GASTROENTEROLOGY 2013;145:1464–1478
AGA
American Gastroenterological Association Institute Technical Review on
the Use of Thiopurines, Methotrexate, and Anti–TNF-a Biologic Drugs
for the Induction and Maintenance of Remission in Inflammatory
Crohn’s Disease
C rohn’s disease (CD) is a chronic inflammatory bowel
disease (IBD) that causes significant morbidity and
represents a considerable burden to society and the health
Gastroenterological Association,8 and is becoming the
common methodology for the streamlined development
of clear, transparent, and actionable guidelines.9
care system.1–5 Based on the latest administrative data, it is An accompanying report10 in this issue of GASTROEN-
estimated that 300,000 to 500,000 Americans have CD.2,6 TEROLOGY integrates the results of this technical review
A recent study reported annual treatment costs of $8265 with the other GRADE criteria to produce a set of
per patient, which extrapolates to yearly costs of $2.5 to $4 recommendations.
billion for the American population with CD.3
Two main principles guide the medical therapy of patients
Methods
with CD. First, because this is a lifelong, relapsing disorder,
therapy to induce remission (inductive therapy) is followed Overview
by therapy to maintain remission (maintenance therapy).7 This technical review (and the accompanying guideline) was
Second, the choice of inductive and maintenance therapies based on the GRADE framework. In developing this technical re-
depends on the severity of the disease and the response to less view, the authors first formulated a series of specific questions that
were to be answered by the guideline. The authors then identified
effective strategies. In this progressive approach to therapy,
the outcomes that were significant to answering each question and
mesalamine, antibiotics, and budesonide are used in patients rated them as critical or important. Next, the group systematically
with mild disease. Systemic corticosteroids, immunomodu- reviewed and summarized the evidence for each outcome across
lators, and anti–tumor necrosis factor (TNF)-a agents are studies, assessed the quality of evidence for each outcome, and
used in patients with moderately severe CD or in patients finally integrated the evidence across all the outcomes to answer
who fail to respond to therapy for mild disease. The immu- each specific question. The quality of the evidence was classified into
nomodulators include the thiopurine analogues, azathio- 4 categories: high, moderate, low, and very low. Assessment of the
prine (AZA) and 6-mercaptopurine (6-MP), and quality for each outcome took into account the study design, risk
methotrexate (MTX). Anti–TNF-a agents approved for use in of bias, inconsistency (or heterogeneity), indirectness, imprecision,
the United States include infliximab (IFX), adalimumab and potential publication bias (see the glossary of terms in
(ADA), and certolizumab pegol (CZP). Supplementary Methods for further explanations).
In this technical review, the American Gastroentero-
logical Association addresses the relative positioning of Outcomes of Interest
immunomodulators and anti–TNF-a biologic agents in Using the PICO format, which frames a clinical question
inducing and maintaining clinical remission in patients by defining a specific population (P), intervention (I), comparator
with inflammatory (luminal) CD. From the standpoint of (C), and outcome (O), we outlined a total of 16 PICO questions
(see Table 1). The patient population with moderate-severe active
patients and clinicians, selecting among immunomodu-
CD was defined as patients with a Crohn’s Disease Activity Index
lator monotherapy, anti–TNF-a monotherapy, and com- (CDAI) of 220 to 450. The population with CD in remission was
bination therapy is a common clinical dilemma. Providing defined as patients with a CDAI <150 and not being treated with
optimal, evidence-based care to the many patients who are
candidates for these potentially costly and/or toxic ther-
apies is of critical importance to the health care system as
Abbreviations used in this paper: ACG, American College of Gastro-
well. The Grading of Recommendations Assessment, enterology; ADA, adalimumab; AZA, azathioprine; BSG, British Society of
Development and Evaluation (GRADE) methodology was Gastroenterology; CD, Crohn’s disease; CDAI, Crohn’s Disease Activity
used in this technical review to assess the evidence on Index; CI, confidence interval; CZP, certolizumab pegol; ECCO, European
immunomodulators and anti–TNF-a-biologic agents in Crohn’s and Colitis Organisation; GRADE, Grading of Recommendations
AGA
Assessment, Development and Evaluation; HR, hazard ratio; IBD, in-
the (1) induction of remission in adult patients who have
flammatory bowel disease; IFX, infliximab; 6-MP, 6-mercaptopurine;
moderately severe inflammatory CD despite therapy with MTX, methotrexate; OR, odds ratio; PICO, population, intervention,
mesalamine, antibiotics, corticosteroids, and/or immuno- comparator, and outcome; RCT, randomized controlled trial; RR, relative
modulators and (2) maintenance of medically induced risk; SIR, standardized incidence ratio; TNF, tumor necrosis factor.
remission. GRADE has been adopted by several national © 2013 by the AGA Institute
0016-5085/$36.00
and international societies, including the American
http://dx.doi.org/10.1053/j.gastro.2013.10.046
,December 2013 AGA 1465
Table 1. PICO Questions
Population(s) Intervention(s) Comparator Outcome(s)
1 a. Adults with moderate-severe CD AZA or 6-MP Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
2 a. Adults with moderate-severe CD MTX Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
3 a. Adults with moderate-severe CD Anti–TNF-a Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections
and lymphoma
4 a. Adults with moderate-severe CD Thiopurine or Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission MTX + anti–TNF-a b. Disease relapse; AE: serious infections and lymphoma
5 a. Adults with moderate-severe CD Thiopurine MTX a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
6 a. Adults with moderate-severe CD Thiopurine or MTX Anti–TNF-a a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
7 a. Adults with moderate-severe CD Thiopurine or MTX + Thiopurine a. Induction of remission; AE: serious infections
b. Adults with CD in remission anti–TNF-a or MTX b. Disease relapse; AE: serious infections and lymphoma
8 a. Adults with moderate-severe CD Thiopurine or MTX + Anti–TNF-a a. Induction of remission; AE: serious infections
b. Adults with CD in remission anti–TNF-a b. Disease relapse; AE: serious infections and lymphoma
NOTE. The following PICOs were excluded from the technical review because of absent or insufficient data: 4a, 4b, 6b, and 7b. Moderate-severe CD
was defined as a CDAI from 220 to 450. Remission was defined as a CDAI <150 and not being treated with corticosteroid therapy. Disease relapse
was defined as a CDAI 150 or corticosteroid use or surgery.
AE, adverse events.
corticosteroids. The interventions were immunomodulatory We selected serious infections and lymphoma as important
monotherapy, anti-TNF monotherapy, or combination therapy. adverse events potentially associated with serious morbidity or,
The comparators were immunomodulatory monotherapy or rarely, mortality. Serious infection was defined as infection that
anti-TNF monotherapy. PICO questions for which there were led to hospitalization (the definition used by the majority of
either no data or insufficient data (eg, in adult patients with studies). These outcomes were considered important but not
moderate-severe CD, should combination therapy with immu- critical for decision making. Because lymphoma is usually asso-
nomodulators plus anti-TNF therapy versus placebo be used to ciated with long-term therapy, we assessed the risk of lymphoma
induce remission?) could not be addressed in this technical re- only during maintenance therapy.
view. Evidence profiles were used to display the summary esti-
mates as well as the body of evidence for each clinical question.
For randomized controlled trials (RCTs) evaluating induction, Literature Search
the efficacy outcome considered critical for decision making was Three separate literature searches were conducted: one
corticosteroid-free clinical remission, defined as a CDAI <150 or for evidence summaries (such as meta-analyses); one for RCTs for
a Harvey–Bradshaw Index <4. For RCTs evaluating maintenance, the efficacy, infection, and lymphoma outcomes; and one for
the critical efficacy outcome was disease relapse, defined as a observational evidence to supplement the data on infection and
CDAI 150, corticosteroid use, or surgery. lymphoma. An information specialist developed each search with
We excluded trials without validated end points, such as trials input from the project team. All search results were imported
only reporting subjective symptom improvement and trials using bibliographic management software for de-duplication
evaluating only corticosteroid sparing. With regard to studies of and title and abstract screening.
induction with immunomodulators, which are agents with a The following bibliographic databases were searched through
delayed onset of action, we included studies with at least 12 the Ovid interface: MEDLINE, MEDLINE In-Process & Other
weeks of therapy. Conversely, we excluded trials that evaluated Non-Indexed Citations, and EMBASE. Parallel searches included
remission after more than 26 weeks of immunomodulator the Cochrane Database of Systematic Reviews, Database of Ab-
therapy, because an agent cannot properly be considered stracts of Reviews of Effects, Cochrane Central Register of
inductive and clinically relevant if its onset of action occurs Controlled Trials (CENTRAL), Cochrane Methodology Register,
beyond 26 weeks. For studies evaluating maintenance, only and HTA Database. The search strategy comprised controlled
medically induced remission was evaluated (ie, we excluded trials vocabulary, including the National Library of Medicine’s MeSH
that evaluated maintenance of surgically induced remission). It (Medical Subject Headings), and keywords. The main search
should be noted that the thiopurine and MTX maintenance concepts included and combined were “Crohn disease” and
studies were conducted before the anti–TNF-a era (ie, the pa- “immunomodulator therapy” and “anti-tumor necrosis factor.”
tients achieved remission with corticosteroids, mesalamine, and/ Methodological filters were applied to limit retrieval to RCTs,
or antibiotics). Interventions were analyzed based on their ability meta-analyses, systematic reviews, and health technology assess-
AGA
to reduce an undesirable outcome: failure to achieve clinical ments. The results were limited to English, human, and 1995
remission (in induction trials) or failure to prevent disease onward. The second search consisted of the main search con-
relapse (in maintenance trials). Based on clinical judgment, we cepts “Crohn disease” and “immunomodulator therapy” and
considered a relative reduction of failure to achieve (or maintain) “anti-tumor necrosis factor” plus “lymphoma.” The results were
remission of 20% as the minimum clinically important difference limited to English language and 2010 onward, because prior sys-
for immunomodulators or anti–TNF-a agents when compared tematic reviews using appropriate search strategies had adequately
with placebo and 10% when comparing drug classes. covered earlier time frames. A search for observational evidence on
, 1466 AGA GASTROENTEROLOGY Vol. 145, No. 6
harm was performed from this search (see Supplementary
The following issues were identified for rating down the quality: representativeness of patient population unclear, consecutive patients not enrolled (convenience sample), outcome adjudication not
fewer to 24 more)e
Data on serious infections were obtained from the study by Lichtenstein et al, which evaluated risk across all immunomodulators (AZA/6-MP/MTX); however, in 95% of cases, this consisted of treatment with a
(AZA/6-MP) (95% CI)
fewer to 38 more)
Risk difference with
82 fewer per 1000
10 more per 1000
Methods for the detailed search strategies). Updated information
Anticipated absolute effects
thiopurines
on serious infection and lymphoma (The Crohn’s Therapy,
(from 182
Resource, Evaluation, and Assessment Tool [TREAT] registry)
(from 2
became available during the writing process and was thus included
in this review (Lichtenstein et al, manuscript under review).11,12
Based on these searches, we identified existing systematic re-
views and used AMSTAR,13 a validated instrument, to evaluate
Risk with
the quality of systematic reviews. Systematic reviews that were of
controls
628 per
1000
1000
high quality, were up-to-date, and used the aforementioned
46 per
outcomes of interest (eg, corticosteroid-free remission based on
CDAI) were selected for inclusion in the evidence profiles. When
systematic reviews were not up-to-date or were incomplete, we
effect (95%
Relative
(0.96–
(0.71–
OR, 1.23
1.57)d
RR, 0.87
performed our own meta-analysis (random effects model for 3 or
1.06)
CI)
more studies and fixed effects model for 2 studies) using the
thiopurine (AZA/6-MP). Serious infection was defined as “any infection reported as serious by the investigator and any infection that required hospitalization.”
Cochrane Collaboration’s RevMan 5.1 software.14
Harm data from RCTs were sometimes of inadequate quality
(AZA/6-MP)
thiopurines
or had very few events. When this occurred, we identified
Table 2. Should Thiopurines (AZA/6-MP) Versus Placebo Be Used for Adults With Active (Moderate to Severe) CD (CDAI 220–450)?
Study event rates (%)
146/1849
With
102/197
observational studies and assessed them for risk of bias using the
(51.8)
(7.9)
Newcastle–Ottawa tool.15 We selected the observational studies
with the highest methodological quality and greatest number of
events for the outcomes of serious infections and lymphoma.
Undetected 4222 Very 148/3255
115/183
control
(62.8)
With
(4.6)
Results
Induction of Remission
risk of bias,
Overall quality
imprecision
imprecision
lowc due to
Moderatea
Thiopurines versus placebo. Five RCTs in 380 of evidence
Undetected 4442
due to
patients (thiopurine, 197; placebo, 183) compared AZA or
6-MP with placebo for the induction of remission.16–20 All
studies were blinded and likely achieved allocation
Failure of remission (critical outcome; remission assessed with CDAI <150 or Harvey–Bradshaw Index <4)
concealment. The patient populations differed with respect
Publication
to disease duration and location. The majority of patients
bias
had ileal or ileocolonic disease for at least 3 to 4 years and
had received prior medical therapy and/or had undergone
prior surgery. Except for one study that used 6-MP at a low
Imprecision
dosage of 50 mg/day,18 all studies examined the use of AZA
Seriousa
Seriousc
at a dosage of 2.5 mg/kg/day. Tapering doses of cortico-
steroids were given concomitantly in both the placebo and
AZA/6-MP arms in all trials except for one study.19
independently confirmed, incomplete outcome reporting, and imprecision.
CI includes substantial benefit but also potential harm (CI crosses 1.0).
Compared with placebo, thiopurine therapy showed a
indirectness
indirectness
Indirectness
trend toward fewer failures to achieve remission at 12 to 17
No serious
No serious
weeks (relative risk [RR], 0.87; 95% confidence interval,
0.71–1.06).21 Based on a placebo failure rate of 62.8%, thio-
purine therapy would result in 82 fewer failures per 1000
Time frame was not reported in the original manuscript.
inconsistency
inconsistency
patients (95% CI, from 182 fewer to 38 more) (Table 2). The CI
Inconsistency
included substantial benefit but also potential harm (crossing
No serious risk No serious
No serious
1.0). The overall quality of evidence was deemed moderate as a
result of imprecision. A low dose of 6-MP was used in one
trial,18 but a sensitivity analysis without this trial did not
Serious infectionsb (important outcome)
This was an adjusted effect estimate.
substantially change the overall estimate of the results.
The quality of data on serious infections varied among
Risk of
bias
of bias
the primary studies. Only 2 of the 5 studies specifically
Seriousc
reported serious infections.16,17 One study18 did not pro-
AGA
vide any data on overall infections, and the remaining 2
studies did not provide any detailed data specific to
of studies), follow-
participants (no.
380 (5 studies),
serious infections.20,22 Data on the risk of serious in-
5394 (1 study),
Khan et al21
5.2 y (mean
Lichtenstein
up, author
12–17 wk,
follow-up),
No. of
fections were thus obtained from a large prospective,
et al11
observational cohort study of 6273 patients with a mean
follow-up of 5.2 years.11 On multivariate analysis,
a
b
d
e
c
AGA
American Gastroenterological Association Institute Technical Review on
the Use of Thiopurines, Methotrexate, and Anti–TNF-a Biologic Drugs
for the Induction and Maintenance of Remission in Inflammatory
Crohn’s Disease
C rohn’s disease (CD) is a chronic inflammatory bowel
disease (IBD) that causes significant morbidity and
represents a considerable burden to society and the health
Gastroenterological Association,8 and is becoming the
common methodology for the streamlined development
of clear, transparent, and actionable guidelines.9
care system.1–5 Based on the latest administrative data, it is An accompanying report10 in this issue of GASTROEN-
estimated that 300,000 to 500,000 Americans have CD.2,6 TEROLOGY integrates the results of this technical review
A recent study reported annual treatment costs of $8265 with the other GRADE criteria to produce a set of
per patient, which extrapolates to yearly costs of $2.5 to $4 recommendations.
billion for the American population with CD.3
Two main principles guide the medical therapy of patients
Methods
with CD. First, because this is a lifelong, relapsing disorder,
therapy to induce remission (inductive therapy) is followed Overview
by therapy to maintain remission (maintenance therapy).7 This technical review (and the accompanying guideline) was
Second, the choice of inductive and maintenance therapies based on the GRADE framework. In developing this technical re-
depends on the severity of the disease and the response to less view, the authors first formulated a series of specific questions that
were to be answered by the guideline. The authors then identified
effective strategies. In this progressive approach to therapy,
the outcomes that were significant to answering each question and
mesalamine, antibiotics, and budesonide are used in patients rated them as critical or important. Next, the group systematically
with mild disease. Systemic corticosteroids, immunomodu- reviewed and summarized the evidence for each outcome across
lators, and anti–tumor necrosis factor (TNF)-a agents are studies, assessed the quality of evidence for each outcome, and
used in patients with moderately severe CD or in patients finally integrated the evidence across all the outcomes to answer
who fail to respond to therapy for mild disease. The immu- each specific question. The quality of the evidence was classified into
nomodulators include the thiopurine analogues, azathio- 4 categories: high, moderate, low, and very low. Assessment of the
prine (AZA) and 6-mercaptopurine (6-MP), and quality for each outcome took into account the study design, risk
methotrexate (MTX). Anti–TNF-a agents approved for use in of bias, inconsistency (or heterogeneity), indirectness, imprecision,
the United States include infliximab (IFX), adalimumab and potential publication bias (see the glossary of terms in
(ADA), and certolizumab pegol (CZP). Supplementary Methods for further explanations).
In this technical review, the American Gastroentero-
logical Association addresses the relative positioning of Outcomes of Interest
immunomodulators and anti–TNF-a biologic agents in Using the PICO format, which frames a clinical question
inducing and maintaining clinical remission in patients by defining a specific population (P), intervention (I), comparator
with inflammatory (luminal) CD. From the standpoint of (C), and outcome (O), we outlined a total of 16 PICO questions
(see Table 1). The patient population with moderate-severe active
patients and clinicians, selecting among immunomodu-
CD was defined as patients with a Crohn’s Disease Activity Index
lator monotherapy, anti–TNF-a monotherapy, and com- (CDAI) of 220 to 450. The population with CD in remission was
bination therapy is a common clinical dilemma. Providing defined as patients with a CDAI <150 and not being treated with
optimal, evidence-based care to the many patients who are
candidates for these potentially costly and/or toxic ther-
apies is of critical importance to the health care system as
Abbreviations used in this paper: ACG, American College of Gastro-
well. The Grading of Recommendations Assessment, enterology; ADA, adalimumab; AZA, azathioprine; BSG, British Society of
Development and Evaluation (GRADE) methodology was Gastroenterology; CD, Crohn’s disease; CDAI, Crohn’s Disease Activity
used in this technical review to assess the evidence on Index; CI, confidence interval; CZP, certolizumab pegol; ECCO, European
immunomodulators and anti–TNF-a-biologic agents in Crohn’s and Colitis Organisation; GRADE, Grading of Recommendations
AGA
Assessment, Development and Evaluation; HR, hazard ratio; IBD, in-
the (1) induction of remission in adult patients who have
flammatory bowel disease; IFX, infliximab; 6-MP, 6-mercaptopurine;
moderately severe inflammatory CD despite therapy with MTX, methotrexate; OR, odds ratio; PICO, population, intervention,
mesalamine, antibiotics, corticosteroids, and/or immuno- comparator, and outcome; RCT, randomized controlled trial; RR, relative
modulators and (2) maintenance of medically induced risk; SIR, standardized incidence ratio; TNF, tumor necrosis factor.
remission. GRADE has been adopted by several national © 2013 by the AGA Institute
0016-5085/$36.00
and international societies, including the American
http://dx.doi.org/10.1053/j.gastro.2013.10.046
,December 2013 AGA 1465
Table 1. PICO Questions
Population(s) Intervention(s) Comparator Outcome(s)
1 a. Adults with moderate-severe CD AZA or 6-MP Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
2 a. Adults with moderate-severe CD MTX Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
3 a. Adults with moderate-severe CD Anti–TNF-a Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections
and lymphoma
4 a. Adults with moderate-severe CD Thiopurine or Placebo a. Induction of remission; AE: serious infections
b. Adults with CD in remission MTX + anti–TNF-a b. Disease relapse; AE: serious infections and lymphoma
5 a. Adults with moderate-severe CD Thiopurine MTX a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
6 a. Adults with moderate-severe CD Thiopurine or MTX Anti–TNF-a a. Induction of remission; AE: serious infections
b. Adults with CD in remission b. Disease relapse; AE: serious infections and lymphoma
7 a. Adults with moderate-severe CD Thiopurine or MTX + Thiopurine a. Induction of remission; AE: serious infections
b. Adults with CD in remission anti–TNF-a or MTX b. Disease relapse; AE: serious infections and lymphoma
8 a. Adults with moderate-severe CD Thiopurine or MTX + Anti–TNF-a a. Induction of remission; AE: serious infections
b. Adults with CD in remission anti–TNF-a b. Disease relapse; AE: serious infections and lymphoma
NOTE. The following PICOs were excluded from the technical review because of absent or insufficient data: 4a, 4b, 6b, and 7b. Moderate-severe CD
was defined as a CDAI from 220 to 450. Remission was defined as a CDAI <150 and not being treated with corticosteroid therapy. Disease relapse
was defined as a CDAI 150 or corticosteroid use or surgery.
AE, adverse events.
corticosteroids. The interventions were immunomodulatory We selected serious infections and lymphoma as important
monotherapy, anti-TNF monotherapy, or combination therapy. adverse events potentially associated with serious morbidity or,
The comparators were immunomodulatory monotherapy or rarely, mortality. Serious infection was defined as infection that
anti-TNF monotherapy. PICO questions for which there were led to hospitalization (the definition used by the majority of
either no data or insufficient data (eg, in adult patients with studies). These outcomes were considered important but not
moderate-severe CD, should combination therapy with immu- critical for decision making. Because lymphoma is usually asso-
nomodulators plus anti-TNF therapy versus placebo be used to ciated with long-term therapy, we assessed the risk of lymphoma
induce remission?) could not be addressed in this technical re- only during maintenance therapy.
view. Evidence profiles were used to display the summary esti-
mates as well as the body of evidence for each clinical question.
For randomized controlled trials (RCTs) evaluating induction, Literature Search
the efficacy outcome considered critical for decision making was Three separate literature searches were conducted: one
corticosteroid-free clinical remission, defined as a CDAI <150 or for evidence summaries (such as meta-analyses); one for RCTs for
a Harvey–Bradshaw Index <4. For RCTs evaluating maintenance, the efficacy, infection, and lymphoma outcomes; and one for
the critical efficacy outcome was disease relapse, defined as a observational evidence to supplement the data on infection and
CDAI 150, corticosteroid use, or surgery. lymphoma. An information specialist developed each search with
We excluded trials without validated end points, such as trials input from the project team. All search results were imported
only reporting subjective symptom improvement and trials using bibliographic management software for de-duplication
evaluating only corticosteroid sparing. With regard to studies of and title and abstract screening.
induction with immunomodulators, which are agents with a The following bibliographic databases were searched through
delayed onset of action, we included studies with at least 12 the Ovid interface: MEDLINE, MEDLINE In-Process & Other
weeks of therapy. Conversely, we excluded trials that evaluated Non-Indexed Citations, and EMBASE. Parallel searches included
remission after more than 26 weeks of immunomodulator the Cochrane Database of Systematic Reviews, Database of Ab-
therapy, because an agent cannot properly be considered stracts of Reviews of Effects, Cochrane Central Register of
inductive and clinically relevant if its onset of action occurs Controlled Trials (CENTRAL), Cochrane Methodology Register,
beyond 26 weeks. For studies evaluating maintenance, only and HTA Database. The search strategy comprised controlled
medically induced remission was evaluated (ie, we excluded trials vocabulary, including the National Library of Medicine’s MeSH
that evaluated maintenance of surgically induced remission). It (Medical Subject Headings), and keywords. The main search
should be noted that the thiopurine and MTX maintenance concepts included and combined were “Crohn disease” and
studies were conducted before the anti–TNF-a era (ie, the pa- “immunomodulator therapy” and “anti-tumor necrosis factor.”
tients achieved remission with corticosteroids, mesalamine, and/ Methodological filters were applied to limit retrieval to RCTs,
or antibiotics). Interventions were analyzed based on their ability meta-analyses, systematic reviews, and health technology assess-
AGA
to reduce an undesirable outcome: failure to achieve clinical ments. The results were limited to English, human, and 1995
remission (in induction trials) or failure to prevent disease onward. The second search consisted of the main search con-
relapse (in maintenance trials). Based on clinical judgment, we cepts “Crohn disease” and “immunomodulator therapy” and
considered a relative reduction of failure to achieve (or maintain) “anti-tumor necrosis factor” plus “lymphoma.” The results were
remission of 20% as the minimum clinically important difference limited to English language and 2010 onward, because prior sys-
for immunomodulators or anti–TNF-a agents when compared tematic reviews using appropriate search strategies had adequately
with placebo and 10% when comparing drug classes. covered earlier time frames. A search for observational evidence on
, 1466 AGA GASTROENTEROLOGY Vol. 145, No. 6
harm was performed from this search (see Supplementary
The following issues were identified for rating down the quality: representativeness of patient population unclear, consecutive patients not enrolled (convenience sample), outcome adjudication not
fewer to 24 more)e
Data on serious infections were obtained from the study by Lichtenstein et al, which evaluated risk across all immunomodulators (AZA/6-MP/MTX); however, in 95% of cases, this consisted of treatment with a
(AZA/6-MP) (95% CI)
fewer to 38 more)
Risk difference with
82 fewer per 1000
10 more per 1000
Methods for the detailed search strategies). Updated information
Anticipated absolute effects
thiopurines
on serious infection and lymphoma (The Crohn’s Therapy,
(from 182
Resource, Evaluation, and Assessment Tool [TREAT] registry)
(from 2
became available during the writing process and was thus included
in this review (Lichtenstein et al, manuscript under review).11,12
Based on these searches, we identified existing systematic re-
views and used AMSTAR,13 a validated instrument, to evaluate
Risk with
the quality of systematic reviews. Systematic reviews that were of
controls
628 per
1000
1000
high quality, were up-to-date, and used the aforementioned
46 per
outcomes of interest (eg, corticosteroid-free remission based on
CDAI) were selected for inclusion in the evidence profiles. When
systematic reviews were not up-to-date or were incomplete, we
effect (95%
Relative
(0.96–
(0.71–
OR, 1.23
1.57)d
RR, 0.87
performed our own meta-analysis (random effects model for 3 or
1.06)
CI)
more studies and fixed effects model for 2 studies) using the
thiopurine (AZA/6-MP). Serious infection was defined as “any infection reported as serious by the investigator and any infection that required hospitalization.”
Cochrane Collaboration’s RevMan 5.1 software.14
Harm data from RCTs were sometimes of inadequate quality
(AZA/6-MP)
thiopurines
or had very few events. When this occurred, we identified
Table 2. Should Thiopurines (AZA/6-MP) Versus Placebo Be Used for Adults With Active (Moderate to Severe) CD (CDAI 220–450)?
Study event rates (%)
146/1849
With
102/197
observational studies and assessed them for risk of bias using the
(51.8)
(7.9)
Newcastle–Ottawa tool.15 We selected the observational studies
with the highest methodological quality and greatest number of
events for the outcomes of serious infections and lymphoma.
Undetected 4222 Very 148/3255
115/183
control
(62.8)
With
(4.6)
Results
Induction of Remission
risk of bias,
Overall quality
imprecision
imprecision
lowc due to
Moderatea
Thiopurines versus placebo. Five RCTs in 380 of evidence
Undetected 4442
due to
patients (thiopurine, 197; placebo, 183) compared AZA or
6-MP with placebo for the induction of remission.16–20 All
studies were blinded and likely achieved allocation
Failure of remission (critical outcome; remission assessed with CDAI <150 or Harvey–Bradshaw Index <4)
concealment. The patient populations differed with respect
Publication
to disease duration and location. The majority of patients
bias
had ileal or ileocolonic disease for at least 3 to 4 years and
had received prior medical therapy and/or had undergone
prior surgery. Except for one study that used 6-MP at a low
Imprecision
dosage of 50 mg/day,18 all studies examined the use of AZA
Seriousa
Seriousc
at a dosage of 2.5 mg/kg/day. Tapering doses of cortico-
steroids were given concomitantly in both the placebo and
AZA/6-MP arms in all trials except for one study.19
independently confirmed, incomplete outcome reporting, and imprecision.
CI includes substantial benefit but also potential harm (CI crosses 1.0).
Compared with placebo, thiopurine therapy showed a
indirectness
indirectness
Indirectness
trend toward fewer failures to achieve remission at 12 to 17
No serious
No serious
weeks (relative risk [RR], 0.87; 95% confidence interval,
0.71–1.06).21 Based on a placebo failure rate of 62.8%, thio-
purine therapy would result in 82 fewer failures per 1000
Time frame was not reported in the original manuscript.
inconsistency
inconsistency
patients (95% CI, from 182 fewer to 38 more) (Table 2). The CI
Inconsistency
included substantial benefit but also potential harm (crossing
No serious risk No serious
No serious
1.0). The overall quality of evidence was deemed moderate as a
result of imprecision. A low dose of 6-MP was used in one
trial,18 but a sensitivity analysis without this trial did not
Serious infectionsb (important outcome)
This was an adjusted effect estimate.
substantially change the overall estimate of the results.
The quality of data on serious infections varied among
Risk of
bias
of bias
the primary studies. Only 2 of the 5 studies specifically
Seriousc
reported serious infections.16,17 One study18 did not pro-
AGA
vide any data on overall infections, and the remaining 2
studies did not provide any detailed data specific to
of studies), follow-
participants (no.
380 (5 studies),
serious infections.20,22 Data on the risk of serious in-
5394 (1 study),
Khan et al21
5.2 y (mean
Lichtenstein
up, author
12–17 wk,
follow-up),
No. of
fections were thus obtained from a large prospective,
et al11
observational cohort study of 6273 patients with a mean
follow-up of 5.2 years.11 On multivariate analysis,
a
b
d
e
c
