Alimentary Pharmacology & Therapeutics
Vitamin B12 deficiency in hypersecretors during long-term acid
suppression with proton pump inhibitors
B. I. HIRSCHOWITZ, J. WORTHINGTON & J. MOHNE N
Division of Gastroenterology and
Hepatology, University of Alabama
School of Medicine, Birmingham, AL,
USA
Correspondence to:
Dr B. I. Hirschowitz, Division of
Gastroenterology and Hepatology,
University of Alabama School of
Medicine, LHR 406, UAB Medical
center, 1530 3rd Ave. South,
Birmingham, AL 35294, USA.
E-mail: bih@uab.edu
Publication data
Submitted 24 October 2007
First decision 4 December 2007
Resubmitted 14 February 2008
Second decision 17 February 2008
Resubmitted 19 February 2008
Accepted 20 February 2008
Epub OnlineAccepted 27 February
2008
1110
SUMMARY
Background
Proton pump inhibitors (PPIs) may cause cyanocobalamin (vitamin B12)
malabsorption, but measuring serum B12 alone may underestimate the
prevalence. However, B12 deficiency elevates methylmalonic acid and
homocysteine, both additional markers of B12 deficiency.
Aim
To determine the true prevalence of B12 deficiency and whether acid
suppression by PPI caused it.
Methods
Sixty-one acid hypersecretors (basal acid output >15 mmol ⁄ h), 46 with
gastrinoma [Zollinger–Ellison (ZE) syndrome] and 15 without [acid
hypersecretor without gastrinoma (pseudo-ZE)], were treated with lan-
soprazole to determine its long-term (up to 18 years) pharmacological
and clinical efficacy and safety, particularly as regards malabsorption of
B12.
Results
Of 61 patients, six (10%) had low serum B12. Additional tests uncov-
ered B12 deficiency in 13 (31%) of 41 still-available patients, despite
normal serum B12. B12 replacement reduced elevated homocysteine
and methylmalonic acid, supporting the diagnosis. Also, measuring both
basal and stimulated gastric secretion, we found that acid suppression
was neither prolonged nor profound enough to explain the B12
deficiency.
Conclusions
In long-term recipients of PPIs, B12 deficiency was more frequent
(29%) than detected by measuring only serum B12, and there was not
enough acid suppression to explain this deficiency.
Aliment Pharmacol Ther 27, 1110–1121
ª 2008 The Authors
Journal compilation ª 2008 Blackwell Publishing Ltd
doi:10.1111/j.1365-2036.2008.03658.x
 V I T A M I N B 1 2 W I T H L O N G - T E R M P P I F O R Z E S Y N D R O M E 1111
INTRODUCTION
Having established the long-term efficacy of lansop-
razole in acid hypersecretors with gastrinoma
[Zollinger–Ellison (ZE) syndrome] or without [acid
hypersecretor without gastrinoma (pseudo-ZE)]
through the closely monitored control of acid secre-
tion1 and the ensuing benefit for clinical outcome,2 we
now examine potential adverse effects of prolonged
acid suppression and, specifically, the possibility of
cyanocobalamin (vitamin B12) malabsorption.3–5 Low
serum vitamin B12 levels have been reported in some
patients treated with omeprazole,3–5 including those
with ZE5 and gastro-oesophageal reflux disease
(GERD)6, 7 treated long-term.
Mild or moderate B12 deficiency is frequently undi-
agnosed,8–12 as many may have normal serum B12
with no typical manifestations of deficiency. Hence,
there is a need for additional sensitive tests. One meta-
bolic consequence of B12 deficiency is the impairment
of methylmalonyl CoA isomerization to succinyl
CoA with accumulation of methylmalonic acid
(MMA).8, 13, 14 A second effect is a block of methionine
synthetase (methylation) in the conversion of homocy-
steine (Hcy) to methionine with (reversible) hyperhom-
ocysteinemia,11, 13, 15 which, in turn, is a marker for
this deficiency. These effects occur in patients with
even mild B12 deficiency. Therefore, measurement of
MMA and Hcy has provided alternative, sensitive indi-
cators8, 11–16 of the deficiency. In at-risk populations,
such as the elderly9, 12, 17, 18 and those who have had
gastric surgery,13 these tests have uncovered signifi-
cantly more instances of B12 deficiency than could be
diagnosed by serum B12 alone. In confirmation,
reversibility of elevated Hcy and MMA by B12 replace-
ment is a further proof of the deficient state.15
On the premise that mild deficiency was being over-
looked, we sought to determine more accurately the
prevalence of B12 deficiency in very long-term proton
pump inhibitor (PPI) treatment of acid hypersecre-
tors,1, 2, 5 first serially measuring serum B12 and, later
in the course of the study, also measuring MMA and
Hcy. These additional tests had not been used in prior
studies of B12 malabsorption in omeprazole-treated
patients.3–7
Moreover, B12 deficiency may not be a trivial mat-
ter, as even mild deficiencies may result in elevation
of Hcy,12, 13, 15 and in many epidemiological studies,
Hcy elevation has been described as having adverse
consequences,19 notably on the cardiovascular
ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121
Journal compilation ª 2008 Blackwell Publishing Ltd
system20, 21 and cognition.18, 22 However, recent
experimental studies have cast some doubts on these
conclusions.21, 23–25
Absorption of dietary B12 is a multistep process26
which may be subject to disruption at one or more
key steps and, where found with PPI treatment, has
been ascribed to impaired release of food-bound B129
because of excessive acid suppression. However, gas-
tric acid output, when measured in other studies,3–7
was reported in the basal state [basal acid output
(BAO)] only, without stimulation, leaving uncertain
the true rates of achlorhydria and the explanation of
this mechanism of the PPI effect on B12 absorption.
Together, these uncertainties prompted us to re-
examine in detail the problem and its implications in
both ZE and pseudo-ZE patients whose acid hyper-
secretion was being treated continuously with lansop-
razole for long periods with carefully controlled
gastric acid inhibition.
MATERIALS AND METHODS
In a continuing Institutional Review Board-approved
protocol, started in 1990 to study the long-term phar-
macological and therapeutic efficacy and safety of
lansoprazole, 61 acid hypersecretor patients (46 ZE, 15
pseudo-ZE) were treated with individually optimized
doses of lansoprazole (median dose 75, range 15–
450 mg ⁄ day) to good effect.1, 2 Duration of prior
omeprazole treatment for 17.3  3 months at a med-
ian dose of 40 mg ⁄ day was taken into account in cal-
culating the total duration of PPI use. All patients
were followed up throughout by the same physician
and support staff.
Upper gastrointestinal endoscopy and gastric analy-
sis were performed every 6 months after starting lan-
soprazole, together with comprehensive routine lab
analyses, including fasting serum gastrin, complete
blood count (CBC) with indices and platelet counts.
Fasting gastric analysis was performed in the two
hours before the next scheduled dose of lansoprazole.
Each study comprised six consecutive 10-min aspira-
tion periods for 1 h of basal secretion followed for 1 h
with stimulation by 6 lg ⁄ kg1 pentagastrin (which
accurately reflects maximal acid output with food27)
or more recently, as pentagastrin has regrettably
become unavailable in the US, by modified sham feed-
ing (vagal stimulation).28 In each sample, volume, pH,
titratable acid and pepsin were measured. The dose of
lansoprazole was individually adjusted as needed to
1112 B . I . H I R S C H O W I T Z et al.

maintain BAO at less than 5 mmol ⁄ h, but to avoid atrophy even in those few cases with Helicobacter
excessive acid suppression where maximum acid out- pylori infection.30
put (MAO) was also =0. These data were used to deter-
mine whether, and to what extent, acid secretion was
Biochemical methods
sufficiently suppressed to pH above 3.5, i.e. above the
range of peptic activity,29 to interfere with the release Unless otherwise stated, all tests were performed at the
of food-bound B12. We have further defined achlor- University of Alabama, Birmingham Special Chemistry
hydria, for the purposes of this study, as pH above 5.0 Laboratory.
– i.e., outside the range of any kind of intra-gastric
proteolysis, and above the bactericidal level.
Gastrin. Fasting serum gastrin was measured by the
IMMULITE 2000 gastrin kit (DPC; Siemens, Amsterdam,
Defining B12 deficiency the Netherlands). The reference values from 143 normal
persons yielded a median of 32 pg ⁄ mL and a nonpara-
One aim of this study was to determine to what extent
metric 95% range of 13–115 pg ⁄ mL with a CV of 9%.
prolonged acid suppression caused B12 absorption.
Since 1997, after a mean period of 40.3  4.8 months
(Supplementary material) on PPI, serum B12 as well as Serum Vitamin B12. B12 was measured by the Bayer
serum and erythrocyte (RBC) folate were measured at ADVIA Centaur VB12 competitive immunoassay by
least annually. No B12 studies were performed before direct chemiluminescence technology (Siemens, Terry-
treatment with PPI, as very few patients had not town, NY, USA), with a normal range of 211–
already been treated with PPI at the time of enrolment 911 pg ⁄ mL (median 382) with a CV of 3.8–5.9% at
into the study. In addition, since 2004, after PPI treat- levels in the normal range.
ment for 109  9.5 months, we also measured serum
concentrations of both MMA and Hcy as additional
Serum and erythrocyte (RBC) folate. Serum and
markers of possible B12 deficiency8, 11–16 in the 41
erythrocyte (RBC) folate were measured using the
patients who were still active in the programme and
Bayer ADVIA Centaur Folate kit (Siemens), with a nor-
who all had normal B12 levels. Total median duration
mal range for serum folate of 5.4–24 ng ⁄ mL (median
of PPI treatment was 140 (6–247) months (Supplemen-
12.5) and RBC folate of 280–791 ng ⁄ mL with a CV of
tary material).
8%. Minimum detection level was 0.35 ng ⁄ mL.
B12 deficiency was diagnosed on the basis of at least
one measurement of serum B12 below the lower limit
of normal (211 pg ⁄ mL). Furthermore, elevated levels of Methylmalonic acid. MMA was measured by Quest
both Hcy and MMA were considered to be likely mark- Diagnostics (San Juan Capistrano, CA, USA) with a
ers for B12 deficiency. After excluding folate normal range of 53–376 nmol ⁄ mL.8, 11
deficiency, elevated Hcy – even with normal MMA –
suggested possible B12 deficiency. This possibility was
Homocysteine. Hcy was measured using the IMMU-
tested by the response to treatment with B12 (1 mg by
LITE 2000 kit, validated in 120 volunteers aged 22–
IM injection weekly for 4 weeks) and re-tested in the
66 years, with a central 95% range of 5–12 lmol ⁄ L
next 4 weeks. Where elevated MMA was normalized,
(median 7.7), CV of 6–8%. Our values were not age-
the diagnosis was considered confirmed.15 Where Hcy
or gender-adjusted.
was elevated, but with normal MMA and normalization
of Hcy by B12, as well as a decrease in MMA by 40%
or more, the diagnosis was considered probable. Failure Pepsin. Pepsin was measured by haemoglobin digestion
of B12 to lower MMA and Hcy were taken as negative: at acid pH.28 Ranges for pepsin output in our lab under
two cases were rejected. various conditions have been reported elsewhere.1, 28
We did not measure antibodies to parietal cells,
H+K+ATPase, or intrinsic factor (IF), as these would
Statistics
not be expected in patients with ZE.5 Biopsies of the
gastric body mucosa showed abundant healthy-look- Categorical data are expressed as counts and analysed
ing parietal cells in all cases, with no evidence of using Fisher’s exact test or a chi-squared test. Normally

ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121

Journal compilation ª 2008 Blackwell Publishing Ltd
 V I T A M I N B 1 2 W I T H L O N G - T E R M P P I F O R Z E S Y N D R O M E 1113

distributed continuous data are expressed as the
mean  S.E.M. and analysed using t-tests and analysis
of variance. Data not normally distributed are
expressed as median and analysed using the Wilcoxon
rank-sum test. A P-value of <0.05 was considered sta-
tistically significant. All analyses were performed using
SAS statistical software, version 9.0 (Cary, NC, USA).
RESULTS
Diagnosing B12 deficiency
Measures to detect deficiencies in individual patients
were applied in two ways:
(i) Serial serum B12 levels in 61 patients prior to
receiving B12 during long-term acid inhibition
remained at comparable levels for about 8 years of
continuous PPI use and then declined progressively
but still remained within normal range (Figure 1), from
a median of between 890 and 950 pg ⁄ mL in the first
6 years, to 495 pg ⁄ mL, a drop of 46%, after 12 years
(P < 0.02 for overall trend).
Six (10%) patients had low (140–200 pg ⁄ mL) serum
B12 levels on one or more occasion, but without overt
symptoms or signs of B12 deficiency.11, 15, 31 They
were treated with B12, 1 mg IM every 6 months and
with 1 mg ⁄ day orally. Serum B12 remained normal on
follow-up, except for one who relapsed four times over
8 years, most recently with marginal B12 (256 pg ⁄ mL)
had elevated Hcy (12.7 lmol ⁄ L) and MMA
(630 nmol ⁄ mL), which fell to 260 with B12 treatment
(see below). He was counted only once in group A.
Three of the other five patients with low serum B12
died before Hcy or MMA levels were measured.
(ii) Hcy and MMA. Forty-one patients with normal
B12 levels were available for measuring Hcy and
MMA levels, in addition to B12 levels. Hcy and MMA
were first measured from 4 to 227 months (median
120) after starting PPI treatment and followed for
22 months. Of these 41 patients, 15 had elevated Hcy
levels, but normal serum B12 and normal RBC folate.
Ten of these also had elevated MMA levels – high-
normal (360 nmol ⁄ mL in one) and abnormal levels
(>390 nmol ⁄ mL) in the other nine (Figure 2).
Confirmation by B12 administration. The significance
of elevated Hcy and the confidence in diagnosing defi-
ciency were further tested by treatment with B12
(1 mg ⁄ week by injection for 4 weeks). Folic acid was
not given. Serum B12 rose from 302 to 1004 pg ⁄ mL.
Hcy fell to normal in thirteen of 15 patients, while
MMA fell to normal in nine of 10 who had both MMA
and Hcy elevation (Figure 2). The 10th patient, who
had chronic renal failure and diabetes with elevated
Hcy and MMA, failed to respond and was rejected.

2000

1800 Median

1600
Serum B12 (pg/mL)

1400

1200

1000

800

600
Figure 1. Progression of serum
B12 levels in patients at vari- 400
ous durations of proton pump
inhibitor treatment. All values 200
after B12 replacement have
been excluded. Overall trend 0
Total
analysis P = 0.02 (data not
years on PPI <3 3 to <6 6 to <9 9 to <12 ≥12
normally distributed, analysis
Number
on median values). Time of of patients 25 29 35 31 25
finding low serum B12 levels
Median 890 947 866 744 495
in six individual patients Number
noted. with B12<200 1 2 1 1 3

ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121

Journal compilation ª 2008 Blackwell Publishing Ltd
1114 B . I . H I R S C H O W I T Z et al.
Figure 2. Individual values of serum B12, homocysteine and methylmalonic acid before and after B12 treatment in 13
patients presumed to be B12-deficient on the basis of response to B12 treatment.
One to four weeks after receiving B12, MMA had
fallen by more than 40% in four of five patients who
had initially normal MMA levels but elevated Hcy. In
all five the Hcy had fallen to normal. The diagnosis
of B12 deficiency was considered to be probable in
these four patients with decreases in both Hcy and
MMA.
After B12 treatment, individual (Figure 2) and mean
values of B12, Hcy and MMA were significantly differ-
ent from each other and now were comparable to con-
trols (Table 1).
Prevalence of B12 deficiency
On the basis of the apparently specific Hcy and MMA
responses to B12 treatment,11, 15 we concluded that 13
of 41 patients (31%) were B12 deficient. The diagnosis
was considered definite in nine (including one with a
prior low serum B12) and probable in another four
with normal MMA but elevated Hcy, both responsive
to B12 (see above). Counting these four patients, the
total number of B12-deficient patients, including the
six with low serum B12 levels, was 18 of 61, or 29%).
The 15 patients in group C (Table 2) could possibly
have included additional cases of deficiency.
Characteristics of the patient population
Having determined B12 status, patients were divided
into those who were probably deficient and those who
were not and described in Table 2 under three
headings: group A – the 18 B12-deficient patients
described above; group B comprises 26 patients who
had normal serial B12, as well as normal Hcy and
MMA, and two who failed to respond appropriately to
B12 administration. That left 15 patients (group C),
considered separately here, who had had normal B12
levels but had not had Hcy or MMA measured. The 18
B12-deficient patients in group A had normal haemo-
globin, mean corpuscular volume and RBC folate
(Table 2). Also, demographic data, duration of PPI
treatment, pretreatment acid output and lansoprazole
doses were not different from those who were not
deficient (Table 2).
Explaining B12 deficiency: gastric secretion and
confounders
The putative explanation for B12 deficiency during
PPI therapy would be excessive postprandial acid
suppression.5, 6
ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121
Journal compilation ª 2008 Blackwell Publishing Ltd
 V I T A M I N B 1 2 W I T H L O N G - T E R M P P I F O R Z E S Y N D R O M E 1115

Table 1. B12, homocysteine and methylmalonic acid serum levels (mean  S.E.) (in 13 additional B12-deficient patients)
before and after B12 treatment, and in control group B

B12 deficient and treated (n = 13) Not B12 deficient (n = 28)

P-values

Variable Before treatment I After treatment II Group B patients III I vs. II II vs. III I vs. III

B12 (pg ⁄ mL) 382  29 1004  147 618  102 0.001 0.04 0.015
Homocysteine (lmol ⁄ L) 15.6  1.1 10.4  1.0 7.6  0.4 0.002 0.003 <0.001
MMA (nmol ⁄ mL) 432.3  81.2 192.7  30.0 162  11.6 0.004 0.346 <0.001

For individual values see Figure 2.

MMA, methylmalonic acid.

Table 2. Comparison of three Not B12 deficient

groups: A (B12 deficient), B
(not B12 deficient) and C (not Homocysteine Homocysteine
B12 deficient by serum B12 B12 deficient measured not measured
levels, but homocysteine and Variable mean  S.E. A (n = 18) B (n = 28) C (n = 15)
methylmalonic acid were not
measured)
Age (years) 56.3  2.9 53.8  2.3 52.9  3.6
Race (B:W) 3:15 5:23 6:9
Gender (M:F) 14.4 17:11 10.5
Diagnosis (ZE:pseudo) 15.3 19.9 12.3
Gastric resection 3 3 1
Chronic pancreatitis 1 0 1
Baseline BAO (mmol ⁄ h)  18.0  2.2 20.8  3.1 22.0  2.4
Lansoprazole (mg ⁄ day) (median) 60 90 75
B12 (pg ⁄ mL) 370.6  40.8 585.3  100.4 556.6  95.3
Haemoglobin (g ⁄ dL) 14.0  0.4 14.3  0.3 12.6  0.1
Mean corpuscular volume (fL) 87.1  2.2 89.8  1.1 88.9  0.2
RBC folate (ng ⁄ mL) 611  46 640  32 na

ZE, Zollinger–Ellison syndrome; BAO, basal acid output.

  Before lansoprazole treatment.

(Table 3 and Supplementary material). Separately, the

Confounders. In this population, there were other
possible causes for B12 deficiency including gastric
resection (antrectomy)13 in seven ZE patients, three of
whom were B12 deficient, but four others were not.
Another possible confounder is chronic pancreatitis:
one of two such patients was also B12 deficient. The
data were analysed with and without the confounders
and in both cases reached the same conclusion – that
28% of the 52 patients without confounders were B12
deficient – almost exactly the same number (29%) as
when using all 61 patients including the confounders.
nine confounders were comparable to the patients with
intact stomachs, both demographically and with
respect to BAO under treatment (2.2 vs. 2.2 mmol ⁄ h)
(Table 2 and Supplementary material).
Acid secretion. Taking the population as a whole in
the first instance, irrespective of possible confounders,
we analysed several aspects of gastric secretion:
These data were obtained in patients on stable, opti-
mal maintenance doses of lansoprazole.1, 2 The data

ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121

Journal compilation ª 2008 Blackwell Publishing Ltd
1116 B . I . H I R S C H O W I T Z et al.

lansoprazole.1 pH was measured under basal and sti-

Table 3. Mean  S.E.M. of basal and stimulated acid and
pepsin output in B12-deficient patients (A) and in those mulated conditions (pentagastrin or vagal stimulation
not deficient (B) by sham feeding) (Figure 3). These data provide an
overview of the pattern of acid secretion in hyper-
A: B12 B: not B12
secretors under optimal treatment with PPI.
deficient deficient
(n = 18) (n = 28) P-value (ii) Acid and pepsin output on lansoprazole, accord-
ing to B12 status. The mean value of individual
Acid (mmol ⁄ h) patient means – was not different in either the basal
BAO 2.10  0.43 2.21  0.34 0.86 or pentagastrin-stimulated state between B12-deficient
MAO 7.98  1.78 12.74  1.76 0.10 (group A) and nondeficient patients (group B) (Table 3).
Pepsin (PU 1000 ⁄ h) BAO prior to lansoprazole treatment was also not sig-
BPO 99.9  17.8 105.2  16.4 0.83
nificantly different (Table 2).
MPO 310.7  78.5 320  35.6 0.90
(iii) Peptic digestion to release food-bound B12 in
Each patient is represented by the mean of all the tests per- the stomach requires a pH of <3.5.29 Other gastric pro-
formed on that person during lansoprazole treatment. teolysis can occur at pH 3.5–5.0. In both deficient
BAO, basal acid output; BPO, basal pepsin output; MPO, (group A) and not deficient (group B) patients, almost
maximum pepsin output. 60% of basal samples had pH below 3.5 (Table 4) and
median pH was under 3, i.e. within the range of opti-
mum pepsin activity; 33% in both groups had pH
include pH, acid measured by titration, and pepsin in below 2.0. There was a small but significantly lower
each of approximately 9000 10-min samples of gastric response to pentagastrin stimulation in group A, but
aspirate from approximately 750 semiannual gastric even so, 73% of patients in A and 84% in B secreted
analyses, each comprising 1-h basal and 1-h stimulated acid to pH <3.5. Overall, pentagastrin failed to reduce
periods. The data were analysed in different ways. pH to below 5.0 (Table 4) in 17.5% and 8.8% of indi-
(i) Distribution of pH values, irrespective of B12 vidual samples in groups A and B respectively.
status, was obtained by using each 10-min sample in (iv) Basal achlorhydria and results of stimulation.
all the gastric analyses from 61 patients receiving Basal achlorhydria (BAO ¼ 0), in which all six 10-min

100

90

80
Cumulative incidence (%)

70

60

50

40

30

20 Figure 3. Cumulative gastric

juice pH values during optimal
10
Achlorhydria Pepsin active
lansoprazole treatment. Indi-
0 vidual 10-min aspirates from
7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 over 750 gastric analyses in
all 61 patients, irrespective of
pH
B12 status, in the basal state
Basal (n = 4583) and after stimulation by pen-
Sham (n = 586) tagastrin or sham feeding
Pentagastrin (n = 3930) (vagal activation).

ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121

Journal compilation ª 2008 Blackwell Publishing Ltd
 V I T A M I N B 1 2 W I T H L O N G - T E R M P P I F O R Z E S Y N D R O M E 1117

Taking into account all the data, the stomach con-

Table 4. Gastric juice pH values in groups A (B12 defi-
cient) and B (not deficient, under basal conditions and tents would have been exposed to a pH of <3.5 almost
when stimulated by pentagastrin or vagally (sham feed- 60% of the time in the basal state, approximately 80%
ing) with maximal stimulation and, even with the less
potent vagal (sham feeding) stimulation alone, without
No. of Median % % between %
Group observations pH £3.5 3.5 and 5 ‡5.0 food, 60–68% of the time (Table 4). In short, acid sup-
pression was neither profound enough nor prolonged
Basal enough to explain the B12 malabsorption.
A 1678 2.8 56.4 12.9 30.7
B 2905 2.9 56.9 13.2 29.9
P-value 0.99 0.32 0.76 0.58 Consequences of B12 deficiency, irrespective of
Pentagastrin cause
A 1496 1.85 72.9 9.6 17.5
B 2434 1.55 83.5 7.7 8.8 There were no clinically obvious manifestations of
P-value 0.001 1.55 0.03 <0.001 B12 deficiency despite the decline in serum B12 after
Sham feeding an average of about eight or more years of PPI. How-
A 183 2.6 60.1 16.4 23.5 ever, elevation of Hcy does pose a risk for cardiovas-
B 403 2.2 67.5 9.4 23.1 cular and other diseases.9, 12, 18, 31 In this study, the 15
P-value 0.03 0.09 0.01 0.91
patients with elevated Hcy comprised a relatively small
sample; analysis for such diseases would lack statisti-
These data represent all individual 10-min samples of gastric
juice while taking optimized lansoprazole doses. cal power and is not presented here.

samples have pH >5.0, was found in 18% of 234 gas-
tric analyses in group A patients and in 22% of 434
studies in group B (Table 5). In the subset of tests with
basal achlorhydria, pentagastrin stimulated acid secre-
tion to a pH of <5.0 in 76% of tests group A and in
90% of group B (P < 0.02). Those who failed to secrete
any acid, with pH above 5.0 in both basal and pentag-
astrin-stimulated states, i.e. complete therapeutic
achlorhydria, accounted for only 4.3% of all gastric
analyses in group A and 2.3% in B, not significantly
different and showing again that basal achlorhydria
alone may not reflect the true secretory status of the
stomach that would be expected with food.27
DISCUSSION
PPIs are highly effective in the treatment of acid-
related disorders and generally require long-term use.
Despite the marked benefits, there remain questions
regarding possible risks posed by prolonged suppres-
sion of gastric acid secretion.9, 12, 18, 31, 32 These are
extensively reviewed by Jensen.32 These risks include
malabsorption of vitamin B12.3–7, 32 Short-term (1–2
weeks) omeprazole in healthy volunteers,4 and long-
term in GERD,6, 7 may reduce B12 uptake. In the only
study of B12 in long-term omeprazole for ZE5 (mean
4.5 years), 6% of 111 patients had low serum B12 lev-
els (102–188 pg ⁄ mL), particularly affecting those with
apparent therapeutic prolonged basal achlorhydria. No

Table 5. Complete gastric pH after stimulation

analyses in groups A and B, Column 1 when BAO = 0 [n (%)
with analysis of the subset Total no. Column 2 of column 2]
with basal achlorhydria in of tests No. of tests BAO, MAO = 0* n (%)
response to pentagastrin Group n (%): with BAO = 0 pH £5.0 pH ‡5.0 % of column 1
stimulation
A 234 42 (18%) 32 (76%) 10 (24%) 10 (2.3%)
B 434 94 (22%) 84 (89%) 10 (11%) 10 (4.3%)
P-value 0.018 0.045 0.152

BAO, basal acid output.

* Total therapeutic achlorhydria, i.e. with both BAO = 0 and MAO = 0 (pH >5.0).

ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121

Journal compilation ª 2008 Blackwell Publishing Ltd
1118 B . I . H I R S C H O W I T Z et al.
patient in that study had clinical B12 deficiency, and
none had IF antibodies.
Two problems with that report5 are: first, the diag-
nosis of B12 deficiency, based as it was only on serum
B12 levels, underestimates the prevalence,9–11, 13, 15
and second, the definition of achlorhydria was based
solely on BAO without added or confirmatory stimula-
tion, resulting in an overestimate of achlorhydria.
Diagnosis of B12 deficiency
As subclinical B12 deficiency may be found with nor-
mal serum B12 levels,10, 11, 13, 17 diagnosing B12 defi-
ciency based solely on B12 levels leaves many cases
unsuspected. Of the two markers of B12 deficiency,
Hcy and MMA elevation, the more specific is
MMA.8, 13, 14 Hcy is less specific, as it may also result
from folic acid deficit. But that was excluded here in
all cases of Hcy elevation by normal serum and eryth-
rocyte folate levels. Thus, reversing elevated Hcy and
MMA by treating with B12, was considered to be spe-
cific and robust evidence for B12 deficiency.
Assuming that this reasoning was valid, we found a
higher-than-reported prevalence of B12 deficiency in
PPI-treated acid hypersecretors. In our study, six
(10%) of 61 patients had low serum B12 levels, but
further 13 of 41 patients (31%), with normal serum
B12 were diagnosed as presumably deficient on the
basis of Hcy and MMA responses to B12 replace-
ment.8–12 If we exclude the nine patients with antrec-
tomy or pancreatitis, there would be 14 in group A, 25
in B and 13 in C. Thus, 14 of 52 (28%) were deficient
in the overall remaining population without counting
confounders. Without the exclusion almost the same
proportion, 18 ⁄ 61 or 29.5% is found. There seemed to
be no good reason to exclude possible confounders
from calculating the estimates of prevalence in this
population.
PPI and B12 deficiency
B12 absorption begins with peptic cleavage in the
stomach, at acid pH of food-bound B12, which then
binds to salivary R protein.26, 33, 34 Next, in the duode-
num, pancreatic enzymes release R-protein-bound
B12, which then binds to IF. The B12–IF complex is
absorbed in the terminal ileum after binding to its
receptor, cubilin.35 Absorbed B12 is cleaved from IF
by cathepsin L and then transported in the blood
bound to transcobalamin II.26 An enterohepatic cycle
promotes the conservation of B12.36 These multiple
steps represent a number of choke points at which the
process of B12 uptake could be disturbed.
Assuming that pepsin is the principal proteolytic
enzyme responsible for the first step, B12 deficiency
associated with PPI treatment3, 4 could result from acid
suppression sufficient to raise pH above 3.5 after
meals, rendering pepsin ineffective.29 However, in our
patients, with underlying acid and pepsin hypersecre-
tion, the exposure to an appropriate acid environment
(pH <3.5), with lansoprazole in the unstimulated stom-
ach, was well above 50% and presumably would be
adequate for food-B12 release in the stomach. Thus,
there was no good evidence of excessive or prolonged
acid suppression with the carefully controlled use of
lansoprazole1, 2 to account for the observed B12 defi-
ciency. Pentagastrin stimulation, with a reduction of
pH to below 3.5 in 83 to 91% of samples, is equivalent
to the stimulation expected from food.27 The case may
be different for PPI use in patients such as those with
GERD with underlying normal or low gastric acid
secretion.
In the duodenum, the B12 bound to salivary R pro-
tein4, 33, 34 is released through pancreatic enzyme pro-
teolysis at alkaline pH26, 37 and then binds to IF of
gastric origin. In man, IF is secreted by the gastric
parietal cell, probably by a mechanism different from
that of acid, as IF secretion, though reduced by vagot-
omy and histamine H2 receptor antagonist, remains
unaffected by omeprazole38 and presumably all PPIs,
including lansoprazole. A direct effect on IF is thus
unlikely.
Adverse conditions in the duodenum may hinder the
release of protein-R-bound B12 and thus prevent bind-
ing to IF. Gastric bypass for obesity or resection for
ulcer disease with jejunal anastomosis would not only
reduce acid and pepsin secretion, as well as the resi-
dence time of food in the stomach, but would also
bypass the duodenum, partly contributing to disrup-
tion of the transfer of B12 from protein R to IF.13
There were three such patients in group A, all with
active acid secretion but four other postantrectomy
patients were not B12 deficient. Chronic pancreatic
insufficiency37 resection as in 1 patient in group A
could interfere with the duodenal phase of B12
absorption13, 26 through reduction or loss of bicarbon-
ate and proteolytic enzymes.
In untreated ZE excess acid could inactivate pancre-
atic enzymes, resulting, among other effects, in B12
malabsorption.39 However, given the gastric secretion
ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121
Journal compilation ª 2008 Blackwell Publishing Ltd
 V I T A M I N B 1 2 W I T H L O N G - T E R M P P I F O R Z E S Y N D R O M E 1119
data under optimal lansoprazole control, this seems
unlikely.
Bacterial colonization
If gastric contents exceed the bactericidal pH 5.0
(which was rare in this study), oral and occasionally
colonic, bacteria may colonize (>105 organisms) the
small bowel40, 41 with resulting B12 sequestration. The
clinical significance of this is controversial40, 41, espe-
cially with relatively low bacterial populations that are
mostly of oral origin.40 Such colonization does not
result in significant either B12, fat, or carbohydrate
malabsorption – or, for that matter, in Hcy elevation.41
We did not investigate our patients for possible small
bowel overgrowth, but given the pH data, we would
not expect this to have been important here.
Consequences of B12 deficiency
Although mild B12 deficiency has been demonstrated
in at least one-fourth of our patients on long-term
PPI, the clinical significance is less obvious. It is by
no means clear, despite a downward trend after
8 years of continuous PPI treatment to a decline of
46% after 12 years, whether it would eventually pro-
gress to full-blown clinical B12 deficiency.
Beyond the well-known serious conse-
9, 12, 18, 31
quences of overt deficiency, there is also the
secondary risk that subtle B12 deficiency will elevate
serum Hcy. That, in turn, may promote the risk of var-
ious cardiac and vascular conditions,13, 20 altered cog-
nition12, 18 and other diverse effects.19
The largely epidemiological methodology for finding
the associations between Hcy elevation and cardiovas-
cular complications has been questioned,42 as has the
significance of Hcy elevation in recent specific experi-
mental tests of the hypothesis showing that reduction
of Hcy by vitamins B and B12 provided little cardio-
vascular,23, 24 cognitive25 or stroke21 benefit. However,
it should be noted that these interventions were in
patients who already had established vascular disease.
At best, these studies showed that reducing Hcy did
not reverse existing vascular disease.
Assuming that there is an increased relative risk and
that hyperhomocysteinemia is an early effect of subtle
B12 deficiency, this secondary effect could be both
clinically important and easily treated.11, 15, 23, 24 Sys-
tematic surveys should be conducted in populations at
risk.
ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121
Journal compilation ª 2008 Blackwell Publishing Ltd
CONCLUSIONS
(i) Prolonged PPI use was associated with a 46%
decline in median serum B12 levels and subnormal
levels in 10% of patients. Moreover, using additional
sensitive diagnostic tools, we uncovered subclinical
B12 deficiency in 31% of patients with normal B12
levels.
(ii) As we did not find meaningful prolonged or
excessive acid or pepsin suppression, this explanation
for the demonstrated B12 deficiency with PPI was
unconvincing. Other mechanism(s) whereby PPIs might
lead to the demonstrated B12 deficiency remain to be
elucidated.
(iii) Even taking into account the possible con-
founders, antrectomy and pancreatitis, did not alter
the proportion (at least 28%) of B12 deficiency in
patients undergoing prolonged PPI treatment.
(iv) Hyperhomocysteinemia caused by even mild
B12 deficiency may pose a potentially modifiable
additional risk for cardiovascular disease.
ACKNOWLEDGEMENTS
Declaration of personal interests: We also wish to
thank Jean Price, MT ASCAP, and Susan Irwin, RN,
for performing all gastric analyses, Dorothy Faulk
for maintaining the clinical database and Yanhui
Sun and Bonnie Agee for statistical assistance. We
are indebted to Drs Carlos Krumdieck, Mel C. Wilcox
and Naomi Fineberg, as well as Elizabeth Crane of
TAP, Inc., for critical and helpful reviews of the
manuscript, and Don Scott for expert copy editing.
We especially appreciate the willingness of our
patients to undergo the extensive tests required by
the demanding protocol. Declaration of funding
interests: This study was funded in full by TAP
Pharmaceuticals, Lake Forest, IL, USA (NCT
#00204373).
SUPPLEMENTARY MATERIAL
The following supplementary material is available for
this article:
Table S1. Duration of PPI treatment before and after
first measurement of serum B12, and later of Homo-
cystine and MMA.
Tables S2. (a,b) These tables compare the demo-
graphics and other relevant data with and without the
possible confounders – ZE with antrectomy and ⁄ or
1120 B . I . H I R S C H O W I T Z et al.

chronic pancreatitis. In either case the proportion of http://www.blackwell-synergy.com/doi/abs/10.1111/

B12-deficient patients remained the same (28% vs.
29%).
Table S3. Demographics and other relevant data in
the nine possible confounders – ZE with antrectomy
(ZEA) and ⁄ or chronic pancreatitis.
This material is available as part of the online article
from:
j.1365-2036.2008.03658.x
(This link will take you to the article abstract).
Please note: Blackwell Publishing are not responsi-
ble for the content or functionality of any supplemen-
tary materials supplied by the authors. Any queries
(other than missing material) should be directed to the
corresponding author for the article.

and treatment. Am J Med 2005; 118: 20 Nygard O, Nordrehaug JE, Refsum H,

REFERENCES
1 Hirschowitz BI, Simmons J, Mohnen J.
Long-term lansoprazole control of gas-
tric acid and pepsin secretion in ZE and
non-ZE hypersecretors: a prospective
10-year study. Aliment Pharmacol Ther
2001; 15: 1795–806.
2 Hirschowitz BI, Simmons J, Mohnen J.
Clinical outcome using lansoprazole in
acid hypersecretors with and without
Zollinger–Ellison syndrome: a 13-year
prospective study. Clin Gastroenterol
Hepatol 2005; 3: 39–48.
3 Howden CW. Vitamin B12 levels during
prolonged treatment with proton pump
inhibitors. J Clin Gastroenterol 2000;
30: 29–33.
4 Marcuard SP, Albernaz L, Khazanie PG.
Omeprazole therapy causes malabsorp-
tion of cyanocobalamin (vitamin B12).
Ann Intern Med 1994; 120: 211–5.
5 Termanini B, Gibril F, Sutliff VE, Yu F,
Venzon DJ, Jensen RT. Effect of long-
term gastric acid suppressive therapy on
serum vitamin B12 levels in patients
with Zollinger–Ellison syndrome. Am J
Med 1998; 104: 422–30.
6 Schenk BE, Festen HP, Kuipers EJ, Klin-
kenberg-Knol EC, Meuwissen SG. Effect
of short- and long-term treatment with
omeprazole on the absorption and
serum levels of cobalamin. Aliment
Pharmacol Ther 1996; 10: 541–5.
7 Koop H, Bachem MG. Serum iron, ferri-
tin, and vitamin B12 during prolonged
omeprazole therapy. J Clin Gastroenterol
1992; 14: 288–92.
8 Allen RH, Stabler SP, Savage DG, Lin-
denbaum J. Diagnosis of cobalamin
deficiency I: usefulness of serum meth-
ylmalonic acid and total homocysteine
concentrations. Am J Hematol 1990;
34: 90–8.
9 Andres E, Affenberger S, Vinzio S, et al.
Food-cobalamin malabsorption in
elderly patients: clinical manifestations
1154–9.
10 Carmel R. Subtle cobalamin deficiency.
Ann Intern Med 1996; 124: 338–40.
11 Savage DG, Lindenbaum J, Stabler SP,
Allen RH. Sensitivity of serum meth-
ylmalonic acid and total homocysteine
determinations for diagnosing cobala-
min and folate deficiencies. Am J Med
1994; 96: 239–46.
12 Stabler SP, Lindenbaum J, Allen RH.
Vitamin B-12 deficiency in the elderly:
current dilemmas. Am J Clin Nutr 1997;
66: 741–9.
13 Sumner AE, Chin MM, Abrahm JL, et al.
Elevated methylmalonic acid and total
homocysteine levels show high preva-
lence of vitamin B12 deficiency after
gastric surgery. Ann Intern Med 1996;
124: 469–76.
14 Stabler SP. Screening the older popula-
tion for cobalamin (vitamin B12) defi-
ciency. J Am Geriatr Soc 1995; 43:
1290–7.
15 Lindenbaum J, Savage DG, Stabler SP,
Allen RH. Diagnosis of cobalamin defi-
ciency: II. Relative sensitivities of serum
cobalamin, methylmalonic acid, and
total homocysteine concentrations. Am
J Hematol 1990; 34: 99–107.
16 Stabler SP, Lindenbaum J, Allen RH.
The use of homocysteine and other
metabolites in the specific diagnosis of
vitamin B-12 deficiency. J Nutr 1996;
126: 1266S–72S.
17 Wagner DA, Bolt DL, Curington CL, Cin-
tron M, Toskes PP. Comparison of meth-
ylmalonic acid to B12 serology and a
newly developed B12 breath test. Gas-
troenterology 2005; 128: A416.
18 Stabler SP. Vitamins, homocysteine, and
cognition. Am J Clin Nutr 2003; 78:
359–60.
19 Selhub J. The many facets of hype-
rhomocysteinemia: studies from the Fra-
mingham cohorts. J Nutr 2006; 136:
1726S–30S.
Ueland PM, Farstad M, Vollset SE.
Plasma homocysteine levels and mortal-
ity in patients with coronary artery
disease. N Engl J Med 1997; 337: 230–
6.
21 Toole JF, Malinow MR, Chambless LE,
et al. Lowering homocysteine in patients
with ischemic stroke to prevent recur-
rent stroke, myocardial infarction, and
death: the Vitamin Intervention for
Stroke Prevention (VISP) randomized
controlled trial. JAMA 2004; 291: 565–
75.
22 Seshadri S, Beiser A, Selhub J, et al.
Plasma homocysteine as a risk factor
for dementia and Alzheimer’s disease. N
Engl J Med 2002; 346: 476–83.
23 Lonn E, Yusuf S, Arnold MJ, et al.
Homocysteine lowering with folic acid
and B vitamins in vascular disease. N
Engl J Med 2006; 354: 1567–77.
24 Clarke R, et al. Homocysteine-lowering
trials for prevention of cardiovascular
events: a review of the design and
power of the large randomized trials.
Am Heart J 2006; 151: 282–7.
25 McMahon JA, Green TJ, Skeaff CM,
Knight RG, Mann JI, Williams SM. A
controlled trial of homocysteine lower-
ing and cognitive performance. N Engl J
Med 2006; 354: 2764–72.
26 Russell-Jones GJ, Alpers DH. Vitamin
B12 transporters. Pharm Biotechnol
1999; 12: 493–520.
27 Halter F, Keller M. A comparison
between intragastric titration and aspi-
ration technique under basal conditions
and after food or pentagastrin stimula-
tion. Am J Dig Dis 1978; 23: 723–9.
28 Hirschowitz BI, Groarke J. Vagal effects
on acid and pepsin secretion and serum
gastrin in duodenal ulcer and controls.
Dig Dis Sci 1993; 38: 1874–84.
29 Hirschowitz BI, Keeling D, Lewin M,
et al. Pharmacological aspects of acid
secretion. Dig Dis Sci 1995; 40: 3S–23S.

ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121

Journal compilation ª 2008 Blackwell Publishing Ltd

30 Hirschowitz BI, Haber MM. Helicobacter
pylori effects on gastritis, gastrin and
enterochromaffin-like cells in Zollinger–
Ellison syndrome and non-Zollinger–
Ellison syndrome acid hypersecretors
treated long-term with lansoprazole.
Aliment Pharmacol Ther 2001; 15: 87–
103.
31 Oh R, Brown DL. Vitamin B12 defi-
ciency. Am Fam Physician 2003; 67:
979–86.
32 Jensen RT. Consequences of long-term
proton pump blockade: insights from
studies of patients with gastrinomas.
Basic Clin Pharmacol Toxicol 2006; 98:
4–19.
33 Alpers DH. What is new in vitamin
B(12)? Curr Opin Gastroenterol 2005;
21: 183–6.
34 Del Corral A, Carmel R. Transfer of
cobalamin from the cobalamin-binding
protein of egg yolk to R binder of
human saliva and gastric juice. Gastro-
enterology 1990; 98: 1460–6.
ª 2008 The Authors, Aliment Pharmacol Ther 27, 1110–1121
Journal compilation ª 2008 Blackwell Publishing Ltd
V I T A M I N B 1 2 W I T H L O N G - T E R M P P I F O R Z E S Y N D R O M E 1121
35 Fedosov SN, Fedosova NU, Berglund L,
Moestrup SK, Nexo E, Petersen TE.
Composite organization of the cobala-
min binding and cubilin recognition
sites of intrinsic factor. Biochemistry
2005; 44: 3604–14.
36 Carmel R. Malabsorption of food cobal-
amin. Baillieres Clin Haematol 1995; 8:
639–55.
37 Allen RH, Seetharam B, Allen NC, Po-
dell ER, Alpers DH. Correction of cobal-
amin malabsorption in pancreatic
insufficiency with a cobalamin analogue
that binds with high affinity to R pro-
tein but not to intrinsic factor. In vivo
evidence that a failure to partially
degrade R protein is responsible for
cobalamin malabsorption in pancreatic
insufficiency. J Clin Invest 1978; 61:
1628–34.
38 Festen HP, Tuynman HA, Den Hollander
W, Meuwissen SG. Repeated high oral
doses of omeprazole do not affect
intrinsic factor secretion: proof of a
selective mode of action. Aliment Phar-
macol Ther 1989; 3: 375–9.
39 Shimoda SS, Saunders DR, Rubin CE.
The Zollinger–Ellison syndrome with
steatorrhea. II. The mechanism of fat
and vitamin B 12 malabsorption. Gas-
troenterology 1968; 55: 705–23.
40 Thorens J, Froehlich F, Schwizer W,
et al. Bacterial overgrowth during treat-
ment with omeprazole compared with
cimetidine: a prospective randomised
double blind study. Gut 1996; 39: 54–9.
41 Saltzman JR, Kowdley KV, Pedrosa MC,
et al. Bacterial overgrowth without clin-
ical malabsorption in elderly hypochlo-
rhydric subjects. Gastroenterology 1994;
106: 615–23.
42 Splaver A, Lamas GA, Hennekens CH.
Homocysteine and cardiovascular dis-
ease: biological mechanisms, observa-
tional epidemiology, and the need for
randomized trials. Am Heart J 2004;
148: 34–40.