ARTICLE 1

Proton Pump Inhibitor Initiation and Withdrawal affects

Liver
Gut Microbiota and readmission risk in cirrhosis
Jasmohan S Bajaj, MD1, Chathur Acharya, MD1, Andrew Fagan, BS1, Melanie B White, RN1, Edith Gavis, RN1, Douglas M Heuman, MD1,
Phillip B Hylemon, PhD2, Michael Fuchs, MD1, Puneet Puri, MD1, Mitchell L Schubert, MD1, Arun J Sanyal, MD1, Richard K Sterling, MD1,
R Todd Stravitz, MD1, Mohammad S Siddiqui, MD1, Velimir Luketic, MD1, Hannah Lee, MD1, Masoumeh Sikaroodi, PhD3 and
Patrick M Gillevet, PhD3

OBJECTiVES: Cirrhosis is associated with gut microbial dysbiosis, high readmissions and proton pump inhibitor
(PPI) overuse, which could be inter-linked. Our aim was to determine the effect of PPI use, initiation
and withdrawl on gut microbiota and readmissions in cirrhosis.

METHODS: Four cohorts were enrolled. Readmissions study: Cirrhotic inpatients were followed throughout the
hospitalization and 30/90-days post-discharge. PPI initiation, withdrawal/continuation patterns
were analyzed between those with/without readmissions. Cross-sectional microbiota study: Cirrhotic
outpatients and controls underwent stool microbiota analysis. Beneficial autochthonous and
oral-origin taxa analysis vis-à-vis PPI use was performed. Longitudinal studies: Two cohorts of
decompensated cirrhotic outpatients were enrolled. Patients on chronic unindicated PPI use were
withdrawn for 14 days. Patients not on PPI were started on omeprazole for 14 days. Microbial
analysis for oral-origin taxa was performed pre/post-intervention.

RESuLTS: Readmissions study: 343 inpatients (151 on admission PPI) were enrolled. 21 were withdrawn and
45 were initiated on PPI resulting in a PPI use increase of 21%. PPIs were associated with higher 30
(p = 0.002) and 90-day readmissions (p = 0.008) independent of comorbidities, medications, MELD
and age. Cross-sectional microbiota: 137 cirrhotics (59 on PPI) and 45 controls (17 on PPI) were
included. PPI users regardless of cirrhosis had higher oral-origin microbiota while cirrhotics on PPI
had lower autochthonous taxa compared to the rest. Longitudinal studies: Fifteen decompensated
cirrhotics tolerated omeprazole initiation with an increase in oral-origin microbial taxa compared to
baseline. PPIs were withdrawn from an additional 15 outpatients, which resulted in a significant
reduction of oral-origin taxa compared to baseline.

CONCLuSiONS: PPIs modulate readmission risk and microbiota composition in cirrhosis, which responds to
withdrawal. The systematic withdrawal and judicious use of PPIs is needed from a clinical and
microbiological perspective in decompensated cirrhosis.
SUPPLEMENTARY MATERIAL accompanies this paper at https://doi.org/10.1038/s41395-018-0085-9

Am J Gastroenterol https://doi.org/10.1038/s41395-018-0085-9

IntroductIon [2–7]. In compensated cirrhotic patients, PPIs have also been

Proton pump inhibitors are one of the most highly prescribed associated with disruptions in the stool microbiota with increase
medications [1]. In patients with cirrhosis, these medications are in salivary-origin taxa such as Streptococcaceae [8]. However,
used extensively and in many cases, do not have an FDA-approved most infections in cirrhotic patients occur in decompensated
indication for their continued use [2]. In cirrhotic patients, PPIs cirrhosis and the effect of PPI introduction and withdrawal in
are associated with infections, as well as worsening prognosis these patients is unclear [9]. Other studies have shown that stool

1
Gastroenterology, Hepatology and Nutrition, Virginia Commonwealth university and McGuire VA Medical Center, Richmond, VA, uSA. 2Microbiology, Virginia
Commonwealth university and McGuire VA Medical Center, Richmond, VA, uSA. 3Microbiome Analysis Center, George Mason university, Manassas, VA, uSA.
Correspondence: J.S.B. (email: [email protected])
Received 13 February 2018; accepted 1 April 2018

© 2018 the American college of Gastroenterology The American Journal of GastroenteroloGy

 2 Bajaj et al.

microbiota in cirrhosis is enriched with these oral origin bacteria patients using less than this were considered non-users. We also
but it is not clear whether this is a consequence of cirrhosis or of enrolled age and gender matched healthy controls that were on
the widespread acid suppression that is prevalent in this popu- and not on PPI. All controls on PPI were on them for uncom-
lation [8, 10]. This interaction between PPI and infections are plicated gastro-esophageal reflux disease (GERD). All subjects
Liver

relevant since readmissions are a major burden in cirrhosis and underwent stool microbiota assessment pertaining to autochtho-
changes in PPI use practice could affect this high rate of readmis- nous taxa [13] (Lachnospiraceae, Ruminococcaceae), potentially
sions [11, 12]. Our aims were to (a) determine the impact of PPI pathogenic taxa (Enterobacteriaceae) and oral-origin taxa (Strep-
introduction and withdrawal on readmissions in cirrhotic inpa- tococcaceae, Veillonellaceae, and Porphyromonadaceae) [14, 15].
tients at 30 and 90 days (b) evaluate the effect of chronic PPI use The cirrhosis dysbiosis ratio (ratio of Lachnospiraceae + Rumino-
on the relative abundance of oral-origin taxa in outpatients with coccaceae + Veillonellaceae + Clostridium Cluster XIV/ Entero-
cirrhosis and healthy controls in a cross-sectional study and (c) bacteriaceae + Bacteroidaceae) was calculated [15]. Comparisons
determine the effect of systematic PPI initiation and withdrawal were performed between patients on and not on PPI. Multi-vari-
on gut microbial taxa of oral origin in decompensated cirrhotic able predictors of autochthonous taxa and oral taxa relative abun-
outpatients in two separate longitudinal studies. dance were analyzed within cirrhosis.

Cirrhotic patients studied before and after PPI initiation

Methods A separate cohort of outpatients with decompensated cirrhosis
There were four prospective cohorts included in this study. All of defined by prior variceal bleeding, controlled ascites (on diuretics)
these were approved by the Institutional Review Board at Virginia and jaundice who were not on PPI were recruited. We excluded
Commonwealth University and McGuire VA Medical Center and patients who were allergic to PPI, those who were unwilling to
all subjects provided written informed consent before study ini- consent, on treatments for hepatic encephalopathy (rifaximin,
tiation. lactulose), those on SBP prophylaxis or those on recent (6 weeks)
antibiotic or probiotic therapy. All patients who agreed under-
Readmissions study went blood and stool collection at baseline and were given 40 mg
Inpatients with cirrhosis who were hospitalized for non-elective QD of oral omeprazole for 14 days (Fig. 3a). A phone call to assess
reasons and were without HIV infection or prior transplant were safety, tolerability and adherence was made at day 7 and repeat
included. PPI use details (indication, dose and type) at admis- stool and blood collection was performed at day 14. All subjects
sion, discharge, as well as withdrawal/addition of PPI during the were instructed to continue a standard diet. Safety and tolerabil-
hospitalization were recorded. Patients gave written informed ity, emergence of new cirrhosis complications, MELD score and
consent, which was obtained from the patients’ legally author- gut microbiota were analyzed for before/after PPI therapy. Sample
ized representatives in case they were confused. We only included size analysis for the cross-sectional study was based on published
patients who were discharged alive and without liver transplant data in which we found a significant increase in Streptococcaceae
to home and had outcomes noted at one and three months post- abundance in compensated cirrhotic patients (0.0 to 9.0%) and in
discharge. The inpatient course, including infections, noso- controls (0.0 to 5%) [8]. We assumed that a similar change would
comial infections, organ failures, acute on chronic liver failure be found in controls and cirrhotic patients with/without PPI
(ACLF), and ICU transfer were recorded in addition to major cross-sectionally, requiring at least 11 subjects per group in either
co-morbid conditions. The major outcomes were readmissions controls with and without PPI using 0.05 level of significance and
at one and three months from discharge. Variables pertaining 80% power.
to PPI withdrawal/addition during the hospitalization were ana-
lyzed. Determinants of readmissions at both time points were Cirrhotic patients studied before and after PPI withdrawal
analyzed using binary logistic regression with clinical variables Another cohort of outpatients with decompensated cirrhosis
(age, gender, MELD score, co-morbid conditions, infections on again defined by defined by prior variceal bleeding, controlled
admission, PPI use change, rifaximin, lactulose, beta-blockers ascites (on diuretics) and jaundice who were on PPI for an indica-
and SBP prophylaxis) that were p < 0.10 on univariate analysis. tion other than an FDA-approved indication for at least 1 month
Sample size was calculated based on a prior study with a similar prior to enrollment were included. FDA approved PPI use indica-
population that showed a 53% 90-day readmission risk [12]. We tions [16] are (a) Healing and maintenance of erosive esophagi-
hypothesized that without PPI this would reduce to 37.5%, which tis, (b) Treatment of gastro esophageal reflux disease (GERD), (c)
at an alpha of 0.05 and power of 80% would require at least 161 Risk reduction for peptic ulcer with non-steroidal anti-inflam-
patients per group. matory drugs (d) Helicobacter pylori eradication, in combina-
tion with antibiotics; (e) Pathological hyper-secretory conditions,
Cross-sectional microbiota study including Zollinger-Ellison syndrome; (f) Short-term treatment
In a group of cirrhotic outpatients who could provide informed and maintenance of peptic ulcer therapy. We excluded patients
consent, stool was collected for microbiota composition. Con- who were unwilling to consent or withdraw PPI, on treatments
comitant medications including PPI use, cirrhosis severity and for hepatic encephalopathy (rifaximin, lactulose), those on SBP
complications were recorded. PPI use was defined as continuous prophylaxis or those on recent (6 weeks) antibiotic or probiotic
use of PPIs daily for at least 1 month prior to the stool collection, therapy. Patients were then asked to withdraw from PPI therapy

The American Journal of GastroenteroloGy www.nature.com/ajg

 Proton Pump Inhibitor Initiation and Withdrawal affects... 3

Pasien sirosis rawat inap

(n = 343)

Liver
Diberikan PPI (n =151) Tanpa Pemberian PPI(n =192)

No Yes No Yes
Pulang dengan Pulang dengan
PPI PPI
Dihentikan Tetap Ditambahkan Tetap
N =24 menggunakan PPI Menggunakan
PPI N =45 PPI N =147
N =127

Pulang dengan PPI (n =172) Pulang tanpa PPI (n =171)

30 hari pengobatan kembali (n = 86) 30 hari pengobatan kembali (n =55)

90 hari pengobatan kembali (n = 97) 90 hari pengobatan kembali (n =72)

Fig. 1 Readmissions study, progression of cirrhotic patients through the inpatient study with respect of proton pump inhibitor continuation, withdrawal, and
initiation

for 14 days. A phone call to assess safety, tolerability and adher-

ence was made at day 7 and repeat stool and blood collection was
performed at day 14. This duration was chosen given prior results
in healthy controls and compensated cirrhotic patients [8]. All
subjects were instructed to continue a standard diet. Safety and
tolerability, emergence of new cirrhosis complications, MELD
score and gut microbiota were analyzed for before/after PPI with-
drawal.
Sample size analysis for both longitudinal studies was based on
published data in which we found a significant increase in salivary
Streptococcaceae abundance in compensated cirrhotic patients (0.0
to 9.0%) [8]. We assumed that a similar disruption would occur in
both situations in which there are PPI compared to the situations
without PPI (at baseline for group 1 and end-of-study for group 2)
but to a smaller extent due to the baseline dysbiosis in decompen-
sated patients. Using a paired t-test at a 0.05 level of significance
and assuming a change of 7% and a standard deviation of 8%, a
sample size of at least 13 would give us 80% power to detect this
effect size.

Microbiota analysis
Detailed methods. Stool samples were analyzed after DNA ex-
traction and multi-tagged sequencing using published techniques
(Supplementary information) [15, 17]. The unique fraction
metric or UNIFRAC measures the differences between micro-
bial communities using the phylogenetic information and is a
validated method to determine diversity changes between groups
[18, 19]. This was used to compare overall microbial differences
between PPI users and non-users in controls and cirrhotic groups.
A principal component analysis was performed in the cross-
sectional study to determine clustering in controls compared
© 2018 the American college of Gastroenterology The American Journal of GastroenteroloGy
4 Bajaj et al.
to cirrhotic patients with/without PPI use. In
addition, taxa be- tween and within groups were
compared using non-parametric approaches
(Kruskal–Wallis tests and Wilcoxon matched pairs
ranked tests) with a focus on bacterial taxa of interest
such as po- tentially beneficial autochthonous taxa
(Lachnospiraceae, Rumi- nococcaceae) and salivary-
origin families (Porphyromonadaceae, Veillonellaceae
and Streptococcaceae).

results
Readmissions study
There were 343 inpatients with cirrhosis recruited
prospectively from VCU and Richmond VAMC
centers. Of these, 151 patients were on PPI on
admission (Fig. 1, Table 1). The reasons for PPI use
were GERD (n = 43), peptic ulcer disease (n = 21),
Bar- rett’s esophagus (n = 18), concomitant NSAID
therapy (n = 7), after variceal bleeding or after
banding (n = 24), abdominal pain (n = 14), and rest
of the 24 were for unknown reasons. The mean ± SD
duration of PPI therapy was 7.45 ± 4.36 years. Most
patients were on omeprazole (n = 106) followed by
esomepra- zole (n = 23), pantoprazole (n = 22),
and lansoprazole (n = 3). Patients on PPI had higher
rates of HE, SBP prophylaxis and beta- blocker use
compared to those not on PPI but admission MELD
score, serum albumin and WBC counts were
statistically similar between groups (Table 1). PPI
users had a similar rate of most inpatient outcomes
but had higher inpatient organ failures com- pared to
non-users. Of these inpatients on PPI, there was
with- drawal of PPI in 24 patients and continuation in
127 patients. The reasons for withdrawal were
mostly recognition of unnecessary use (n = 16),
change to occasional H2 receptor blockers (n = 5)
and inadvertent stoppage (n = 4). Of the 192
patients who were

The American Journal of GastroenteroloGy www.nature.com/ajg

 Proton Pump Inhibitor Initiation and Withdrawal affects... 5

Table 1 Pasien Rawat Inap berdasarkan Penggunaan PPI hospitalized without PPI, 45 had PPI added during the hospi-
talization and 147 remained off PPIs. The 45 patients with PPI
Tanpa PPI Dengan PPI addition were due to acid peptic disease (n = 31), abdominal pain
N = 192 N = 151
(n = 9) and post-banding (n = 5). Most of these 45 patients were
Liver

Usia (tahun) 56.5 ± 9.5 58.2 ± 9.0 discharged on omeprazole (n = 36) followed by esomeprazole
Laki-Laki 56 44 (n = 6) and lansoprazole (n = 3). Ultimately 172 patients were dis-
Etiology (HCV/Alc/HCV + Alc/NASH/ 78/21/38/42/13 67/10/29/35/10 charged on PPI and 172 were discharged without PPI and at the
others) time of discharge, the profile of the patients was largely similar to
Parameter Penilaian that of the admission variables.
At one-month post-discharge, we found that 141 patients were
Skor MELD 16.7 ± 7.8 17.5 ± 8.3
readmitted, mostly due to liver disease complications (n = 95, HE
Perhitungan Sel Darah 7.4 ± 4.1 7.1 ± 4.1 n = 40, infections n = 28, renal/metabolic n = 13 and rest n = 14).
Putih (/X1000 mm3)
The readmission proportion was significantly higher in those dis-
Serum albumin (g/dl) 2.7 ± 0.6 2.8 ± 0.6 charged on PPI compared to those who were not (n = 86, 50%
Serum Na (meq/L) 135.9 ± 5.4 135.4 ± 5.7 compared to n = 55, 32%, p = 0.02). This was also significant on
Terpapar Infeksi 40 42 multi-variable analysis where age (OR 0.97, 95% CI 0.94−0.99,
Beta-blockers 66 79* p = 0.04) and PPI use (OR: 1.99, 95% CI 1.28−3.08, p = 0.002)
were predictors of readmission. Other criteria, including admis-
Rifaximin 34 56*
sion and discharge MELD score, co-morbid conditions, serum
Lactulose 72 91* albumin, specific HE, and SBP-related medications and alcoholic
Profilaksis SBP 10 21* etiology of cirrhosis were not additionally contributory to this pre-
Kondisi Co-Morbid diction.
Diabetes 67 48
At three months post-discharge, 169 patients (49%) were read-
mitted. Most of these were due to liver-related reasons (n = 113, HE
Penyakit Serebrovaskular 3 2
n = 45, infections n = 34, renal/metabolic reasons n = 16, upper GI
Penyakit Cardiovaskular 40 36 bleeding n = 18) followed by others (cardiovascular n = 24, respir-
(Hipertensi, CHF, Penyakit
Jantung Koroner, Penyakit atory n = 19, orthopedic n = 8, and others n = 5). Of these 169, a
vaskuler perifer) significantly higher proportion of patients on PPI were readmit-
Penyakit Paru Obstruktif Kronis 12 5 ted at 90 days (n = 97, 54%) compared to the patients not on PPI
on discharge (n = 72, 41%, p = 0.008). Patients who were readmit-
Kondisi selama Perawatan ted were also had a similar MELD score on discharge (readmitted
19.9 ± 7.4 vs. not 18.9 ± 10.2, p = 0.27), rate of discharge beta-
Pemberian GIT selama pemberian
blocker use (44% vs. 39%, p = 0.34), rifaximin use (29% vs. 33%,
Transfer ke ICU 30 30
p = 0.37), lactulose use (59% vs. 56%, p = 0.63) and SBP prophy-
Infeksi Sekunder 9 12 laxis (16% vs. 14%, p = 0.68). There was a trend towards a lower age
Pemasangan Ventilasi 14 16 of patients who were readmitted, 56.3 ± 8.9 vs. others 58.2 ± 9.7
Mekanik (p = 0.06). On univariate analysis, variables studied were age, gen-
Dialisis 5 14* der, co-morbid conditions, MELD on discharge, SBP prophylaxis,
HE grade 3 or 4 20 28* rifaximin, lactulose, beta-blockers, PPI use and alcoholic etiology
Shock 14 20* of cirrhosis. The only variables with p < 0.10 were age and PPI use
and in a multi-variable analysis, the significant predictors were PPI
ACLF 12 17
use (OR 1.6, 95% CI 1.1−2.5, p = 0.03) and age (OR 0.96, 95% CI
Saat Pemulangan
0.94−0.98, p = 0.03).
Skor MELD 18.8 ± 9.3 19.5 ± 8.5 Patterns of PPI use were consistent in that all patients on PPI
Serum albumin 2.9 ± 0.8 2.8 ± 0.7 continued their therapy at one and three months while the patients
Beta-Blockers 66 73* not on PPI remained off PPI in those same time periods. Addition-
ally, 5 patients who were not on PPI were started on H2 receptor
Rifaximin 44 63*
blockers.
Lactulose 99 97*
Profilaksi SBP 24 27 Cross-sectional microbiota study
Terapi PPI 171 172 137 cirrhotic outpatients and 45 healthy controls were included.
Data yang ditampilkan adalah rata-rata ±SD dan angka belum terproses, There were no significant differences between groups with respect
*p<0.05 diantara kelompok dalam Kruskal-Wallis dan t-test yang to age, gender and race. Of these 59 cirrhotic patients were on PPI
disesuaikan. ACLF : gejala akut pada kegagalan hati kronis berdasarkan
Konsorsium Amerika Utara sebagai Kriteria Studi Tahap Akhir Penyakit
for an average of 8.2 ± 2.6 years for GERD (n = 16), peptic disease
Liver, PPI : proton pump inhibitor, HE : ensepalopati hepar, SBP : (n = 12), NSAID therapy (n = 5) and unknown reasons (n = 26).
peritonitis bakteri spontan, HCV : hepatitis C virus, Alc : alkoholik,
NASH : steatohepatitis non alkoholik

© 2018 the American college of Gastroenterology The American Journal of GastroenteroloGy

6 Bajaj et al.

a Changes in gut microbiota in subjects with and without PPI

**
0.5 *

Liver
0.4
Relative abundance
*

0.3
** **

0.2 *
*

0.1
* * *
*

0.0

CtrN CtrP CirrN CirrP CtrN CtrP CirrN CirrP CtrN CtrP CirrN CirrP CtrN CtrP CirrN CirrP
Lachno Rumino Strepto Veillo

Principal component analysis

Controls Cirrhotic patients

b PCO case scores (Bray Curtis) PCO case scores (Bray Curtis)
0.47 0.6

0.38 0.5

0.28 0.4
0.19
0.2
Axis 2

0.09
Axis 2

0.1

–0.47 –0.38 –0.28 –0.19 –0.09 0.09 0.19 0.28 0.38 0.47
–0.09
–0.5 –0.4 –0.2 –0.1 0.1 0.2 0.4 0.5 0.6
–0.19 –0.1

–0.28 –0.2

–0.38
–0.4
–0.47
–0.5
Axis 1 Axis 1

Fig. 2 Cross-sectional microbiota study, a Relative changes in gut microbiota composition of key taxa in outpatient subjects with and without proton pump
inhibitor use. Data are presented as median and 95% Ci and comparisons are performed using Kruskal–Wallis tests, *p < 0.05, **p < 0.01, CtrN: controls
not on proton pump inhibitors, CtrP: controls on proton pump inhibitors, CirrN: cirrhotic patients not on proton pump inhibitors, CirrP: cirrhotic patients on
proton pump inhibitors, Lachno: Lachnospiraceae, Rumino: Ruminococcaceae, Strepto: Streptococceae, Veillo: Veillonellaceae. b Principal component
analysis of stool microbiota from healthy controls who are on PPi (red) compared to those not on PPi (yellow) showed clustering of the PPi users. c Prin-
cipal component analysis of stool microbiota from cirrhotic patients who are on PPi (red) compared to those not on PPi (yellow) did not show any specific
clustering

The leading PPIs used were omeprazole (n = 45) followed by between controls and cirrhotic patients without PPI. Similarly,
esomeprazole (n = 12) and lansoprazole (n = 2). The 17 healthy the Streptococcaceae relative abundance was statistically similar
controls were on PPI due to GERD for 6.9 ± 4.5 years and were between PPI users regardless of their control or cirrhosis status.
mostly on omeprazole (n = 11) followed by esomeprazole (n = 6). On principal component analysis, there was a clustering in con-
Cirrhotic patients on PPI had a higher MELD score and propor- trols (Fig. 2a) who were using PPI but not in cirrhotic patients
tion with HE than those who were not (Table 2). (Fig. 2c). The cirrhosis dysbiosis ratio was significantly lower (i.e.,
Microbiota analysis showed a significant difference between there was more dysbiosis), in PPI users compared to non-users
groups with respect to UNIFRAC (p < 0.001). There was a signifi- (0.42 ± 0.32 vs. 0.59 ± 0.26, p = 0.005). However, the autochtho-
cantly lower relative abundance of autochthonous taxa and higher nous family, Lachnospiraceae was significantly lower in cirrhotic
Streptococcaceae in PPI users compared to the rest in cirrhotic patients compared to controls regardless of PPI use status. Since
patients and in healthy controls (Fig. 2a). Interestingly, there was within the cirrhosis group, PPI use was higher in those with
no significant difference in relative abundance of Streptococcaceae relatively more advanced liver disease, multi-variable regression

The American Journal of GastroenteroloGy www.nature.com/ajg

 Proton Pump Inhibitor Initiation and Withdrawal affects... 7

Table 2 Pasien Rawat Jalan Dengan atau Tanpa Sirosis Berdasarkan Penggunaan PPI

Kontrol (n = 45) Pasien Sirosis (n = 137)

Tanpa PPI Dengan PPI Tanpa PPI Dengan PPI

Liver

(n = 28) (n = 17) (n = 78) (n = 59)

Usia 54.9 ± 8.9 58.8 ± 6.6 56.6 ± 6.7 56.9 ± 7.1

Laki-Laki 71% 71% 74% 75%
Ras (Caucasian/AA/Hispanic/lainya) 12/10/6/0 9/5/4/0 54/22/2/0 48/10/1/0
Diabetes % - - 37% 27%
Ascites - - 45% 53%
Skor MELD - - 12.1 ± 6.6 15.4 ± 7.2*
Dengan Varises 23 33*
Riwayat HE sebelumnya - - 31% 54%*
Lactulosa - - 29 21
Penambahan Rifaximin - - 12 17*
Profilaksis SBP - - 2 3
Beta-Blockers - - 22 33*
Data disajikan sebagai data rerata ± SD dan bilangan baku dinyatakan lain, * p <0,05 pada kelompok pada Kruskal Wallis dan T-tes yang sesuai
PPI: inhibitor pompa proton, HE: hepatic encephalopathy, SBP : spontaneous bacterial peritonitis

models were created for each family including MELD score, HE (Fig. 3b). No changes in Veillonellaceae were seen nor changes in
status, age, gender, alcoholic etiology, PPI use, SBP prophylaxis other families.
use, rifaximin use, lactulose use, and beta-blocker use.
Within the cirrhosis cohort, the Lachnospiraceae relative abun- Cirrhotic patients studied before and after PPI withdrawal
dance was predicted by MELD score (p < 0.0001) and PPI use We enrolled 15 patients with decompensated cirrhosis on chronic
(p = 0.02) independently, while Ruminococcaceae was predicted PPI (all on omeprazole, 12 on 40 mg QD and 3 on 20 mg QD).
only by MELD score (p < 0.001). The only factor that significantly The decompensation was due to ascites in all patients, which was
predicted Streptococcaceae was PPI use (p = 0.03). The cirrhosis controlled using diuretics. The patients were on PPI for unap-
dysbiosis ratio was predicted with the MELD score (p = 0.001), proved indications (abdominal pain in 6, occasional GERD in
PPI use (p = 0.02) and diabetes (p = 0.05). 5, prior non-peptic bleeding in 2 and unknown indications in 3
Both cirrhotic patients and controls were studied with the vari- patients). No changes in symptoms, including heartburn occurred
ables included being age, control vs. cirrhosis status, gender and in between the time when they were off the PPIs. Again, no
PPI use. As in within the cirrhosis group, the only predictor for significant change in MELD score or new complications was seen
Streptococcaceae relative abundance was PPI use (p = 0.02) inde- during this period (Table 3). On microbiota analysis, there was
pendent of the cirrhosis status. Similarly, Ruminococcacae relative a significant decrease in Porphyromonadaceae, Streptococcaceae,
abundance was predicted by cirrhosis vs. control status (p = 0.001). and Veillonellaceae relative abundance after PPI withdrawal com-
For the Lachnospiraceae relative abundance, control vs. cirrhosis pared to pre-PPI therapy (Fig. 3c).
status (p < 0.0001), PPI use (p < 0.0001) and age (p = 0.005).

Cirrhotic patients studied before and after PPI initiation dIscussIon

We enrolled 15 cirrhotic outpatients not on PPI were prescribed The results of the current study demonstrate that PPI use is often
omeprazole 40 mg to be taken daily before breakfast for 14 days. indiscriminate and initiation and withdrawal of PPIs in decompen-
All patients had decompensation due to ascites which was well- sated cirrhosis can impact clinical outcomes such as readmissions
controlled on diuretics. None of the subjects had current or prior and change the gut microbial composition. PPI use in cirrhosis
SBP. These patients were then followed for safety, tolerability, has been associated with several negative outcomes in population-
as well as microbiota analysis. All patients demonstrated 100% based and prospective inpatient/outpatient studies [11]. Few of
adherence on the medications and did not demonstrate any safety these studies, however, have focused on readmissions, which have
or tolerability concerns. There was no change in the MELD score emerged as a major burden on the healthcare system in the US
nor any new complication development during the trial (Table 3). [20]. In a smaller prior study, PPI use was independently asso-
There was a significant increase in the median relative abundance ciated with newer infection-related readmissions in cirrhotic
of Porphyromonadaceae and Streptococcaceae after PPI initiation patients who were included with infections [11]. In a larger study

© 2018 the American college of Gastroenterology The American Journal of GastroenteroloGy

8 Bajaj et al.

a
STUDY Gp 1: Sirosis Gp 2: Sirosis Dekompensasi
OVERVIEW Dekompensasi tanpa dengan Pengobatan Sirosis
Pengobatan PPI tanpa alasan yang jelas

Liver
Screening:
kelayakan dan diet

Visit 1: Microbiome, Omeprazole 40 Penghentian PPI

MELD mg/day selama 14 days selama 14 hari

Panggilan Telepon
selama pengujian:
keterkaitan dan
komplikasi

Kunjungan 2: Microbiome, Gp 1 Pengobatan PPI Gp 2 tanpa Pengobatan

MELD selama 14 hari PPI selama 14 hari

b Porphyromonadaceae p = 0.001 e Porphyromonadaceae p = 0.004

0.25 0.3
Relative abundance
Relative abundance

0.20
0.2
0.15

0.10
0.1
0.05

0.00 0.0
Off PPI On PPI On PPI Off PPI

c Streptococcaceae p = 0.02 f Streptococcaceae p = 0.01

0.15 0.15
Relative abundance
Relative abundance

0.10 0.10

0.05 0.05

0.00 0.00
Off PPI On PPI On PPI Off PPI

d Veillonellaceae p = 0.52 g Veillonellaceae p = 0.03

0.20 0.10
Relative abundance
Relative abundance

0.08
0.15
0.06
0.10
0.04
0.05
0.02

0.00 0.00
Off PPI On PPI On PPI Off PPI

Fig. 3 Longitudinal studies a Schematic of the study design. b Before and after omeprazole initiation: Relative abundance of key oral-origin microbiota
pre-therapy values compared to post-initiation after 14 days tracked within individuals are presented. Analyses performed using Wilcoxon matched pairs
test with p-values displayed in sub-headings. c Before and after omeprazole withdrawal: Relative abundance of key oral-origin microbiota while on proton
pump inhibitors compared to post-withdrawal for 14 days tracked within individuals are presented. Analyses performed using Wilcoxon matched pairs test
with p-values displayed in sub-headings

The American Journal of GastroenteroloGy www.nature.com/ajg

 Proton Pump Inhibitor Initiation and Withdrawal affects... 9

Table 3 Longitudinal studies for PPI initiation and withdrawal in decompensated cirrhosis

Studi Inisiasi (n = 15) Studi Penghentian (n = 15)

Tanpa PPI 14 hari setelah PPI Dengan PPI 14 hari berhenti PPI
Liver

Usia 58 ± 7 59 ± 6
Jenis Kelamin Laki-Laki 14/1 13/2
Ras (Caucasian/AA/Hispanic/lainya) 9/6/0/0 8/5/2/0
Etiologi (HCV, Alc, HCV + Alc, NASH, others) 6/2/1/4/2 7/2/3/3/0
Diabetes 5 6
MELD score 13.2 ± 4.5 13.5 ± 5.1 13.2 ± 5.3 12.8 ± 4.9
Ascites 15 15
Asupan Kalori selama 24 jam 2241 ± 145 2308 ± 162 2401 ± 236 2387 ± 196
Keterikatan pada regimen – 100% – 100%
The duration of the trials was 14 days each. None of the patients were on hepatic encephalopathy treatment or SBP prophylaxis throughout the trial. No new or worsening
cirrhosis complications or changes in MELD score or dietary intake were noted in both trials
PPI proton pump inhibitor, HCV hepatitis C virus, Alc alcoholic, NASH non-alcoholic steatohepatitis

with infected and uninfected patients, PPI use on discharge had decompensated cirrhosis [8, 21, 23, 24]. In addition, PPI use was
a trend towards predicting 90-day readmissions [12]. However, associated with Lachnospiraceae relative abundance independent
the current study tracks the PPI use including withdrawal as an of the cirrhosis status of the subject. It is important to note that
inpatient, as well as addition of these medications in a granular Lachnospiraceae and Streptococcaceae but not Ruminococcaceae,
manner to provide a more detailed insight into their impact on another autochthonous family, were associated with PPI use. This
readmissions. While 24 patients of those admitted on PPI were likely denotes that the PPI effect is not simply a function of advanc-
withdrawn from therapy, 45 patients were initiated on PPI, lead- ing liver disease, in which both Lachnospiraceae and Ruminococ-
ing to a net increase in PPI therapy on discharge by 15% compared caceae would be affected [15, 26]. When overall microbial change
to when these patients were admitted. More importantly, despite was studied using UNIFRAC, it was significantly different between
having similar MELD scores and concomitant medication use groups, although clustering on principal component analysis was
at the time of readmission, PPI-users had a greater readmission only seen in control PPI users. This is likely due to the higher
rate than non-users. This was independent of usual risk factors impact of PPI in controls who were chosen to be free of other dis-
for readmission such as MELD score, concomitant medications eases and medications. In contrast, cirrhotic patients had other co-
and alcoholic etiology of cirrhosis, while age was protective. As morbid conditions, and were likely to be on other medications that
expected, once discharged on PPI, all patients stayed on PPI ther- could dilute the relative impact of PPI. However, despite this lack
apy or remained off it in those who were discharged without it. of clustering, the cirrhotic PPI users’ overall microbiota were dif-
The strength of this evaluation is the close monitoring of patients, ferent from cirrhotic non-users on UNIFRAC and in the bacterial
their dates of initiation and withdrawal of PPI, as well as their taxa of interest. Ultimately, the association of the PPI use with a
readmissions, which were captured completely. microbial relative abundance pattern that denotes advancing dis-
While there are several potential reasons for these negative out- ease (lower cirrhosis dysbiosis ratio), independent of MELD score
comes in cirrhosis and PPI, the changes in gut microbiota have could be a potential mechanism behind the negative outcomes.
been hypothesized as a major potential mechanism [21–24]. Prior However, within a cross-sectional study, especially in which, like
studies in healthy subjects, compensated cirrhosis, and alcohol- prior studies, the PPI use rate is higher in more advanced patients,
induced rodent models point towards this mechanism but the the effect of cirrhosis compared to PPI is difficult to delineate [2–4,
impact of withdrawal/initiation in advanced or decompensated 27]. Therefore, we performed the two longitudinal studies evaluat-
cirrhosis is unclear [8, 25]. The cross-sectional microbiota study in ing PPI introduction and withdrawal over 14 days in decompen-
outpatients with and without cirrhosis demonstrated that poten- sated cirrhosis. A similar study of PPI introduction in compensated
tially beneficial autochthonous taxa such as Lachnospiraceae were cirrhosis and healthy controls demonstrated a significant disrup-
lower in those with PPI and salivary families such as Streptococ- tion in gut microbiota composition [8]. The population selected
caceae were higher in PPI users. Interestingly, Streptococcaceae was free of HE and antibiotic use but was decompensated due to
relative abundance was higher in PPI users regardless of whether ascites. Patients with pre-existing HE were excluded due to the
they had cirrhosis or not, and was only significantly predicted impact of HE-related medications on the microbiota that could
by PPI use on multi-variable analysis. This confirms studies in potentially obscure the impact of PPI therapy [15, 28]. The results
healthy controls done by our group and others and extends it onto demonstrate that oral origin microbiota significantly increased

© 2018 the American college of Gastroenterology The American Journal of GastroenteroloGy

10 Bajaj et al.

after PPI initiation, and significantly reduced after the PPI therapy and reversed during the same period even in decompensated cir-
was withdrawn. This is encouraging because it demonstrates that rhosis. The systematic withdrawal and judicious use of PPI should
a pattern like compensated cirrhotic patients and controls is fol- be encouraged from a clinical and microbiological perspective in
lowed in decompensated patients and changes related to long-term decompensated cirrhosis. Further studies of PPI withdrawal in

Liver
PPI use can be reversed in as short as 14 days after withdrawal. larger populations to study this systematically are needed.
The relevance of oral-origin bacteria in cirrhosis is controversial
[29]. Prior experience demonstrates that cirrhotic patients have CONFLICT OF INTEREST
altered salivary microbial composition and that the gut micro- Guarantor of the article: JSB is the guarantor.
biota can increase their relative salivary-origin microbiota rela- Specific author contributions: JSB conceptualized the study, MS
tive abundance with only 14 days of PPI therapy [8, 30]. Another and PMG were involved in microbiota analysis, PBH was involved
report cross-sectionally determined again that cirrhotic subjects in the critical revisions, CA, MBW, AF, and EAG were involved in
have higher oral-origin taxa in their stool [10]. But whether this research coordination, Rest of the authors were involved in subject
is a phenomenon related to cirrhosis itself and is complicit in its recruitment and follow-up.
progression, or is an epiphenomenon of impaired bile acid secre- Financial support: This work was partly supported from American
tion or gastric acid suppression due to PPI use or cirrhosis itself is College of Gastroenterology Clinical Research Award, VA Merit
an important question. Our results demonstrate that this is likely Review I0CX001076, McGuire Research Institute and an investiga-
to be affected significantly by PPI use, since withdrawal of PPIs tor-initiated grant by Grifols Pharmaceuticals, all to JSB. None of the
resulted in a dramatic reduction and vice versa, even in decom- study sponsors had any role in the study design, collection, analysis,
pensated cirrhosis. Given the some Streptococcal species can cause and interpretation of the data and in the writing of the report.
SBP; they may not be entirely innocent bystanders in this situation Potential competing interests: None.
[31]. In addition, Porphyromonadaceae has been associated with
systemic inflammation, and brain dysfunction associated with HE
[32, 33]. The lower relative abundance of Lachnospiraceae with a Study Highlights
concomitant Streptococcaceae increase could potentially indicate
that the oral origin taxa could be crowding out the beneficial taxa,
which in turn could also aid in cirrhosis progression. Therefore,
the increase in oral origin taxa in the gut microbiota by PPI in cir- Readmissions are a major healthcare burden in cirrhosis and
rhosis is indicative of a major disruption and could be a marker current clinical data are often inadequate to predict this.
for worse prognosis. The mechanism of these changes are unclear Proton pump inhibitor use is epidemic, is usually due to unin-
but a direct impact of PPI on immune capability, in addition to the dicated reasons in advanced cirrhosis, and is often associated
relatively increased abundance of these taxa due to acid suppres- with negative outcomes.
sion could be hypothesized to synergize in this situation [34]. The impact of proton pump withdrawal and initiation in the
The study is limited by a relatively small number of subjects who inpatient setting on readmissions is unclear
were withdrawn from PPI as inpatients. However, most of the rea- Proton pump inhibitor use can increase in oral-origin micro-
sons for withdrawal were related to a greater awareness of the neg- biota in the stool after short-term treatment in compensated
ative impact of unnecessary PPI use and will likely increase with cirrhosis, which could affect these outcomes.
greater education. We also excluded HE patients from the longitu- However, the effect of proton pump inhibitor withdrawal and
dinal studies and studies including them in larger sample sizes are initiation in decompensated cirrhosis, who experience many
required. We only studied the stool but not the mucosal microbial of the negative outcomes in cirrhosis, needs to be studied
taxa, which could be changed differentially. We also did not find a
significant effect of antibiotics (SBP prophylaxis and rifaximin) on
the multi-variable analysis of microbiota, but this could be related In inpatients with cirrhosis, almost a fifth of patients are either
to the smaller sample size of patients on SBP prophylaxis and the initiated or withdrawn from proton pump inhibitors, which
predominant effects of rifaximin on microbial function rather than can predict 30 and 90-day readmissions independent of cir-
composition [28]. We also assumed that all PPIs would have similar rhosis severity and concomitant medications.
effects on outcomes and microbiota, however, most patients were on
omeprazole and this agent was used for the initiation study as well. proton pump inhibitor therapy have a higher relative abun-
The findings of this study demonstrate that PPI use patterns dance of oral-origin taxa and in cirrhotics, associated with
change significantly during hospitalizations, which can impact 30 lower potentially beneficial autochthonous taxa.
and 90-day readmission risk independent on cirrhosis severity,
hepatic encephalopathy and concomitant medications. Chronic inhibitors, omeprazole initiation and use over two weeks
PPI use is associated with lower autochthonous and higher oral increases oral-origin taxa, while in another group of decompen-
origin taxa in cirrhotic outpatients, which is independent of cir- sated cirrhotic outpatients on chronic proton pump inhibitors,
rhosis severity and concomitant medications. These increases in the baseline high oral-origin microbiota relative abundance is
oral origin taxa can be brought upon by a two-week PPI course reversed after omeprazole is withdrawn for two weeks.

The American Journal of GastroenteroloGy www.nature.com/ajg

 Proton Pump Inhibitor Initiation and Withdrawal affects... 1
1

17. Hamady M, Knight R. Microbial community profiling for human

references microbiome projects: tools, techniques, and challenges. Genome Res.
1. Johnson DA, Katz PO, Armstrong D, et al. The safety of appropriate use of 2009;19:1141–52.
over-the-counter proton pump inhibitors: an evidence-based review and 18. Lozupone CA, Stombaugh JI, Gordon JI, et al. Diversity, stability and
Delphi consensus. Drugs. 2017;77:547–61. resilience of the human gut microbiota. Nature. 2012;489:220–30.
2. Deshpande A, Pasupuleti V, Thota P, et al. Acid-suppressive therapy is 19. J Gregory Caporaso, Justin Kuczynski, Jesse Stombaugh, et al.
Liver

associated with spontaneous bacterial peritonitis in cirrhotic patients: a Nature Methods, 2010; https://doi.org/10.1038/nmeth.f.303.
meta-analysis. J Gastroenterol Hepatol. 2013;28:235–42. 20. Tapper EB. Challenge accepted: confronting readmissions for our patients
3. Dam G, Vilstrup H, Watson H, et al. Proton pump inhibitors as a risk with cirrhosis. Hepatology. 2016;64:26–8.
factor for hepatic encephalopathy and spontaneous bacterial peritonitis in 21. Imhann F, Bonder MJ, Vich Vila A, et al. Proton pump inhibitors affect
patients with cirrhosis with ascites. Hepatology. 2016;64:1265–72. the gut microbiome. Gut. 2016;65:740–8.
4. Cole HL, Pennycook S, Hayes PC. The impact of proton pump inhibitor 22. Freedberg DE, Toussaint NC, Chen SP, et al. Proton pump inhibitors alter
therapy on patients with liver disease. Aliment Pharmacol Ther. 2016; specific taxa in the human gastrointestinal microbiome: a crossover trial.
44:1213–23. Gastroenterology. 2015;149:883–5.
5. Tsai CF, Chen MH, Wang YP, et al. Proton pump inhibitors increase risk 23. Naito Y, Kashiwagi K, Takagi T, et al. Intestinal dysbiosis secondary to
for hepatic encephalopathy in patients with cirrhosis in a population proton-pump inhibitor use. Digestion. 2018;97:195–204.
study. Gastroenterology. 2017;152:134–41. 24. Jackson MA, Goodrich JK, Maxan ME, et al. Proton pump inhibitors alter
6. Bajaj JS, Ratliff SM, Heuman DM, et al. Proton pump inhibitors are the composition of the gut microbiota. Gut. 2016;65:749–56.
associated with a high rate of serious infections in veterans with 25. Llorente C, Jepsen P, Inamine T, et al. Gastric acid suppression promotes
decompensated cirrhosis. Aliment Pharmacol Ther. 2012;36:866–74. alcoholic liver disease by inducing overgrowth of intestinal Enterococcus.
7. Bajaj JS, Zadvornova Y, Heuman DM, et al. Association of proton pump
inhibitor therapy with spontaneous bacterial peritonitis in cirrhotic Nat Commun. 2017;8:837. -
26. Chen Y, Yang F, Lu H, et al. Characterization of fecal microbial communi
patients with ascites. Am J Gastroenterol. 2009;104:1130–4. ties in patients with liver cirrhosis. Hepatology. 2011;54:562–72.
8. Bajaj JS, Cox IJ, Betrapally NS, et al. Systems biology analysis of omepra- 27. Dultz G, Piiper A, Zeuzem S, et al. Proton pump inhibitor treatment
zole therapy in cirrhosis demonstrates significant shifts in gut microbiota is associated with the severity of liver disease and increased
composition and function. Am J Physiol Gastrointest Liver Physiol. mortality in patients with cirrhosis. Aliment Pharmacol Ther.
2014;307:G951–7. 2015;41:459–66.
9. Tandon P, Garcia-Tsao G. Bacterial infections, sepsis, and multiorgan 28. Bajaj JS, Heuman DM, Sanyal AJ, et al. Modulation of the metabiome by
failure in cirrhosis. Semin Liver Dis. 2008;28:26–42. rifaximin in patients with cirrhosis and minimal hepatic encephalopathy.
10. Qin N, Yang F, Li A, et al. Alterations of the human gut microbiome in PLoS ONE. 2013;8:e60042.
liver cirrhosis. Nature. 2014;513:59–64. 29. Acharya C, Sahingur SE, Bajaj JS. Microbiota, cirrhosis, and the
11. O’Leary JG, Reddy KR, Wong F, et al. Long-term use of antibiotics and emerging oral-gut-liver axis. JCI Insight. 2017;5: pii: 94416.
proton pump inhibitors predict development of infections in patients with 30. Bajaj JS, Betrapally NS, Hylemon PB, et al. Salivary microbiota reflects
cirrhosis. Clin Gastroenterol Hepatol. 2015;13:753–9. changes in gut microbiota in cirrhosis with hepatic encephalopathy.
12. Bajaj JS, Reddy KR, Tandon P, et al. The 3-month readmission rate Hepatology. 2015;62:1260–71.
remains unacceptably high in a large North American cohort of patients 31. Bert F, Valla D, Moreau R, et al. Viridans group streptococci causing spon-
with cirrhosis. Hepatology. 2016;64:200–8. taneous bacterial peritonitis and bacteremia in patients with end-stage
13. Nava GM, Stappenbeck TS. Diversity of the autochthonous colonic liver disease. Liver Transpl. 2008;14:710–1.
microbiota. Gut Microbes. 2011;2:99–104. 32. Ahluwalia V, Betrapally NS, Hylemon PB, et al. Impaired gut-liver-brain
14. Patel VC, Shawcross DL. Salivary microbiota-immune profiling in axis in patients with cirrhosis. Sci Rep. 2016;6:26800.
cirrhosis: could this be the noninvasive strategy that will revolutionize 33. Bajaj JS, Ridlon JM, Hylemon PB, et al. Linkage of gut microbiome with
prognostication in hepatology? Hepatology. 2015;62:1001–3. cognition in hepatic encephalopathy. Am J Physiol Gastrointest Liver
15. Bajaj JS, Heuman DM, Hylemon PB, et al. Altered profile of human gut Physiol. 2012;302:G168–75.
microbiome is associated with cirrhosis and its complications. J Hepatol. 34. Kedika RR, Souza RF, Spechler SJ. Potential anti-inflammatory effects of
2014;60:940–7. proton pump inhibitors: a review and discussion of the clinical implica-
16. http://www.accessdata.fda.gov/drugsatfda_docs/label/2008/022056s00101 tions. Dig Dis Sci. 2009;54:2312–7.
9810s087lbl.pdf.

© 2018 the American college of Gastroenterology The American Journal of GastroenteroloGy