Hannink et al.

Systematic Reviews 2013, 2:52

http://www.systematicreviewsjournal.com/content/2/1/52

RESEARCH Open Access

A systematic review of individual patient data

meta-analyses on surgical interventions
Gerjon Hannink1*, Hein G Gooszen1, Cornelis JHM van Laarhoven2 and Maroeska M Rovers1,3

Abstract
Background: Compared to subgroup analyses in a single study or in a traditional meta-analysis, an individual
patient data meta-analysis (IPDMA) offers important potential advantages. We studied how many IPDMAs report on
surgical interventions, how many of those surgical IPDMAs perform subgroup analyses, and whether these
subgroup analyses have changed decision-making in clinical practice.
Methods: Surgical IPDMAs were identified using a comprehensive literature search. The last search was conducted
on 24 April 2012. For each IPDMA included, we obtained information using a standardized data extraction form,
and the quality of reporting was assessed. We also checked whether results were implemented in clinical
guidelines.
Results: Of all 583 identified IPDMAs, 22 (4%) reported on a surgical intervention. Eighteen (82%) of these IPDMAs
presented subgroup analyses. Subgroups were mainly based on patient and disease characteristics. The median
number of reported subgroup analyses was 3.5 (IQR 1.25-6.5). Statistical methods for subgroup analyses were
mentioned in 11 (61%) surgical IPDMAs.
Eleven (61%) of the 18 IPDMAs performing subgroup analyses reported a significant overall effect estimate, whereas
six (33%) reported a non-significant one. Of the IPDMAs that reported non-significant overall results, three IPDMAs
(50%) reported significant results in one or more subgroup analyses. Results remained significant in one or more
subgroups in eight of the IPDMAs (73%) that reported a significant overall result.
Eight (44%) of the 18 significant subgroups appeared to be implemented in clinical guidelines. The quality of
reporting among surgical IPDMAs varied from low to high quality.
Conclusion: Many of the surgical IPDMAs performed subgroup analyses, but overall treatment effects were more
often emphasized than subgroup effects. Although, most surgical IPDMAs included in the present study have only
recently been published, about half of the significant subgroups were already implemented in treatment guidelines.

Background undertake, and even worse, irrelevant to their practice

Surgery has advanced spectacularly in the past 50 years, because of concerns about generalizability [2,3].
but many advances have not come from carefully The results of RCTs are usually implemented in prac-
planned research using valid study designs [1]. Research tice by either treating or testing all patients in case of a
on surgical interventions is associated with several meth- ‘positive’ study or treating or testing no-one in case of a
odological and practical challenges of which few, if any, ‘negative’ study. Clinicians intuitively know that this ap-
apply only to surgery. Surgical innovation is especially proach is oversimplified because in reality some patients
demanding because many of these challenges coincide benefit more than average whereas others do not benefit.
[2]. Perhaps this situation leads many surgeons to view This may explain why around 50% of the RCTs perform
randomized controlled trials (RCTs), although theoretic- subgroup analyses [4,5]. However, misleading claims
ally advantageous, to be too difficult and impractical to about subgroup effects based on a single study are com-
mon [6].
* Correspondence: [email protected] Investigating subgroups is a highly relevant, but com-
1
Department of Operating Rooms, Radboud University Nijmegen Medical plex topic because of two interrelated concerns: failure
Center, PO Box 9101, Nijmegen 6500HB, The Netherlands
Full list of author information is available at the end of the article
to detect a relevant subgroup effect (false negative), and

© 2013 Hannink et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Hannink et al. Systematic Reviews 2013, 2:52 Page 2 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

a misleading claim about a subgroup effect which in of datasets, studying/addressing different questions/
reality does not exist (false positive). Both of these prob- subgroups were included. If obvious duplicate pa-
lems can lead to suboptimal care for patients. Subgroup pers were available, the most elaborate paper was
effects have been extensively and fiercely debated in the included.
clinical, epidemiological, and statistical literature, espe-
cially in the context of single trials or traditional meta- Data extraction and analysis
analyses based on published summary results [7-11]. Data from all included surgical IPDMAs were extracted
Individual patient data meta-analyses (IPDMAs) differ with respect to specific characteristics, that is, publica-
from traditional meta-analyses in that an IPD meta- tion year, number of included trials and patients, do-
analysis uses the ‘raw data’ of individual patients from main, type of intervention, comparison, and outcome
included studies instead of the published summary re- measured. Regarding the subgroups, number, type, justi-
sults of studies in a traditional meta-analysis [12]. Com- fication, statistical methods, and results in relation to
pared to subgroup analyses in a single study or in a the overall effect estimate were studied.
traditional meta-analysis, an IPDMA offers important We classified five types of subgroups, patient charac-
potential advantages, such as: (1) increased possibilities teristics (for example, age or gender), disease character-
to perform more complex statistical analyses that better istics (for example, severity or co-morbidity), household
match the underlying data; (2) more power compared characteristics (for example, socioeconomic status or
to single studies and traditional meta-analyses; (3) higher smoking), intervention characteristics (for example, type
validity of subgroup analyses by avoiding ecological bias of intervention or dose), and methodological characteris-
and by taking the distribution of other patient charac- tics (for example, quality of included trials or trial
teristics into account; (4) improved flexibility and stan- effect). Justification for subgroups analyses was catego-
dardization of defining subgroups across studies; and (5) rized as based on literature, clinical experience, bio-
opportunities to examine the consistency of subgroup logical mechanism, or no justification.
effects across studies [13-17]. We also assessed the quality of reporting of all selected
In this paper we present a systematic overview of IPDMAs. IPDMAs on RCTs should be reported
all IPDMAs on surgical interventions published. We according to the Preferred Reporting Items for System-
studied the number and types of subgroup analyses atic Reviews and Meta-Analyses (PRISMA) [18]. Since
performed, and whether these subgroups analyses this guideline is not specific to IPDMAs, it has been sug-
influenced decision-making in clinical practice. gested that some additional information should be
reported, for instance why the IPDMA approach was ini-
Methods tiated, whether there was a protocol for the IPDMA pro-
Search ject, and whether a one-step or a two-step analysis was
A comprehensive literature search in PubMed, Embase, performed [12]. We judged the quality of reporting
Web of Science, and the Cochrane Library was based on the 18 criteria suggested by Riley et al. [12].
conducted to identify all IPDMAs of RCTs. The last Two independent reviewers (GH and MMR) selected eli-
search was conducted on 24 April 2012. Keywords used gible surgical IPDMAs and extracted data (duplicate in-
to develop our search strategy were ‘individual patient dependently). Any disagreements were resolved by
data’ and ‘meta-analysis’ (see Additional file 1 for de- consensus.
tailed search strategy). Finally, we reviewed available clinical guidelines for
recommendations based on significant results of sub-
Selection group analyses from IPDMAs to determine the extent to
In first instance, titles and abstracts were screened to which these results were implemented in clinical guide-
identify eligible IPDMAs. Selection of potential eligible lines. We conducted a PubMed search for fields ‘pa-
IPDMAs was restricted to IPD obtained from RCTs tients’ (for example, carotid stenosis), ‘intervention’ (for
comparing surgical interventions. Patients had to be ran- example, carotid stenting) and ‘comparison’ (for ex-
domized over a surgical intervention in at least one ample, endarterectomy), extracted from IPDMAs with
treatment arm, and the surgical procedures had to be significant subgroup analyses, and limited our search to
performed under general, spinal, epidural, or regional ‘Practice Guideline’. We only included publications in
anesthesia. IPDMAs regarding drug-eluting medical de- English. We also searched the National Library of
vices and surgical trials in which a drug was the com- Guidelines (http://guidance.nice.org.uk/), and the Na-
parison were excluded. tional Guideline Clearinghouse (http://www.guidelines.
Full text papers were retrieved when meta-analytic gov/).
techniques for individual patient data of RCTs were All steps in this review were carried out according to a
used. IPDMAs using the same dataset or combination pre-defined protocol (Additional file 2).
Hannink et al. Systematic Reviews 2013, 2:52 Page 3 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

Results The remaining non-surgical IPDMAs predominantly

Search focused on cancer, cardiovascular disease, and diabetes,
In the search for IPDMAs, 3,597 potential eligible papers and most assessed whether a treatment or intervention
were identified. After studying the abstracts, 583 papers, was effective, often in subgroups of patients. Before 2000
published between 1991 and 2012, indeed reported an only a few IPDMAs were published, whereas a consider-
IPDMA. After detailed evaluation, 22 (4%) IPDMAs able rise in the number of published IPDMAs is seen be-
reported on a surgical intervention and met our inclu- tween 2005 and 2012 (Figure 2). This growth is most
sion criteria (Figure 1; Additional file 3). likely the result of an increased awareness the potential
Of the 22 surgical IPDMAs, 12 focused on cardiovas- advantages of IPDMAs, and the initiation of collabora-
cular interventions, three on inguinal hernia repair, three tions to specifically perform such studies.
on gynecological interventions, two on orthopedic inter-
ventions, one on a gastroenterological intervention, and Summary of IPDMAs using IPD (or part of IPD) from the
one on ventilation tubes for otitis media (Table 1). The same trials
surgical IPDMA papers were published between 2005 Of the 12 IPDMAs that focused on a cardiovascular
and 2012. Eighteen (82%) of the 22 surgical IPDMAs intervention, four IPDMAs [26-29] used individual pa-
tried to identify subgroups of patients that benefit more tient data from the same 22 trials (6,763 patients) evalu-
or less from the surgical intervention. ating the clinical effects of primary percutaneous coronary

Figure 1 Flowchart of study selection process for IPDMA of surgical interventions.

 http://www.systematicreviewsjournal.com/content/2/1/52
Hannink et al. Systematic Reviews 2013, 2:52
Table 1 Characteristics of the 22 identified surgical IPDMAs
Author and year RCTs Patients Patients Intervention Comparison Outcome Subgroups (n) Overall Significant
(n) (n) effect effect
estimate estimates
in
subgroups
(n/N)
Jorgenson et al., 2007 7 2,091 Women with Cervical cerclage Expectant Primary Obstetric history, cervical length NS 0/2
[19] cervical management, no (2)
Pregnancy loss or neonatal death
insufficiency cerclage
before discharge from hospital
Secondary
Preterm delivery and maternal
morbidity
Hlatky et al., 2009 [20] 10 7,812 Patients with Coronary artery Percutaneous All-cause mortality Age, sex, diabetes, smoking, NS 2/14
multivessel bypass graft coronary hypertension,
coronary disease intervention hypercholesterolaemia, PVD,
stability of symptoms, previous
MI, heart failure, LV function, no.
of diseased vessel, proximal LAD,
balloon vs.stent (14)
Daniels et al., 2010 5 862 Patients with Laparoscopic No LUNA Derived measure of worst pain Presence of visual pathology, site NS 1/4
[21] chronic pelvic uterosacral nerve level experienced of pain, age, parity (4)
pain ablation (LUNA)
Burzotta et al., 2009 11 2,686 Patients with ST- Percutaneous Standard Primary Manual vs. non-manual S 1/7
[22] elevation coronary percutaneous thrombectomy devices, diabetes,
All-cause mortality
myocardial intervention with coronary primary vs. rescue PCI, treated vs.
infarction (STEMI) thrombectomy intervention Secondary non-treated with IIb/IIIa-inhibitors,
ischemic time, infarct-related
Survival free from MI, TLR, or TVR, artery, pre-PCI TIMI flow (7)
major adverse coronary events
(MACE), death+MI
Carotid Stenting 3 3,433 Patients with Carotid stenting Endarterectomy Primary Age, sex, diabetes, hypertension, S 1/16
Trialists’Collaboration, symptomatic SBP, hypercholesterolaemia,
2010 [23] carotid stenosis Any stroke or death smoking, coronary heart disease,
Secondary peripheral artery disease, most
recent ipsilateral ischemic event,
Disabling stroke or death, all- history of stroke, degree of
cause death, any stroke, ipsilateral ischemic stroke,
myocardial infarction, severe local contralateral severe carotid
hematoma, severe wound stenosis or occlusion, treatment
infection within 14 days, patients recruited
per center, center recruitment
rate (16)
Middleton et al., 2010 17 2,814 Patients with Hysterectomy, Endometrial Dissatisfaction rates Uterine cavity length, age, S 1/5

Page 4 of 12
[24] heavy menstrual endometrial destruction (1st & presence of fibroids/polyps,
bleeding destruction (1st & 2nd generation), parity, baseline bleeding score (5)
2nd generation), levonorgestrel
levonorgestrel releasing intra-
 http://www.systematicreviewsjournal.com/content/2/1/52
Hannink et al. Systematic Reviews 2013, 2:52
Table 1 Characteristics of the 22 identified surgical IPDMAs (Continued)
releasing intra- uterine system
uterine system (MIRENA)
(MIRENA)
Mercado et al., 2005 4 3,051 Patients with Percutaneous Coronary artery Primary Age, gender, diabetes, smoking, NS 0/5
[25] multi-system coronary bypass graft number of diseased vessels (5)
Composite of death, MI, or stroke
coronary artery intervention with
at 1 year FU
disease multiple stenting
Secondary
Death, composite of death or MI,
repeat revascularization,
composite of death, MI, stroke,
and repeat revalscularization
Boersma et al., 2006 22 6,767 Patients with PCI Fibrinolysis All-cause mortality Age, sex, diabetes, prior MI, MI S 1/11
[26] acute myocardial location, heart rate, SBP,
infarction fibrinolytic agent, front-loaded
tPA, site volume (11)
Timmer et al., 2007 19 6,315 Patients with PCI Fibrinolysis Death, recurrent MI, death or Diabetes (1) S 0/1
[27] acute myocardial recurrent MI, stroke
infarction
de Boer et al., 2010 22 6,767 Patients with Primary PCI Fibrinolysis Primary Age (1) S 0/1
[28] acute myocardial
infarction All-cause mortality
Secondary reMI, stroke, composite
of all-cause mortality or reMI,
composite of all-cause mortality,
reMI, or stroke
de Boer et al., 2011 22 6,767 Patients with Primary PCI Fibrinolysis All-cause mortality High-risk patients (1) S 0/1
[29] acute myocardial
infarction
Fox et al., 2010 [30] 3 5,467 Patients with Routine invasive Selective invasive Primary High-risk groups based on S 1/1
non-ST-elevation strategy strategy baseline characteristics (1)
Composite of CV death or non-
myocardial
fatal MI
infarction
Secondary
All-cause death, non-fatal MI
alone
Damman et al., 2012 3 5,467 Patients with Routine invasive Selective invasive Primary Age (1) S 1/1
[31] non-ST-elevation strategy strategy
myocardial Composite of CV death or non-
fatal MI, CV death, MI
infarction
Damman et al., 2012 3 5,467 Patients with Routine invasive Selective invasive All-cause mortality Procedure-related MI, S 1/2
[32] non-ST-elevation strategy strategy spontaneous MI (2)

Page 5 of 12
myocardial
infarction
Biau et al., 2009 [33] 6 423 Primary S 2/3
 http://www.systematicreviewsjournal.com/content/2/1/52
Hannink et al. Systematic Reviews 2013, 2:52
Table 1 Characteristics of the 22 identified surgical IPDMAs (Continued)
Patients with Reconstruction with Reconstruction Positive pivot-shift test Secondary Gender, age at surgery, trial effect
symptomatic patellar tendon with hamstring (3)
unilateral anterior autograft tendon autograft Positive Lachman test
cruciate ligament
injury
Rovers et al., 2005 7 1,234 Children with Short-term Watchful waiting Mean time spent with effusion, Hearing level at baseline, history NS 2/9
[34] otitis media with ventilation tubes hearing, language development of acute otitis media, upper
effusion respiratory infections, attending
day care, socioeconomic status,
siblings, season, history of
breastfeeding, parental smoking
(9)
Salerno et al., 2007 4 305 Cirrhotic patients Transjugular Paracentesis Primary NA S NA
[35] with refractory intrahepatic
Death from any cause before LT
ascites portosystemic shunt
(TIPS) Secondary
Liver-related death
Staples et al., 2011 2 209 Patients with Vertebroplasty Sham Scores for pain and function Onset of pain, pain scores at NS 0/2
[36] osteoporotic baseline (2)
vertebral
compression
fractures
McCormack et al., 25 4,165 Patients with Laparoscopic repair Open repair Duration of operation, ‘opposite’ NA S NA
2003 [37] clinical diagnosis method initiated, conversion,
of groin hernia (Transabdominal hematoma, seroma, wound/
for whom preperitoneal repair superficial infection, mesh/deep
(TAPP) or totally
surgical infection, port site hernia, vascular
extraperitoneal
management was injury, visceral injury, length of
judged repair (TEP)) hospital stay, time to return to
appropriate usual activities, persisting pain,
persisting numbness, hernia
recurrence, known death within
30 days of surgery
Scott et al., 2002 [38] 11 3,347 Patients with Mesh technique Non-mesh Duration of operation, ’opposite’ NA S NA
clinical diagnosis technique method initiated, conversion,
of groin hernia hematoma, seroma, wound/
for whom superficial infection, serious
surgical complications, length of
management was postoperative hospital stay, time
judged to return to usual activities,
appropriate persisting pain, persisting
numbness, hernia recurrence,
known death

Page 6 of 12
EU Hernia Trialists 35 6,901 Patients with Laparoscopic repair, Open repair, non- Hernia recurrence, persisting pain NA S NA
Collaboration, 2002 clinical diagnosis mesh methods mesh methods
[39] of groin hernia
for whom
surgical
 http://www.systematicreviewsjournal.com/content/2/1/52
Hannink et al. Systematic Reviews 2013, 2:52
Table 1 Characteristics of the 22 identified surgical IPDMAs (Continued)
management was
judged
appropriate
Gregson et al., 2012 8 2,186 Patients with Surgery Conservative Unfavorable outcome Location of hematoma, time from NA 4/5
[40] spontaneous treatment event, age, Glascow Coma Score,
supratentorial volume of hematoma (5)
intracerebral
hemorrhage

Page 7 of 12
Hannink et al. Systematic Reviews 2013, 2:52 Page 8 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

Figure 2 Number of applied IPDMA published up to April 2012,* as identified by a systematic review of PubMed, Embase, Web of
Science, and the Cochrane Library. *Thirty-seven IPDMAs published in 2012 were identified up to 24 April 2012, when the review
was conducted.

intervention versus in-hospital fibrinolysis. In three IPDMAs methods for subgroup analyses were mentioned in 11
[30-32], comparing routine invasive strategies with se- (61%) of the 18 IPDMAs performing subgroup analyses.
lective invasive strategies in 5,467 patients with non- All IPDMAs that mentioned statistical methods for sub-
ST segment elevation acute coronary syndromes, data group analysis used interaction tests.
from the same three trials (FRISC II, ICTUS, and Only five (28%) surgical IPDMAs mentioned the power
RITA-3) were used. In addition, of the three IPDMAs of the subgroup analyses. Three IPDMAs reported that
that focused on inguinal hernia repair one IPDMA their studies were underpowered to detect subgroup ef-
[39] presented a combination of the data used in the fects, one IPDMA reported that their study was well
other two IPDMAs [37,38]. powered to detect subgroups effects, however, did fail to
show differences in subgroups, and one IPDMA men-
Number, justification, type, and methods of subgroups tioned differences in power between different subgroups,
analyses in surgical IPDMAs but not whether these were over- or underpowered.
In 18 (82%) of the full set of surgical IPDMAs assessed, Eleven (61%) of the 18 IPDMAs performing subgroup
subgroup analyses were performed to examine whether analyses reported a significant overall effect estimate,
certain patients benefit more from a specific treatment whereas six (33%) reported a non-significant one. One
than others. The median number of subgroups reported IPDMA (6%) did not report an overall effect estimate
in these IPDMAs was 3.5 (range, 1–16, IQR 5.25 (1.25- and only presented results of subgroup analyses [40].
6.5)). In 12 (67%) of the 18 surgical IPDMAs that stud- Of the IPDMAs that reported non-significant overall
ied subgroups a justification for subgroup analyses was results, three IPDMAs (50%) reported significant results
mentioned. Scientific literature was used for justification in one or more subgroup analyses. Results remained sig-
in these studies. nificant in one or more subgroups in eight of the
The types of subgroups studied varied. Fifteen (83%) IPDMAs (73%) that reported a significant overall result.
IPDMAs studied patient characteristics, five (28%) stud- Thirty-six (40%) of the total number of 90 subgroups ana-
ied household characteristics, 15 (83%) studied disea- lyses were performed on a non-significant or inconclusive
se characteristics, and six (33%) studied intervention- overall effect estimate, whereas 49 (54%) were performed on
related subgroups. Subgroups related to study or trial a significant overall effect estimate. The remaining five (6%)
effects were studied in three (17%) IPDMAs. No IPDMAs subgroups originated from the one IPDMA that did not re-
studied subgroups related to the quality of the included port an overall effect estimate and only presented results of
trials, for example concealment of allocation, blinding, or subgroup analyses, four out of these five subgroups being sig-
completeness of follow-up. nificant [40]. Of the subgroup analyses performed on non-
Twelve (55%) IPDMAs stratified their analysis per trial significant overall results, five (14%) became significant. Nine
before pooling the results (two-step analysis). A one-step (18%) of those performed on IPDMAs with a significant
analysis was performed in four (18%) IPDMAs. Statistical overall result remained significant.
Hannink et al. Systematic Reviews 2013, 2:52 Page 9 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

Eight (67%) of the 12 surgical IPDMAs with significant potential advantages of IPDMAs (see Table 2 for exam-
subgroups reported on what the implications of these ples). In 18 (82%) of the full set of 22 surgical IPDMAs
significant results of their subgroup analyses were for assessed, subgroup analyses were performed to examine
clinical practice. Mainly, the importance of differentiat- whether certain patients benefit more from a specific
ing when evaluating the efficacy and safety of new med- treatment than others. Eight (67%) of the 12 surgical
ical and interventional treatments, and translating these IPDMAs with significant subgroups reported on what
findings in treatment recommendations were empha- the implications of their findings were for clinical prac-
sized. Moreover, it was reported that the influence of tice. Forty-four percent (8 out of 18) of the significant
certain subgroups had not been reported previously, that subgroups were implemented in clinical guidelines.
findings concurred with recent recommendations or Although many IPDMAs performed subgroup ana-
guidelines, and that subgroups not per se needed to be lyses, the overall treatment was usually the main focus
an exclusion criterion for treatment. Eight (44%) of the of the paper. Only occasionally subgroup analyses were
18 significant subgroups were implemented in clinical emphasized. In surgical IPDMAs, similar to IPDMAs in
guidelines. general [42], subgroups were often based on patient and
disease characteristics. The median number of sub-
Quality of reporting groups has been reported to range from 2 to 4, the max-
The quality of reporting of the surgical IPDMAs varied imum number of subgroups from 15 to 50 [6,43-45],
(Figure 3). More than half of the IPDMA failed to report which is comparable to our findings. Justification of sub-
whether or not there was a protocol for the IPDMA pro- group analyses, the methods used to perform subgroup
ject available. The reason why the IPD approach was ini- analyses, and power calculations for performing sub-
tiated and the numbers of patients within each of the group analyses are often not reported in IPDMAs
original studies were generally well reported. For 17 [6,43-48]. However, 11 (65%) of the IPDMAs included in
(77%) IPDMAs, the process used to identify relevant our study justified at least one of the subgroups on
studies for the IPDMA were reported. Details on the which they reported, scientific literature being the mode
statistical analysis were reported in 16 (73%) IPDMAs, of justification used. This is in line with other studies
however, details on the identification process and statis- that found that clinical experience or biochemical justifi-
tical analysis were not described in one IPDMA (4%), cation is rare [44,46,49]. Others showed that the propor-
and were unclear in the remaining five (23%) IPDMAs. tion of studies that used interaction tests for at least one
of their subgroups ranges from 10% to 56% [6,43-48],
Discussion which is slightly lower compared to our findings. So far,
Our systematic review of all IPDMAs on surgical inter- few studies mentioned the importance of the power of
ventions published so far provides an overview of the subgroup analyses [6,44,50,51], and reported that many

Figure 3 Quality of IPDMA reporting surgical interventions. Numbers inside bars are numbers of studies.
Hannink et al. Systematic Reviews 2013, 2:52 Page 10 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

Table 2 Two examples of differences in conclusions with regard to how patient-level characteristics modify treatment
effect
Example Description
Effectiveness of coronary artery bypass grafting vs. percutaneous A two-step meta-analysis of individual patient data from 7,812 patients included
coronary interventions for multivessel disease. in 10 randomized trials comparing coronary artery bypass grafting (CABG) and
percutaneous coronary intervention (PCI) in patients with multivessel coronary
artery disease, showed a similar overall treatment effect on long-term mortality
after CABG and PCI [20]. However, in diabetic patients mortality was substantially
lower in the CABG group than in the PCI group (HR 0.70, 95% CI 0.56-0.87).
Mortality was similar between groups in patients without diabetes (HR 0.98, 95%
CI 0.86-1.12; P=0.01 for interaction). Patient age modified the effect of treatment
on mortality with hazard ratios of 1.25 (95% CI 0.94-1.66) in patients aged <55
years, 0.90 (95% CI 0.75-1.09) in patients aged 55–64 years, and 0.82 (95% CI 0.70-
0.97) in patients aged ≥65 years (P=0.002 for interaction). Treatment effect was
not modified by other subgroups. CABG might be a better option for patients
aged ≥65 years and patients with diabetes since mortality was found to be
lower in these subgroups. These results have been implemented in clinical
guidelines [41].
Effectiveness of routine vs. selective invasive strategy in patients An individual patient data meta-analysis of three randomized trials of routine
with non-ST-segment elevation acute coronary syndrome. versus selective invasive strategies in patients with non-ST-segment elevation
acute coronary syndrome showed that a routine invasive strategy resulted in
significantly less cardiovascular deaths (CV deaths) or non-fatal myocardial
infarctions (MIs) compared to selective invasive strategies [30]. The authors used
patient’s baseline characteristics to develop a multivariable risk prediction model.
A simplified integer risk score was derived from the risk prediction model to
predict a patient’s 5-year probability of CV death or MI, and the patients were
categorized into three risk groups (low, intermediate, and high risk).
The treatment effect was similar between groups in patients with low-risk (HR
0.80 (95% CI 0.63-1.02)) and intermediate-risk (HR 0.81 (95% CI 0.66-1.01)) scores.
In patients with high-risk scores treatment favored routine over selective invasive
strategies (HR 0.68 (95% CI 0.53-0.86)). There were 2.0% (95% CI −4.1-0.1%) and
3.8% (95% CI −7.4- -0.1%) absolute risk reductions in CV death or MI in the low-
and intermediate-risk groups and an 11.1% (95% CI −18.4- -3.8%) absolute risk
reduction in the highest-risk patients. The multivariable risk prediction model has
not yet been implemented in clinical guidelines.

reports put too much emphasis on subgroup analyses several studies included a same set of trials and the pat-
that commonly lacked statistical power. This is in agree- tern of exploring heterogeneity among these studies
ment with the results of the present study. might be similar, that is, there might be a clustering ef-
To the best of our knowledge we are the first to study fect. As this might impact the subsequent analyses as
surgical IPDMAs, and illustrate the merits of this well as the conclusion, we performed a sensitivity ana-
method within surgery. However, some potential limita- lysis. The outcomes and conclusions were not substan-
tions should also be discussed. First, our literature tially influenced by the inclusion of studies using the
search for surgical IPDMAs was limited to IPDMAs same set of trials. Fifth, the time from publication to im-
with IPD obtained from RCTs comparing surgical inter- plementation of a result into a guideline or clinical prac-
ventions, excluding IPDMAs regarding drug-eluting tice can be highly variable, and sometimes takes even
medical devices and surgical trials in which a drug was more than 10 years [52]. Most surgical IPDMAs in-
the comparison, and records were limited to the English cluded in the present study have only recently been
language. We, however, believe that our review provides published, and time to possible implementation has
a good representation of the method within the surgical been rather short. Therefore, we might have under-
field. Second, reporting bias could not be entirely ex- estimated the implementation of results from IPDMAs
cluded, since reporting of subgroup effects in scientific into guidelines and/or clinical practice.
publications might be influenced by reviewers’ and edi-
tors’ opinions. Third, as most studies mentioned mul- Conclusions
tiple subgroups, a clustering effect might occur for One of the challenges in medicine is to rationally imple-
reporting on justification and statistical methods. There- ment available therapies in clinical practice, in the ap-
fore, the results were reported on study level instead propriate patients at the appropriate time. Findings from
of individual subgroup level. Fourth, in the 12 IPDMAs IPDMAs might provide insight into opportunities to im-
of cardiovascular interventions reporting subgroup analyses, prove evidence-based treatment decisions for patients.
Hannink et al. Systematic Reviews 2013, 2:52 Page 11 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

IPDMA is an extremely powerful tool if used correctly 6. Assmann SF, Pocock SJ, Enos LE, Kasten LE: Subgroup analysis and other
and provides the most definitive synthesis of the avail- (mis)uses of baseline data in clinical trials. Lancet 2000, 355:1064–1069.
7. Feinstein AR: The problem of cogent subgroups: a clinicostatistical
able evidence, also for potential subgroups. Despite the tragedy. J Clin Epidemiol 1998, 51:297–299.
recommendations available on reporting clinical trials 8. Brookes ST, Whitley E, Peters TJ, Mulheran PA, Egger M, Davey Smith G:
and meta-analyses, such as PRISMA [18], these guide- Subgroup analyses in randomised controlled trials: quantifying the risks
of false-positives and false-negatives. Health Technol Assess 2001, 5:1–56.
lines have not been specifically developed for IPDMAs. 9. Rothwell PM: Treating individuals 2. Subgroup analysis in randomised
The development of a generally accepted guideline for controlled trials: importance, indications, and interpretation. Lancet 2005,
reporting on IPDMAs including subgroup analyses 365:176–186.
10. Gabler NB, Duan N, Liao D, Elmore JG, Ganiats TG, Kravitz RL: Dealing with
should therefore be encouraged. This seems the only op- heterogeneity of treatment effects: is the literature up to the challenge?
tion to really improve the reporting, analyses, and claims Trials 2009, 10:43.
and applicability of subgroup effects in clinical research. 11. Bender R, Koch A, Skipka G, Kaiser T, Lange S: No inconsistent trial
assessments by NICE and IQWiG: different assessment goals may lead to
different assessment results regarding subgroup analyses. J Clin
Additional files Epidemiol 2010, 63:1305–1307.
12. Riley RD, Lambert PC, Abo-Zaid G: Meta-analysis of individual participant
data: rationale, conduct, and reporting. BMJ 2010, 340:c221.
Additional file 1: Detailed search strategy. 13. Berlin JA, Santanna J, Schmid CH, Szczech LA, Feldman HI: Individual
Additional file 2: PRISMA 2009 Checklist. patient- versus group-level data meta-regressions for the investigation
Additional file 3: PRISMA 2009 Flow Diagram. of treatment effect modifiers: ecological bias rears its ugly head.
Stat Med 2002, 21:371–387.
14. Fisher DJ, Copas AJ, Tierney JF, Parmar MK: A critical review of methods
Competing interests for the assessment of patient-level interactions in individual participant
The authors declare that they have no competing interests. data meta-analysis of randomized trials, and guidance for practitioners.
J Clin Epidemiol 2011, 64:949–967.
15. Lambert PC, Sutton AJ, Abrams KR, Jones DR: A comparison of summary
Authors’ contributions patient-level covariates in meta-regression with individual patient data
GH prepared the protocol with guidance from HG, CvL, and MR. GH and MR meta-analysis. J Clin Epidemiol 2002, 55:86–94.
developed the search strategies. GH and MR selected relevant studies and
16. Tugwell P, Knottnerus JA: Advantages of individual patient data analysis
extracted data; all authors participated in screening/extraction for the initial
in systematic reviews. J Clin Epidemiol 2010, 63:233–234.
unpublished version of the review. GH carried out the analysis and prepared
17. van Walraven C: Individual patient meta-analysis–rewards and challenges.
the manuscript with input from all authors. All authors read and approved
J Clin Epidemiol 2010, 63:235–237.
the final manuscript.
18. Moher D, Liberati A, Tetzlaff J, Altman DG: Preferred reporting items for
systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009,
Funding 339:b2535.
This research received no grant from any funding agency in the public, 19. Jorgensen AL, Alfirevic Z, Tudur Smith C, Williamson PR: Cervical stitch
commercial or not-for-profit sectors. (cerclage) for preventing pregnancy loss: individual patient data
meta-analysis. BJOG 2007, 114:1460–1476.
Author details 20. Hlatky MA, Boothroyd DB, Bravata DM, Boersma E, Booth J, Brooks MM,
1
Department of Operating Rooms, Radboud University Nijmegen Medical Carrie D, Clayton TC, Danchin N, Flather M, Hamm CW, Hueb WA, Kahler J,
Center, PO Box 9101, Nijmegen 6500HB, The Netherlands. 2Department of Kelsey SF, King SB, Kosinski AS, Lopes N, McDonald KM, Rodriguez A,
Surgery, Radboud University Nijmegen Medical Center, Nijmegen, The Serruys P, Sigwart U, Stables RH, Owens DK, Pocock SJ: Coronary artery
Netherlands. 3Department of Epidemiology, Biostatistics & HTA, Radboud bypass surgery compared with percutaneous coronary interventions for
University Nijmegen Medical Center, Nijmegen, The Netherlands. multivessel disease: a collaborative analysis of individual patient data
from ten randomised trials. Lancet 2009, 373:1190–1197.
Received: 11 November 2012 Accepted: 14 June 2013 21. Daniels JP, Middleton L, Xiong T, Champaneria R, Johnson NP, Lichten EM,
Published: 5 July 2013 Sutton C, Vercellini P, Gray R, Hills RK, Jones KD, Aimi G, Khan KS,
International LUNA IPD Meta-analysis Collaborative Group: Individual
References patient data meta-analysis of randomized evidence to assess the
1. McCulloch P: How to improve surgical research. BMJ 2011, 343:d4121. effectiveness of laparoscopic uterosacral nerve ablation in chronic pelvic
2. Ergina PL, Cook JA, Blazeby JM, Boutron I, Clavien PA, Reeves BC, Seiler CM, pain. Hum Reprod Update 2010, 16:568–576.
Balliol C, Altman DG, Aronson JK, Barkun JS, Campbell WB, Cook JA, 22. Burzotta F, De Vita M, Gu YL, Isshiki T, Lefevre T, Kaltoft A, Dudek D, Sardella
Feldman LS, Flum DR, Glasziou P, Maddern GJ, Marshall JC, McCulloch P, G, Orrego PS, Antoniucci D, De Luca L, Biondi-Zoccai GG, Crea F, Zijlstra F:
Nicholl J, Strasberg SM, Meakins JL, Ashby D, Black N, Bunker J, Burton M, Clinical impact of thrombectomy in acute ST-elevation myocardial
Campbell M, Chalkidou K, Chalmers I, de Leval M, et al: Challenges in infarction: an individual patient-data pooled analysis of 11 trials.
evaluating surgical innovation. Lancet 2009, 374:1097–1104. Eur Heart J 2009, 30:2193–2203.
3. Barkun JS, Aronson JK, Feldman LS, Maddern GJ, Strasberg SM, Balliol C, 23. Bonati LH, Dobson J, Algra A, Branchereau A, Chatellier G, Fraedrich G, Mali
Altman DG, Blazeby JM, Boutron IC, Campbell WB, Clavien PA, Cook JA, WP, Zeumer H, Brown MM, Mas JL, Ringleb PA: Short-term outcome after
Ergina PL, Flum DR, Glasziou P, Marshall JC, McCulloch P, Nicholl J, Reeves stenting versus endarterectomy for symptomatic carotid stenosis: a
BC, Seiler CM, Meakins JL, Ashby D, Black N, Bunker J, Burton M, Campbell preplanned meta-analysis of individual patient data. Lancet 2010,
M, Chalkidou K, Chalmers I, de Leval M, et al: Evaluation and stages of 376:1062–1073.
surgical innovations. Lancet 2009, 374:1089–1096. 24. Middleton LJ, Champaneria R, Daniels JP, Bhattacharya S, Cooper KG, Hilken
4. Sun X, Briel M, Busse JW, You JJ, Akl EA, Mejza F, Bala MM, Bassler D, Mertz NH, O’Donovan P, Gannon M, Gray R, Khan KS, International Heavy
D, Diaz-Granados N, Vandvik PO, Malaga G, Srinathan SK, Dahm P, Johnston Menstrual Bleeding Individual Patient Data Meta-analysis Collaborative
BC, Alonso-Coello P, Hassouneh B, Walter SD, Heels-Ansdell D, Bhatnagar N, Group, Abbot J, Barrington J, Bhattacharya S, Bongers MY, Brun JL, Busfield
Altman DG, Guyatt GH: Credibility of claims of subgroup effects in R, Sowter M, Clark TJ, Cooper J, Cooper KG, Corson SL, Dickersin K, Dwyer N,
randomised controlled trials: systematic review. BMJ 2012, 344:e1553. Gannon M, Hawe J, Hurskainen R, Meyer WR, O’Connor H, Pinion S, et al:
5. Bachmann LM, Puhan MA, ter Riet G, Bossuyt PM: Sample sizes of studies Hysterectomy, endometrial destruction, and levonorgestrel releasing
on diagnostic accuracy: literature survey. BMJ 2006, 332:1127–1129. intrauterine system (Mirena) for heavy menstrual bleeding: systematic
Hannink et al. Systematic Reviews 2013, 2:52 Page 12 of 12
http://www.systematicreviewsjournal.com/content/2/1/52

review and meta-analysis of data from individual patients. BMJ 2010, for spontaneous supratentorial intracerebral hemorrhage. Stroke 2012,
341:c3929. 43:1496–1504.
25. Mercado N, Wijns W, Serruys PW, Sigwart U, Flather MD, Stables RH, O'Neill 41. CG126 Stable angina: NICE guideline. ; 2013 [http://guidance.nice.org.uk/
WW, Rodriguez A, Lemos PA, Hueb WA, Gersh BJ, Booth J, Boersma E: CG126/NICEGuidance].
One-year outcomes of coronary artery bypass graft surgery versus 42. Koopman L, van der Heijden GJ, Glasziou PP, Grobbee DE, Rovers MM: A
percutaneous coronary intervention with multiple stenting for systematic review of analytical methods used to study subgroups in
multisystem disease: a meta-analysis of individual patient data from (individual patient data) meta-analyses. J Clin Epidemiol 2007,
randomized clinical trials. J Thorac Cardiovasc Surg 2005, 130:512–519. 60:1002–1009.
26. Boersma E: Does time matter? A pooled analysis of randomized clinical 43. Pocock SJ, Assmann SE, Enos LE, Kasten LE: Subgroup analysis, covariate
trials comparing primary percutaneous coronary intervention and adjustment and baseline comparisons in clinical trial reporting: current
in-hospital fibrinolysis in acute myocardial infarction patients. Eur Heart J practice and problems. Stat Med 2002, 21:2917–2930.
2006, 27:779–788. 44. Hernandez AV, Boersma E, Murray GD, Habbema JD, Steyerberg EW:
27. Timmer JR, Ottervanger JP, de Boer MJ, Boersma E, Grines CL, Westerhout Subgroup analyses in therapeutic cardiovascular clinical trials: are most
CM, Simes RJ, Granger CB, Zijlstra F: Primary percutaneous coronary of them misleading? Am Heart J 2006, 151:257–264.
intervention compared with fibrinolysis for myocardial infarction in 45. Bhandari M, Devereaux PJ, Li P, Mah D, Lim K, Schunemann HJ, Tornetta P
diabetes mellitus: results from the Primary Coronary Angioplasty vs 3rd: Misuse of baseline comparison tests and subgroup analyses in
Thrombolysis-2 trial. Arch Intern Med 2007, 167:1353–1359. surgical trials. Clin Orthop Relat Res 2006, 447:247–251.
28. de Boer SP, Westerhout CM, Simes RJ, Granger CB, Zijlstra F, Boersma E: 46. Parker AB, Naylor CD: Subgroups, treatment effects, and baseline risks:
Mortality and morbidity reduction by primary percutaneous coronary some lessons from major cardiovascular trials. Am Heart J 2000,
intervention is independent of the patient's age. JACC Cardiovasc Interv 139:952–961.
2010, 3:324–331. 47. Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM: Statistics in
29. de Boer SP, Barnes EH, Westerhout CM, Simes RJ, Granger CB, Kastrati A, medicine–reporting of subgroup analyses in clinical trials. N Engl J Med
Widimsky P, de Boer MJ, Zijlstra F, Boersma E: High-risk patients with 2007, 357:2189–2194.
ST-elevation myocardial infarction derive greatest absolute benefit from 48. Sun X, Heels-Ansdell D, Walter SD, Guyatt G, Sprague S, Bhandari M, Sanders
primary percutaneous coronary intervention: results from the Primary D, Schemitsch E, Tornetta P 3rd, Swiontkowski M: Is a subgroup claim
Coronary Angioplasty Trialist versus thrombolysis (PCAT)-2 collaboration. believable? A user’s guide to subgroup analyses in the surgical literature.
Am Heart J 2011, 161:500–507. e1. J Bone Joint Surg Am 2011, 93:e8.
30. Fox KA, Clayton TC, Damman P, Pocock SJ, de Winter RJ, Tijssen JG, 49. Fernandez YGE, Nguyen H, Duan N, Gabler NB, Kravitz RL: Assessing
Lagerqvist B, Wallentin L: Long-term outcome of a routine versus Heterogeneity of Treatment Effects: Are Authors Misinterpreting Their
selective invasive strategy in patients with non-ST-segment elevation Results. Health Serv Res 2010, 45:283–301.
acute coronary syndrome a meta-analysis of individual patient data. 50. Moreira ED Jr, Stein Z, Susser E: Reporting on methods of subgroup
J Am Coll Cardiol 2010, 55:2435–2445. analysis in clinical trials: a survey of four scientific journals. Braz J Med
31. Damman P, Clayton T, Wallentin L, Lagerqvist B, Fox KA, Hirsch A, Biol Res 2001, 34:1441–1446.
Windhausen F, Swahn E, Pocock SJ, Tijssen JG, de Winter RJ: Effects of age 51. Moreira ED, Susser E: Guidelines on how to assess the validity of results
on long-term outcomes after a routine invasive or selective invasive presented in subgroup analysis of clinical trials. Rev Hosp Clin Fac Med
strategy in patients presenting with non-ST segment elevation acute Sao Paulo 2002, 57:83–88.
coronary syndromes: a collaborative analysis of individual data from the 52. Sheldon TA, Cullum N, Dawson D, Lankshear A, Lowson K, Watt I, West P,
FRISC II - ICTUS - RITA-3 (FIR) trials. Heart 2012, 98:207–213. Wright D, Wright J: What's the evidence that NICE guidance has been
32. Damman P, Wallentin L, Fox KA, Windhausen F, Hirsch A, Clayton T, implemented? Results from a national evaluation using time series
Pocock SJ, Lagerqvist B, Tijssen JG, de Winter RJ: Long-term cardiovascular analysis, audit of patients’ notes, and interviews. BMJ 2004, 329:999.
mortality after procedure-related or spontaneous myocardial infarction
in patients with non-ST-segment elevation acute coronary syndrome: a doi:10.1186/2046-4053-2-52
collaborative analysis of individual patient data from the FRISC II, ICTUS, Cite this article as: Hannink et al.: A systematic review of individual
and RITA-3 trials (FIR). Circulation 2012, 125:568–576. patient data meta-analyses on surgical interventions. Systematic Reviews
33. Biau DJ, Katsahian S, Kartus J, Harilainen A, Feller JA, Sajovic M, Ejerhed L, 2013 2:52.
Zaffagnini S, Ropke M, Nizard R: Patellar tendon versus hamstring tendon
autografts for reconstructing the anterior cruciate ligament: a meta-
analysis based on individual patient data. Am J Sports Med 2009,
37:2470–2478.
34. Rovers MM, Black N, Browning GG, Maw R, Zielhuis GA, Haggard MP:
Grommets in otitis media with effusion: an individual patient data
meta-analysis. Arch Dis Child 2005, 90:480–485.
35. Salerno F, Camma C, Enea M, Rossle M, Wong F: Transjugular intrahepatic
portosystemic shunt for refractory ascites: a meta-analysis of individual
patient data. Gastroenterology 2007, 133:825–834.
36. Staples MP, Kallmes DF, Comstock BA, Jarvik JG, Osborne RH, Heagerty PJ,
Buchbinder R: Effectiveness of vertebroplasty using individual patient
data from two randomised placebo controlled trials: meta-analysis. Submit your next manuscript to BioMed Central
BMJ 2011, 343:d3952. and take full advantage of:
37. McCormack K, Scott NW, Go PM, Ross S, Grant AM: Laparoscopic
techniques versus open techniques for inguinal hernia repair. Cochrane
Database Syst Rev 2003, 1:CD001785. • Convenient online submission
38. Scott NW, McCormack K, Graham P, Go PM, Ross SJ, Grant AM: Open mesh • Thorough peer review
versus non-mesh for repair of femoral and inguinal hernia. • No space constraints or color figure charges
Cochrane Database Syst Rev 2002, 4:CD002197.
39. EU Hernia Trialists Collaboration: Repair of groin hernia with synthetic • Immediate publication on acceptance
mesh: meta-analysis of randomized controlled trials. Ann Surg 2002, • Inclusion in PubMed, CAS, Scopus and Google Scholar
235:322–332.
• Research which is freely available for redistribution
40. Gregson BA, Broderick JP, Auer LM, Batjer H, Chen XC, Juvela S,
Morgenstern LB, Pantazis GC, Teernstra OP, Wang WZ, Zuccarello M,
Mendelow AD: Individual patient data subgroup meta-analysis of surgery Submit your manuscript at
www.biomedcentral.com/submit