabeprazole), with no recorded diagnosis in their clinical history of dyspepsia, upper gastrointestinal bleeding, gastro-oesophageal reflux disease, gastroduodenal ulcer, Barrett’s disease, Zollinger–Ellison syndrome or H. pylori eradication, and not receiving treatment with drugs that increase the risk of upper gastrointestinal bleeding, such as non-steroidal anti-inflammatory drugs, anticoagulants, antiplatelet agents or corticosteroids in combination with venlafaxine or selective serotonin reuptake inhibitors.

Exclusion criteria were residence in private nursing homes with in-house physicians, due to the unavailability of information on pharmacological treatment, and the absence of any recorded Charlson Comorbidity Index value.

Strategy

The strategy consisted of sending discontinuation proposals to each patient’s medication record. Implementation was possible through the use of SAPE and OBSERVA, two electronic tools. The former allows identification of the target population, and the latter enables the sending of discontinuation proposals to each patient meeting the criteria. These proposals were accompanied by a message explaining the rationale and were reviewed by medical professionals, who decided whether to discontinue or maintain PPI treatment. This strategy was complemented by patient information leaflets explaining when PPI treatment is or is not necessary, the potential adverse effects, and how to withdraw treatment.⁹

Definitions

• •

  Deprescribing: any discontinuation of the prescription in the electronic prescribing system lasting more than one month, regardless of the time elapsed since the proposal was issued. To establish this one-month threshold, a sensitivity analysis between 30 and 60 days was performed. The results were similar, and since rebound acid hypersecretion has been described from 1 to 2 weeks after discontinuation,^10,11 a one-month threshold was established for the definition of deprescribing.

• •

  Cohort entry date: for all patients, both in the deprescribed group and in the group that maintained the prescription, this date corresponded to the date of the PPI discontinuation proposal.

Data sources

All data required for the study were obtained from the databases of the Servicio Navarro de Salud-Osasunbidea and, specifically, from BARDENA (Base de Análisis de Resultados de Navarra), which integrates data from different SNS-O sources¹² and enables homogeneous data collection from clinical records across the entire population.

Study variables

• •

  Exposure variable: PPI deprescribing (yes/no).

• •

  Primary outcome variable: emergency hospital admission or death from any cause.

• •

  Secondary outcome variables: upper gastrointestinal bleeding, other gastrointestinal events (ulcers, gastritis or oesophagitis), pneumonia, Clostridioides difficile infection, and bone fractures were identified from diagnoses recorded in primary care clinical records, hospital records, and hospital discharge coding (CMBD). Deficiencies of different micronutrients were established when laboratory values recorded in the clinical history were below the reference range defined by the analysing laboratory.

Baseline demographic and clinical variables

To describe baseline population characteristics, the following variables were collected: sex, age, Charlson index (categorised as 0, 1, 2, > 2), previous emergency hospital admissions (yes/no), number of primary care visits in the previous year (categorised as < 3, 3–7, 8–12, > 12), and number of prescribed medications (categorised as 0, 1–3, 4–5, > 5).

Statistical analysis

• 1

  Descriptive analysis:

A descriptive analysis of baseline population characteristics was performed. Categorical variables were described using frequencies and percentages, and patients in whom PPIs were deprescribed were compared with those who maintained treatment using the Chi-square test (χ²). Quantitative variables were described using mean and standard deviation. Comparisons between groups were performed using Student’s t-test. Statistical significance was set at p < 0.05.

To visualise and analyse patient pathways and transition times within the study cohort, a process mining analysis was conducted. This analysis was performed using the bupaR and DiagrammeR packages in the R environment.¹³ Based on cohort event data, an event log was constructed for each patient, along with transitions between states (such as “Cohort entry”, “Deprescribing”, “Prescription”, “Death”) with corresponding timestamps. A process flow diagram was generated to visualise aggregated patient trajectories, highlighting the most frequent transitions within the process.

• 2

  Association analysis (nested case–control study):

To analyse the association between deprescribing and each event, several case–control studies were nested within the cohort, one for each outcome variable. For each outcome (primary and secondary):

• •

  All cases (patients who experienced the event) were identified.

• •

  For each case, one control was selected from the cohort who had not experienced the event prior to the case index date. Control selection was performed using incidence density sampling (density sampling) with the ccwc function from the Epi package in R¹³.

Matching criteria between cases and controls included: sex, age, year of intervention (2017, 2018, 2022), Charlson index categories, number of primary care visits, and number of prescribed medications. For the primary outcome (admission or death), and for the variables “admission” and “death”, an additional matching criterion was included: the presence or absence of an emergency admission in the previous year.

The association between deprescribing (exposure) and each event was quantified by calculating odds ratios (ORs) with their 95% confidence intervals (95% CIs). Given the matched design, the analysis was performed using conditional logistic regression. No additional adjustments beyond matching were made in the models.

All analyses were conducted both for the overall population and stratified by sex, as specified in the protocol.¹⁴

Results

A total of 28,161 patients were included across the three implementation periods of the deprescribing strategy (8577 patients in 2017, 7237 in 2018 and 12,347 in 2022). After excluding those without Charlson index data, the final sample size was 27,826 patients, representing more than 4% of the total population of Navarra (Fig. 1).

Figure 1.

Flow diagram of patient inclusion in the study.

Table 1 compares the general characteristics of the population, distinguishing between those in whom PPIs were deprescribed at any time for at least one month and those who continued treatment. In addition, the total population is broken down into the three groups of patients to whom the deprescribing proposal was sent at the three time points (2017, 2018 and 2022). In 2018 and 2022, patients in whom PPIs were deprescribed were younger and had a lower Charlson index than those who continued PPI treatment. The proportion of men and women among deprescribed and treated patients did not differ across the implementation periods.

Table 1.

Description of the total population included in the study, as well as the populations at each of the three implementation time points of the strategy, according to whether or not the proton pump inhibitor was deprescribed.

		Total			2017			2018			2022
		Not deprescribed	Deprescribed	p	Not deprescribed	Deprescribed	p	Not deprescribed	Deprescribed	p	Not deprescribed	Deprescribed	p
		n = 16,576	n = 11,250		n = 5120	n = 3432		n = 3850	n = 3307		n = 7600	n = 4517
Age	Mean (SD)	72.0 (14.6)	66.7 (17.6)	<0.001	83.2 (7.4)	83.2 (7.5)	0.740	63.8 (12.1)	59.2 (14.0)	<0.001	68.5 (14.8)	59.5 (17.0)	<0.001
Sex	Female	9987 (60.1%)	6892 (61.0%)	0.081	3355 (65.5%)	2351 (68.5%)	0.005	2146 (55.6%)	1882 (57.1%)	0.209	4486 (59.0%)	2659 (58.9%)	0.878
	Male	6589 (39.8%)	4358 (38.7%)		1765 (34.5%)	1081 (31.5%)		1710 (44.4%)	1419 (42.9%)		3114 (41.0%)	1858 (41.1%)
Number of drugs	Mean (SD)	3.7 (3.1)	3.8 (2.7)	0.010	5.1 (3.3)	5.1 (2.8)	0.831	3.7 (2.9)	3.6 (2.4)	0.093	2.7 (2.7)	2.9 (2.3)	<0.001
Admission previous year	No	15,086 (91.0%)	10,368 (92.1%)	0.002	4628 (90.4%)	3145 (91.6%)	0.054	3485 (90.5%)	3044 (92.0%)	0.0525	6973 (91.8%)	4179 (92.5%)	0.141
	Yes	1484 (9.0%)	888 (7.9%)		492 (9.6%)	287 (8.4%)		365 (9.5%)	263 (8.0%)		627 (8.2%)	338 (7.5%)
Primary care visits	Average (SD)	9.1 (8.1)	8.7 (7.9)	<0.001	9.4 (8.1)	9.0 (7.6)	0.036	9.2 (8.2)	8.6 (8.0)	<0.001	8.7 (8.0)	8.6 (8.0)	0.327
Charlson score	Mean (SD)	1.8 (2.2)	1.3 (1.9)	<0.001	2.0 (2.3)	1.8 (2.1)	<0.001	1.6 (2.2)	1.1 (1.8)	<0.001	1.6 (2.2)	1.1 (1.8)	<0.001

Fig. 2 presents the analysis performed using process mining. It shows the trajectories followed by patients from the start of follow-up to PPI deprescribing and to death, also displaying the reintroduction of PPI treatment where applicable. In all cases, a PPI was considered discontinued if at least one month had elapsed since prescription cessation. Of the 27,826 patients analysed, 40.4% experienced PPI deprescribing at some point during follow-up, and in 66% of these cases no drug from this class was prescribed again. Among patients in whom treatment was deprescribed, the median (range) time from the proposal to discontinue treatment to actual cessation was 8.57 months (0.03–89.07). Of these, 6094 patients were re-prescribed a PPI at some point, with a median time to reintroduction of 4.60 months (0.03–84.52); among these, treatment was again discontinued in 67.6% of cases, with a median time of 1.93 months (1.00–84.29).

Figure 2.

Flow diagram of patient trajectories from study initiation to the end of follow-up.

In total, 7742 individuals died or were hospitalised, accounting for 28% of the cohort. During the study period, almost 10% of patients died. The incidence of secondary outcomes ranged from 0.5% for upper gastrointestinal bleeding and C. difficile infections to higher incidences for cyanocobalamin or calcium deficiency (above 10%) and iron deficiency (above 20%) (Table 2).

Table 2.

Results of the analysed variables.

	Deprescribed (n = 11,256)	Not deprescribed (n = 16,570)	OR	95% CI
Death or hospitalisation	1570	6172	0.78	0.70–0.88
Death	563	1678	0.97	0.80–1.19
Hospitalisation	1351	5495	0.74	0.65–0.84
Upper gastrointestinal bleeding	22	62	0.60	0.22–1.65
Other gastrointestinal adverse effects^a	583	1673	1.62	1.33–1.97
Fractures	302	986	0.69	0.53–0.90
Pneumonia	513	1833	0.75	0.61–0.92
C. difficile infection	20	76	0.67	0.27–1.63
Cyanocobalamin deficiency	850	2652	0.73	0.62–0.87
Magnesium deficiency	106	484	0.38	0.23–0.61
Calcium deficiency	798	3016	0.64	0.54–0.75
Iron deficiency	1329	6191	0.68	0.60–0.78

CI: confidence interval; OR: odds ratio adjusted for age, sex, Charlson index, year of intervention (2017, 2018, 2022), number of primary care visits and number of prescribed medications. For the variables “death or hospitalisation”, “death” and “hospitalisation”, the presence or absence of emergency admission in the previous year was also included.

a

Ulcers, gastritis or oesophagitis.

PPI deprescribing showed a protective effect for the primary combined outcome of death and hospitalisation (OR = 0.78; 95% CI = 0.70–0.88). When analysed separately, the effect remained significant for hospitalisation (OR = 0.74; 95% CI = 0.65–0.84), with no differences observed in mortality (OR = 0.97; 95% CI = 0.80–1.19). Regarding potential adverse effects associated with PPI use, a lower risk of micronutrient deficiencies (magnesium, calcium, iron and vitamin B₁₂), as well as fractures and pneumonia, was observed in patients in whom PPIs were deprescribed. A total of 1288 patients experienced at least one clinically relevant bone fracture and 2346 patients had at least one episode of pneumonia. No differences were found in the incidence of C. difficile infection (Table 2).

The incidence of upper gastrointestinal bleeding was similar between patients in whom PPIs were deprescribed and those who continued treatment (OR = 0.60; 95% CI = 0.22–1.65). However, patients undergoing deprescribing showed a higher incidence of gastrointestinal adverse effects (ulcers, gastritis, oesophagitis) compared with those who maintained PPI treatment (OR = 1.62; 95% CI = 1.33–1.97). When these events were analysed separately, only the increase in oesophageal disorders reached statistical significance [n = 2166; OR = 1.467 (1.207–1.784); p < 0.001] in patients in whom PPIs were deprescribed compared with those who maintained treatment.

In the sex-stratified analysis, a lower proportion of men was observed. At cohort entry, women had, on average, a higher number of prescribed medications (3.9 vs 3.3; p < 0.001) and more primary care visits in the previous year than men (9.5 vs 7.9; p < 0.001), although a lower proportion of women had been hospitalised in the previous year (Table 3).

Table 3.

Description of baseline characteristics stratified by sex.

	Female	Male	p
Age, mean (SD)	71.7 (16.3)	66.7 (15.4)	<0.001
Hospital admission previous year, n (%)	1328 (7.8)	1046 (9.4)	<0.001
Prescribed medications, mean (SD)	3.9 (3.0)	3.3 (2.8)	<0.001
Primary care visits previous year, mean (SD)	9.5 (8.1)	7.9 (7.7)	<0.001
Charlson score, mean (SD)	1.5 (2.1)	1.6 (2.2)	<0.001

SD: standard deviation.

In both groups, deprescribing was associated with a lower number of hospital admissions. For secondary outcomes, results were similar in men and women, except for pneumonia and fractures: in women, deprescribing was associated with fewer fractures and pneumonias, whereas this association was not observed in men (Table 4).

Table 4.

Effect of PPI deprescribing stratified by sex.

	Female				Male
	Deprescribed (n = 6892)	Not deprescribed (n = 9987)	OR (95% CI)	p	Deprescribed (n = 4358)	Not deprescribed (n = 6589)	OR (95% CI)	p
Death or hospitalisation	1057	4137	0.84 (0.73–0.97)	0.014	513	2035	0.68 (0.55–0.83)	<0.001
Death	418	1185	0.93 (0.75–1.14)	0.479	143	495	0.68 (0.48–0.97)	0.031
Hospitalisation	887	3671	0.76 (0.66–0.88)	<0.001	439	1849	0.76 (0.62–0.93)	0.007
Upper gastrointestinal bleeding	9	37	0.67 (0.11–3.99)	0.657	11	27	0.80 (0.22–2.98)	0.739
Other gastrointestinal events^a	411	1103	1.55 (1.23–1.96)	<0.001	194	548	1.17 (0.83–1.66)	0.374
Fractures	253	857	0.67 (0.50–0.89)	0.006	42	136	1.31 (0.64–2.69)	0.467
Pneumonia	342	1220	0.68 (0.53–0.87)	0.002	166	618	1.00 (0.69–1.45)	1.000
C. difficile infection	11	59	1.2 (0.37–3.93)	0.763	8	21	0.50 (0.05–5.51)	0.571
Vitamin B₁₂ deficiency	558	1710	0.80 (0.66–0.98)	0.028	280	954	0.72 (0.54–0.95)	0.022
Hypomagnesaemia	67	285	0.31 (0.17–0.58)	<0.001	40	198	0.47 (0.21–1.05)	0.065
Hypocalcaemia	528	1880	0.67 (0.55–0.82)	<0.001	264	1142	0.63 (0.47–0.84)	0.002
Iron deficiency	1043	4571	0.632 (0.55–0.73)	<0.001	324	1582	0.67 (0.51–0.87)	0.003

CI: confidence interval; OR: odds ratio adjusted for age, sex, Charlson index, year of intervention (2017, 2018, 2022), number of primary care visits and number of prescribed medications; PPI: proton pump inhibitor. For the variables “death or hospitalisation”, “death” and “hospitalisation”, the presence or absence of emergency admission in the previous year was also included.

a

Ulcers, gastritis or oesophagitis.

Discussion

Deprescribing can be defined as the systematic process by which medications whose potential risks outweigh their potential benefits are identified and withdrawn in order to achieve therapeutic goals in an individual. Accordingly, PPI deprescribing is an increasingly widespread clinical strategy, particularly in patients in whom the indication for treatment is unclear in terms of the benefit–risk balance.

In the case of proton pump inhibitors (PPIs), the prevalence of theoretically unjustified use is high^2,3; in 2016, it accounted for 4% of the population of Navarre, which is consistent with the large size of our cohort. The advanced mean age of the cohort, together with multimorbidity, accurately reflects our population structure, which is increasingly characterised by ageing and polypharmacy.

Process mining analysis shows that a proportion of patients restart treatment after deprescribing, which may be related to the development of gastrointestinal symptoms; however, this association cannot be established with certainty given the observational design. In any case, this finding was not associated with an increase in hospital admissions. Similarly, the analysis shows that, in a subset of these patients, PPIs are deprescribed again, which may reflect intermittent rather than chronic use, a more appropriate approach in patients without a clear indication.

The increase in oesophageal adverse effects observed in our study is consistent with rebound acid hypersecretion, a recognised effect that may occur after PPI deprescribing,^4,15 and this was the reason why our discontinuation proposal recommended gradual withdrawal. The impact of symptom recurrence on patients’ quality of life is an important aspect that was not evaluated and should be considered in future studies.

The deprescribing rate observed in this study (41%) is similar to that reported in others, such as the study by Ayoub et al., which achieved a 42% deprescribing rate in an outpatient prospective study. Nallapeta et al., using a more individualised intervention consisting of weekly sessions for physicians and patient education, achieved a 51% deprescribing rate.¹⁵ In the hospital setting, the study by Barraquer and Roy reported deprescribing in 49% of patients without indication.¹⁶ It should be noted that our study involved a population-based strategy encompassing all individuals assigned to a primary care list.

In this population, the deprescribing strategy for PPIs in patients without risk factors for gastrointestinal bleeding was not associated with increased mortality or hospital admissions. The results support the recommendations of the American Gastroenterological Association.⁴ The strategy was not associated with an increase in mortality or admissions and was linked to a reduction in the combined outcome. This analysis represents one of the main strengths of this study, as no other studies have been identified in the literature that have investigated the effect of PPI deprescribing on mortality and hospital admissions. Only one study was identified in which PPI use was associated with increased all-cause mortality compared with the use of H2 receptor antagonists (HR 1.25; 95% CI = 1.23–1.28). In that study, PPI users also had a higher risk of death compared with non-users of PPIs and non-users of any acid-suppressive therapy [(HR 1.15; 95% CI = 1.14–1.15); (HR 1.23; 95% CI = 1.22–1.24), respectively].¹⁷

When reviewing the most characteristic adverse effects of PPIs, it was observed that patients in whom PPIs were deprescribed had a lower incidence of pneumonia and fractures, consistent with some studies.^18,19 It is important to note that pneumonias diagnosed both in primary care and in hospital emergency departments were identified, regardless of whether they resulted in hospital admission. The same approach was applied to fractures, thereby reducing the risk of information bias.

The results observed in this study show a clear association between PPI use and deficiencies in micronutrients such as calcium, magnesium, iron, and vitamin B12, consistent with findings reported in the existing literature.^6,20,21C. difficile infection is a potential adverse effect described in the summary of product characteristics and reported in some studies,^7,22–24 however, no such association was identified in our study, in line with other reports.^25,26 The incidence of infections was very low and, consequently, the confidence interval was wide, precluding firm conclusions. The higher incidence of gastrointestinal adverse effects in deprescribed patients is consistent with the development of symptoms associated with rebound acid hypersecretion⁴ and highlights the need to inform patients accordingly and to provide follow-up, with reintroduction of treatment if symptoms persist. No increase in gastrointestinal bleeding was identified; its incidence was very low, consistent with the cohort’s inclusion criteria, which included, among others, a low baseline risk of gastrointestinal bleeding.

With regard to the sex-stratified analysis, the higher proportion of women is probably due to their greater life expectancy, particularly in the 2017 strategy, whose target population was restricted to individuals over 65 years of age. This is the first study in which outcomes have been analysed by sex, made possible by the long follow-up period, the large sample size, and the high number of events across most of the variables studied. The results were consistent in men and women, with the exception of bone fractures, which may be related to the higher prevalence of osteoporosis in women, associated with hormonal changes after menopause. These are exploratory data that provide an overall description and should be further investigated in future studies to determine whether this finding is incidental or not.

One of the main strengths of this study is its large sample size. This was made possible by the availability of an electronic tool enabling identification of patients eligible for the strategy, as well as the mass sending of proposals through the patient’s electronic health record.²⁷ This is complemented by the long follow-up period, which allowed analysis of the intended health outcomes, together with the possibility of identifying complications managed both in primary care and in hospital, as well as the reasons for hospital admission.

Another strength is the low level of intervention inherent to the strategy, as patients were selected according to agreed inclusion criteria and prescribing was never restricted; instead, the decision to deprescribe PPIs remained at the discretion of the physician, with the advantage that there was no monitoring of professional behaviour, thereby avoiding a potential Hawthorne effect. In this regard, consistency in deprescribing was observed, with PPI withdrawal being more frequent among patients with fewer comorbidities.

Other strengths include the inclusion of a control group, unlike other studies reviewed in the literature,^15,28–32 and the sex-stratified analysis, which has also not been identified in the literature to date.

The results of the primary outcome were driven mainly by the hospitalisation component, although, from an epidemiological perspective, it was also considered important to include mortality within the composite outcome. This could be improved in future studies by using prospective designs that allow the relative contribution of each component to be appropriately weighted, including the assessment of competing risks.

This study has several limitations. As a retrospective observational study, it is subject to potential biases related to the quality of recorded data, such as selection bias. Furthermore, the completeness of recorded information may be influenced by the decision to deprescribe or not (information bias). As inherent to the observational nature of the study, the presence of indication bias due to unmeasured confounding factors cannot be ruled out.

To minimise bias, matching was performed according to sex, age, year of intervention (2017, 2018, 2022), Charlson index categories, number of primary care visits, and number of prescribed medications. In the absence of sufficiently validated records of frailty or functional status, the Charlson index, together with the number of medications and primary care visits, was used as a proxy for complexity and care needs partially related to frailty, with the limitations that this entails. This may partly explain the protective effect observed, particularly in hospitalisations. To address this limitation in future studies, it would be of great interest to promote improved recording of frailty and cognitive impairment indices.

Conclusions

The deprescribing strategy in patients without risk factors for upper gastrointestinal bleeding was not associated with an increase in mortality or hospital admissions, in either men or women.

In patients in whom PPIs were deprescribed, a lower incidence of hospital admissions, pneumonia and bone fractures, as well as deficiencies in micronutrients such as magnesium, calcium, iron and cyanocobalamin, was observed. With regard to gastrointestinal events, an increased risk was observed for the combined outcome of ulcers, gastritis or oesophagitis, although without an increase in the risk of serious complications such as upper gastrointestinal bleeding.

Therefore, the results of this evaluation of a PPI deprescribing strategy in patients without an indication for treatment, involving primary care physicians and supported by electronic tools, support the feasibility of similar strategies, which should be assessed prospectively.

Ethical considerations

This project received a favourable opinion from the Research Ethics Committee for Medicinal Products of the Chartered Community of Navarra at its meeting held on 18 May 2022.

A waiver of informed consent was granted for the following reasons:

• -

  This project is based on an evaluation aimed at improving prescribing safety and optimising pharmacotherapy.

• -

  Obtaining informed consent from patients was not feasible, as this was a retrospective descriptive study.

• -

  There was no randomisation or intervention by the study investigators, and therefore no additional risk to patients.

• -

  The study did not involve access to patients’ clinical records by the investigators.

• -

  All data required for the study were obtained from databases, specifically from BARDENA, a population-based database of Navarra that integrates data from the following sources of the Servicio Navarro de Salud-Osasunbidea (SNS-O). In this database, data are extracted, transformed and loaded ensuring data consistency and anonymisation. The generated datasets are anonymised with a dissociated individual identifier to prevent identification.

• -

  Proper data handling is ensured through confidentiality agreements signed by all participating investigators. The information obtained is used solely for the purposes established in the study protocol and will not be used for any other purpose.

Funding

This project was made possible thanks to funding obtained through the 2022 call for research project grants from the Department of Health of the Government of Navarra.

Acknowledgements

We thank the primary care physicians, particularly Nuria Goñi; the Digestive Diseases Department of the Hospital Universitario de Navarra, especially Federico Bolado; the Healthcare Information Systems Department; and all colleagues from the Subdirectorate of Pharmacy and Benefits who have contributed over time to the design and implementation of the strategy.