Open Access
Open Peer Review

This article has Open Peer Review reports available.

How does Open Peer Review work?

Systematic review of methods used in meta-analyses where a primary outcome is an adverse or unintended event

BMC Medical Research Methodology201212:64

DOI: 10.1186/1471-2288-12-64

Received: 14 June 2011

Accepted: 16 April 2012

Published: 3 May 2012

Abstract

Background

Adverse consequences of medical interventions are a source of concern, but clinical trials may lack power to detect elevated rates of such events, while observational studies have inherent limitations. Meta-analysis allows the combination of individual studies, which can increase power and provide stronger evidence relating to adverse events. However, meta-analysis of adverse events has associated methodological challenges. The aim of this study was to systematically identify and review the methodology used in meta-analyses where a primary outcome is an adverse or unintended event, following a therapeutic intervention.

Methods

Using a collection of reviews identified previously, 166 references including a meta-analysis were selected for review. At least one of the primary outcomes in each review was an adverse or unintended event. The nature of the intervention, source of funding, number of individual meta-analyses performed, number of primary studies included in the review, and use of meta-analytic methods were all recorded. Specific areas of interest relating to the methods used included the choice of outcome metric, methods of dealing with sparse events, heterogeneity, publication bias and use of individual patient data.

Results

The 166 included reviews were published between 1994 and 2006. Interventions included drugs and surgery among other interventions. Many of the references being reviewed included multiple meta-analyses with 44.6% (74/166) including more than ten. Randomised trials only were included in 42.2% of meta-analyses (70/166), observational studies only in 33.7% (56/166) and a mix of observational studies and trials in 15.7% (26/166). Sparse data, in the form of zero events in one or both arms where the outcome was a count of events, was found in 64 reviews of two-arm studies, of which 41 (64.1%) had zero events in both arms.

Conclusions

Meta-analyses of adverse events data are common and useful in terms of increasing the power to detect an association with an intervention, especially when the events are infrequent. However, with regard to existing meta-analyses, a wide variety of different methods have been employed, often with no evident rationale for using a particular approach. More specifically, the approach to dealing with zero events varies, and guidelines on this issue would be desirable.

Background

There is increasing concern regarding adverse outcomes of therapeutic interventions, which may be addressed by bringing together evidence from multiple sources [1]. The Cochrane Collaboration, in its remit to undertake systematic reviews, has considered adverse events as an area worthy of consideration as a defined clinical area that requires specific methodology [2], but the emphasis was placed on the non-statistical elements, such as choice of outcomes, study types and search strategies. The use of quantitative evidence synthesis methods with adverse events specifically in mind has been considered some time ago [3], but there is little empirical information collated on meta-analytic approaches within adverse event meta-analyses.

A recent review of systematic reviews of adverse events in response to drug interventions included 43 review articles published in 2006 [4]. However, the statistical methods could only be reviewed in the 24 of these that included a meta-analysis. This review also considered the search strategies used in the systematic reviews of adverse events, and this field has been better documented since it is the focus of several further studies [59].

The aim of this current systematic review is to investigate the meta-analytic methodology used where the primary outcome of the meta-analysis was an adverse or unintended event, with the aims of identifying which methods have been used and the challenges faced by authors. A wide range of therapeutic interventions are considered, including drugs, surgical procedures, devices, vaccinations, dietary interventions, anaesthetics and public health programmes. It is hoped that this review will assist in development and standardisation of methods in future and we discuss the areas of priority for further research to facilitate the development of guidelines to support meta-analysis of adverse and unintended outcomes data.

Methods

The dataset of meta-analyses for this current survey was identified using a database of systematic reviews collected previously [7]. The aim of this previous study was to develop and evaluate search strategies for the retrieval of systematic reviews, which may or may not include a meta-analysis, where the primary outcome was an adverse event (or unintended but positive event) resulting from a clinical intervention. The databases searched (both electronically and by hand) in the previous study were the Database of Abstracts of Reviews of Effects (DARE) and the Cochrane Database of Systematic Reviews (CDSR). Details of the search strategy have been described [7].

These search strategies yielded a total of 257 systematic reviews (246 from DARE plus 11 Cochrane reviews), published between 1994 and 2006. The initial searches were updated, yielding a further 20 systematic reviews on adverse events [9], producing 277 in total.

From the systematic reviews above, references were selected for inclusion in the current review according to the following criteria:
  1. 1.

    some form of quantitative synthesis (or test for heterogeneity with intention to perform a quantitative synthesis if appropriate) must be performed using more than one observed estimate of effect;

     
  2. 2.

    the study group of interest must have received some form of clinical intervention with intended or potential therapeutic effect; and

     
  3. 3.

    the full study report must be available in English.

     
A quantitative data synthesis may take the form of a pooled estimate, a confidence interval, quoting a p-value only, or performing a meta-regression. Studies entailing only qualitative evidence synthesis, although having an essential role to play in the assessment of adverse events, were not included in this review, which is exclusively aimed at statistical synthesis methods. Meta-analyses of unintended or adverse reactions associated with non-interventional activities, for example recreational drug use, were excluded.
Table 1

Number of included reviews by year of publication

Year of publication

No. included reviews (%)1

Published in journal

Cochrane review

Other publication type

1994

4 (2.4)

4

0

0

1995

8 (4.8)

8

0

0

1996

9 (5.4)

9

0

0

1997

15 (9.0)

15

0

0

1998

16 (9.6)

16

0

0

1999

19 (11.4)

18

1

0

2000

11 (6.6)

11

0

0

2001

22 (13.3)

22

0

0

2002

21 (12.7)

17

2

2

2003

22 (13.3)

20

2

0

2004

13 (7.8)

10

3

0

2005

4 (2.4)

0

4

0

2006

2 (1.2)

0

2

0

1Out of 166 included reviews.

Included reviews were evaluated on multiple aspects. These included type of intervention, number of primary studies incorporated, number of individual meta-analyses performed and source of funding (when determining sponsorship, any form of commercial funding was counted as such, even if other sources of funding were also present; similarly, government sponsorship took precedence over academic). The statistical aspects included general meta-analysis methods and how specific challenges were addressed. These included (i) dealing with sparse data (a count of zero events in one arm of study or zero events across the whole study); (ii) statistical heterogeneity; (iii) dissemination biases (primarily publication bias); (iv) variable quality of primary studies; and (v) use of individual patient data (IPD).

Results

Description of included reviews

In total, 166 reviews including a meta-analysis fulfilled all criteria and were included in this systematic review [10175]. Of these, 14 were Cochrane reviews, the others were published in a wide variety of medical journals or were reviews published by health agencies. Table 1 shows a breakdown of number of included reviews by publication year and type.

The interventions being evaluated for adverse events were similarly diverse; by far the largest category was drug interventions (69.9%, 116/166). The next largest category was surgical interventions (8.4%, 14/166). Other types of intervention, such as forms of anaesthesia, devices, and diagnostic procedures, each accounted for fewer than 5% of the included reviews. Multiple interventions (often multiple therapies for cancer being assessed simultaneously) were considered by 5.4% (9/166) of included reviews.
Table 2

Maximum and minimum numbers of contributing estimates for meta-analyses in the same included review

Maximum no. data observations1

Minimum no. data observations

No. included reviews (%)2

2−5

2−5

17 (10.4)

6−10

2−5

34 (20.7)

6−10

6−10

6 (3.7)

11−20

2−5

50 (30.5)

11−20

6−10

5 (3.0)

11−20

11−20

4 (2.4)

>20

2−5

29 (17.7)

>20

6−10

12 (7.3)

>20

11−20

3 (1.8)

>20

>20

4 (2.4)

1A single primary study may contribute more than one data observation, hence number of data observations may not be the same as number of primary studies.

2Out of 164 included reviews with at least one pooled estimate and number of data observations stated.

With regard to sponsorship, the largest number of included reviews were academically sponsored (45.8%; 76/166). Commercial sponsorship accounted for 16.3% (27/166), while 30.1% (50/166) were sponsored by some form of government body. Other sponsorship sources provided funding for 7.2% (12/166), while funding source was not stated for one included review.

Of the 166 included reviews, all except one included some overall combined estimate. The one included review that did not report a combined estimate included a meta-regression for dose–response with no overall pooled estimate of effect size.

The number of meta-analyses performed in each included review was also very variable. In many cases a large number of meta-analyses were performed, with 44.6% (74/166) of studies having more than 10 meta-analyses. By comparison, 19.3% (32/166) of studies had 6−10 meta-analyses, 29.5% (49/166) had 2–5 meta-analyses, 6.0% (10/166) had only one meta-analysis while 0.60% (1/166) had none (meta-regression only).

The number of primary cohorts (contributing a data observation to the meta-analysis) is shown in Table 2. This table indicates that many included reviews performed multiple meta-analyses with varying numbers of data observations derived from primary studies, possibly reflecting multiple outcomes or subgroup analyses.
Table 3

Percentages of included reviews by year including assessment of heterogeneity, quality and dissemination bias

Year

Total no. included reviews

Heterogeneity (no. (%))

Quality (no. (%))

Dissemination bias (no. (%))

1994

4

2 (50)

1 (25)

1 (25)

1995

8

7 (88)

1 (13)

4 (50)

1996

9

8 (89)

1 (11)

6 (67)

1997

15

12 (80)

8 (53)

5 (33)

1998

16

14 (88)

3 (19)

10 (63)

1999

19

18 (95)

8 (42)

12 (63)

2000

11

9 (82)

4 (36)

4 (36)

2001

22

19 (86)

8 (36)

15 (68)

2002

21

16 (76)

13 (62)

8 (38)

2003

22

16 (73)

10 (45)

15 (68)

2004

13

11 (85)

7 (54)

5 (38)

2005

4

4 (100)

4 (100)

3 (75)

2006

2

2 (100)

2 (100)

1 (50)

Statistical methodology aspects

Table 3 sets out the number of included reviews that cover heterogeneity, quality and dissemination bias, by year of publication, to examine any changes over time.

Outcome measures

Table 4 shows the proportions of included reviews using different effect measures. This is important because the choice of outcome measure may in itself influence the meta-analysis method and results.

In Table 4 the option ‘More than one’ was only selected in the eventuality that there was no obvious primary outcome metric, but instead at least two outcome metrics that appeared to receive approximately equal importance in the meta-analyses. Otherwise, the primary outcome measure was recorded even if there were other outcome metrics used in secondary analyses such as sensitivity analyses.

Many of the outcomes being reported in the primary studies were binary, thus lending themselves to analysis by odds ratio or relative risk, as seen in Table 4. A difference scale (for example risk difference), as opposed to a ratio scale, was chosen only infrequently. Only 9.6% of studies presented more than one outcome measure. This indicated that either more than one clinical outcome was being investigated, requiring a different outcome metric, or that the authors had presented analyses for the same clinical outcome on equal terms for different outcome metrics, possibly intended as a comparison between the two.
Table 4

Measure of effect

Measure of effect

No. included reviews

Percent included reviews1

Comparative measures (between interventions)

Odds ratio

55

33.1

Relative risk

51

30.7

Risk difference

8

4.8

Mean difference

6

3.6

Standardised mean difference

7

4.2

Percent difference

2

1.2

Non-comparative measures

Correlation

2

1.2

Probability (or percent)

13

7.8

Mean difference

2

1.2

Percent difference

3

1.8

Multiple measures

More than one

16

9.6

Other measures

Other

1

0.6

1Out of 166 included reviews.

Meta-analysis methodology

Another major area of interest was the methodology used for the meta-analysis itself. Table 5 sets out the meta-analysis method used to combine studies along with numbers and percentages.
Table 5

Meta-analysis methods

Meta-analysis method

No. included reviews

Percent included reviews1

Standard fixed effect2

54

32.5

Other fixed effect2

9

5.4

Standard random effects2

50

30.1

Marginal analysis

3

1.8

Bayesian methods

6

3.6

Multiple analysis methods

33

19.9

Other

1

0.6

Not stated

5

3.0

Meta-regression only

1

0.6

Heterogeneity test only

1

0.6

Unclear

3

1.8

1Out of 166 included reviews.

2See text for definition.

As seen in Table 5, fixed and random effect(s) models were used with roughly equal frequency. The term ‘standard fixed effect’ was used when the authors chose an accepted fixed effect model, such as the Mantel–Haenszel model [176], the inverse variance model or the Peto model [177]. Also, if the authors used any referenced fixed effect model this was recorded as ‘standard fixed effect’. In several cases the authors had used a method of combining data that would be considered as a ‘fixed effect’ model, but appeared to have either used a mean or weighted mean, or a logistic regression method, or had devised their own method for combining data, for example based on sample size of the studies. Such methods were considered non-standard and were recorded as ‘other fixed effect’. Interestingly, all the random effects models were referenced standard models (referred to as standard random effects in Table 5, such as the DerSimonian & Laird model [178]).

A record of ‘Multiple analysis methods’ was only selected when there was more than one method used on an equal basis. If there was an obvious primary method with additional supplementary methods the primary method was chosen.

The one meta-analysis reference referred to as ‘Other’ used a novel approach developed by the authors of ‘summary ranking’ involving assigning a score to the rank order of toxicity in individual primary studies, and then combining the scores to provide an overall rank order for toxicity. One study included a meta-regression as the only quantitative analysis, whilst another study performed a test for heterogeneity but did not produce any pooled estimates. In only three cases was there insufficient detail regarding the methodology to allow the type of analysis to be determined. In five of the reviewed studies the meta-analysis method was not stated.

A Bayesian approach was used by six included reviews, but in some cases the Bayesian model used in the meta-analysis was not fully described. Semi-Bayesian methods, incorporating a Bayesian use of a prior distribution on the rate of the adverse event alongside standard frequentist methods, were also used in one included review. This meta-analysis also experimented with the use of different prior distributions for the parameters. However, none of these studies presented graphical representations of the probability densities for the parameters.

Not taking into account the specific method of the meta-analysis, 74/166 (44.6%) used a fixed effect model. This included one review that stated that a random effects model was used, but in the absence of heterogeneity the presented results were fixed effect. (The novel method using rank summaries was also a fixed effect approach.) Three included reviews where the methodology was not described in sufficient detail to be classed as a standard or other fixed effect model, were, however, able to be classed as fixed effect in their approach.

A random effects model was used by 53/166 (31.9%) included reviews, including four of the reviews that used Bayesian methods. Hence, it is apparent that fixed and random effect(s) models were used with roughly equal frequency.

Both fixed and random effect(s) models were used in 28/166 (16.9%) included reviews, including one that used Bayesian methods. In the other cases it was not applicable (5/166; 3.0%), or not stated (5/166; 3.0%). It was unclear whether fixed or random effect(s) had been used in one included review only.

The reasons why the authors chose a particular model were recorded in 69/166 (41.6%) included reviews. Reasons based upon heterogeneity (or between-study variation) were the most commonly cited (46/69; 66.7%). Increased conservatism (of a random effects model) was also frequently mentioned; 7/69 (10.1%) reviews alluded to this.

Other reasons cited in support for a particular meta-analysis method included differences in primary study types, and so that larger studies would contribute more to the meta-analysis. Arriving at similar results from both fixed and random effect(s) models was also used to justify the chosen approach. Only one included review offered multiple explanations, while nine offered an explanation not mentioned above.

Type of primary study and approaches to inclusion

The types of primary study encountered by included reviews are set out in Table 6. The most frequent study type was some form of trial, the sole study type for 46.4% (77/166) of included reviews. Trials were clearly defined as being randomised and/or controlled in some included reviews. The 26 reviews (26/166, 15.7%) that included both trials and observational studies demonstrated a wide variety of approaches taken to this situation, often reflecting the number of each different type of study. Some included reviews made no attempt to differentiate by study design. In some included reviews there was only one trial, all other studies were observational, and in one of these reviews, the trial was excluded, although its inclusion did not alter the results. In another instance with only one trial, it was excluded from all meta-analyses, only the observational studies (of different designs) being included. In one included review the sole trial was excluded due to no events being observed in one group of the trial; similarly, in another review with only one trial, this study was excluded due to the small number of outcome events.
Table 6

Primary study types incorporated within included reviews

Study types

No. included reviews

% included reviews1

Randomised trials

70

42.2

Other trials2

7

4.2

Observational studies

56

33.7

Mixed (trials and

  

observational studies)

26

15.7

Not stated

7

4.2

1Out of 166 included reviews.

2May include randomised trials but not specifically stated as such.

The most common approach to mixed study types was to perform a sensitivity analysis by analysing all primary studies together and then dividing the studies by some element of study design. For example, one included review combined all studies together and then case–control studies were analysed separately, and cohort studies were combined with trials. Some variation on this theme was followed by several other included reviews. Another approach was to avoid combination of estimates across study designs altogether, by combining results from studies with similar designs.

Graphical representations of data

Graphical representations of data were used in the majority of included reviews. Forest plots were the only graph used in 53.0% (88/166) of reviews, while meta-regression plots were the only plot in 1.8% (3/166). Both forest plots and meta-regression plots appeared in 1.8% (3/166) of included reviews. Other plots were used in 18.1% (30/166) of included reviews, usually a plot of the individual studies but lacking a pooled estimate. Only 25.3% (42/166) of included reviews produced no graphical representations of their results.

Publication Bias

Publication bias was considered in 89/166 (53.6%) of included reviews. This issue may not have been specifically described in terms of publication bias. For example, some included reviews performed searches for unpublished primary studies, indicating that publication bias was within the awareness of the authors when performing a meta-analysis even if it was not taken any further than searching for such primary studies.

Publication bias was discussed but not formally evaluated in 44/166 (26.5%) included reviews. A quantitative analysis was performed by 31/166 (18.7%) included reviews. A sensitivity analysis by publication status was the preferred method of investigating publication bias for one review.

Of 31 included reviews with some form of quantitative analysis, 12 (38.7%) used a test with a p-value. The other 19/31 (61.3%) used an alternative method not resulting in a p-value. The most commonly used tests were Egger’s test [179] and Begg’s test [180]. Kendall’s tau test was mentioned by three included reviews.

The trim and fill method [181] was used to adjust for publication or selection bias in two included reviews. These were the only reviews that attempted to adjust for these types of bias. Graphical methods (funnel plots) were used to investigate for publication bias (or selection bias) in 29/166 reviews (17.5%).

The vast majority of included reviews used only published primary studies (129/166; 77.7%). Published primary studies with unpublished data (obtained through contact with the authors), were used in 20/166 (12.0%) included reviews. Both published and unpublished primary studies were used in 14/166 (8.4%) reviews. In the other reviews the study source(s) was either unclear or not stated. In several reviews where published primary studies only were included, it was made clear that unpublished data had been sought.

Heterogeneity

Heterogeneity was considered in some manner by 138/166 included reviews (83.1%), whether by a quantitative or qualitative assessment, or by a subgroup analysis or meta-regression. Meta-regression was included in nine reviews that had no other assessment of heterogeneity, whilst one review discussed issues regarding combination of primary studies with different criteria, but did not do a formal qualitative or quantitative analysis of heterogeneity. Of the remaining 128 included reviews, a quantitative assessment was performed in 124 (124/166; 74.7%). A qualitative assessment of heterogeneity (for example, inspection of forest plots or noting heterogeneous results) was made in 10/166 (6.0%) included reviews. Six included reviews incorporated both quantitative and qualitative aspects of heterogeneity assessment (6/166; 3.6%).

Considering quantitative analysis methods, 121/166 (72.9%) included reviews employed some form of statistical test for heterogeneity, although with variation in the chosen critical p-value for significance. The chosen significance value was 0.05 for 28 reviews (23.1% of the 121 studies that performed a test), while 23/121 chose a more liberal p-value of 0.1 (19.0%). Only one included review chose 0.2 as the cut-off p-value. In many reviews the actual p-value was quoted without reference to a particular threshold (51/121 reviews (42.1%) did this). In the other reviews no p-value or significance level was stated.

An estimate for heterogeneity was presented by 16/121 included reviews with a quantitative analysis (13.2%). The most frequently-used estimate measure was the I2statistic [182, 183], which was used in 13 reviews. Alternative estimate measures included the between-studies variance. One review used another estimate measure, the R(I) statistic [184]. Only one included review used multiple estimate measures; the estimates used included I2and the Q statistic [185].

Two ways to investigate the causes of heterogeneity are subgroup analysis and meta-regression. Subgroup analysis was performed in 27/128 (21.1%) reviews that included an analysis of heterogeneity, and in two reviews that did not formally assess heterogeneity. Meta-regression was used in 27 included reviews in total (27/166; 16.3%). In nine of these reviews, no formal assessment of heterogeneity had been performed. The covariates used in the meta-regression analyses were often very specific to the nature of the intervention or outcome being considered. A qualitative investigation of sources of heterogeneity was carried out in 17/138 reviews that considered heterogeneity in some way (12.3%).

Individual patient data

Very little use was made of IPD in the reviewed meta-analyses. Only two included reviews of the total 166 (1.2%) included IPD. Of these two, all primary studies included had IPD available (so there was no requirement to combine IPD and summary data). Both reviews used a one-stage method for the meta-analysis. In one review the meta-analysis was stratified by trial and other factors including centre within study for multicentre studies and age divisions. In the other review it was not stated whether the meta-analysis was stratified or not.

Sparse data

The issue of sparse data, whereby statistical methods were required to allow the inclusion of primary studies where the outcome was a count of zero, or a percent of zero, occurred in 65/166 included reviews (39.2%), one of which was a meta-analysis of single-arm studies only. Specific statistical methods may be required to allow incorporation of such a primary study into an overall pooled estimate or for calculation of confidence intervals. For 62 of the 65 reviews in which sparse data occurred, the outcome was on a comparative scale (an odds ratio, relative risk, or risk difference); this methodological area is the focus of this review. The primary issues are (i) outcome measures used by meta-analyses with zero events; (ii) use of continuity corrections; (iii) methods for inclusion of primary studies with zero events that do not involve continuity corrections; and (iv) incorporation of primary studies with zero events in both arms.

Of the 64 included reviews that considered the issue of sparse data in two-arm studies, 41 (64.1%) had datasets involving double-zero (zero events in both arms) primary studies. In the remaining 23 reviews, only single-zero (zero events in only one arm of a two-arm study) primary studies were present, or it was either unclear or not directly stated whether any double-zero studies were included within the dataset.

Of these 64 reviews, 30 (46.9%) presented their outcome as an odds ratio, 24 (37.5%) as a relative risk, and two as a risk difference. In six included reviews, there was more than one outcome with roughly equal importance in the meta-analyses. In two reviews, the outcome, where sparse data were incorporated, was an incidence rate.

Continuity corrections were used to circumvent problems with zero counts (in one or both arms of a study) that result in difficulties with estimating ratio-based outcome measures such as the odds ratio. Continuity corrections are also required to calculate the variance (and hence confidence intervals) for a risk difference. However, it was very difficult to determine an accurate picture of how continuity corrections were used.

Some included reviews clearly stated that continuity corrections had been used (15/64, 23.4%). In 17/64 (26.6%) included reviews, continuity corrections were not used. In 32 cases it was not clearly stated whether continuity corrections had been used or not.

The most popular primary continuity correction was 0.5, used in 14 of the 15 included reviews that stated their continuity correction. Only one meta-analysis reference used an alternative continuity correction with 0.25 being the chosen value. Only one review performed a sensitivity analysis across different continuity corrections, using 0.5, 0.1 and 0.01, and reported that the continuity correction did not alter the results. Three included reviews provided a reason for their choice of continuity correction, and the only reason cited was to minimise bias.

Several methods for incorporating single- and double-zero studies into a meta-analysis that do not involve continuity corrections were encountered in the included reviews. The most frequently used was the Peto method, employed by 12 studies. The use of a difference metric rather than a ratio as the outcome measure was used to circumvent problems with zeroes in seven reviews, although calculation of confidence intervals with such methods would be problematic. Seven included reviews resorted to the use of marginal analysis. Bayesian methods were used to tackle sparsity of events in only two included reviews.

Double-zero studies were included in analyses in 17 of the 41 reviews where double-zero primary studies were clearly present within the dataset. In two further reviews, double-zero studies were included in a sensitivity analysis. Double-zero studies were clearly excluded from analyses in 18 reviews where they were present in the set of primary studies. Such an exclusion was either a deliberate decision by the authors, shown by excluding the primary study in forest plots of the meta-analysis, or was done by default, the primary study being shown on a forest plot, but being given a weighting of zero. The four remaining included reviews were unclear as to whether or not these double-zero primary studies were included.

Of the 19 included reviews where double-zero primary studies were included in some way, seven made explicit use of continuity corrections. In other included reviews it was not clearly stated whether continuity corrections had been used. Other options, such as using a marginal analysis, were employed in some reviews, whilst in other reviews the methodologies used were unclear.

Discussion

This paper has reviewed an extensive sample of published meta-analyses where the primary outcome was an adverse or unintended event. Reviews published in languages other than English were excluded, hence there is a risk that methodologies used in reviews published in other languages may differ from those discussed here. There was considerable variation in the methodology employed across the sample, including designs of the primary studies incorporated in the meta-analyses, use of fixed or random effect(s) measures, how to deal with zero events in study arms, and how to assess study quality and publication bias. In many cases, the statistical aspects were not clearly reported, with insufficient detail to discern the methods used. Often, little justification was given for the approaches to meta-analysis used. This is perhaps, in part, due to the lack of specific guidelines available for meta-analysis of sparse/adverse events. Not only would such guidelines improve the standard of reporting of adverse event meta-analyses, they could also circumvent disagreements in the literature due to the use of alternative methods producing different conclusions, as was the case for the recent high-profile concerns of elevated cardiovascular risk in those taking rosiglitazone [186190].

Based on this review, one of the major areas of confusion appears to be with regard to the use of continuity corrections for dealing with sparse data for comparative outcomes. Indeed the term continuity correction is possibly misleading, as these are in effect nothing more than arbitrary factors added to a cell count of zero. Where a study has zero events in both groups, we believe it contains no information regarding the magnitude of the odds ratio or relative risk, but adding a continuity correction to both sides (incorrectly) keeps it in the analysis for methods which require it. Although not required for the risk difference scale, and such a study does contain information, a correction factor is required for the estimation of associated variance which causes further confusion. This latter point also raises the unanswered question of whether the presence of double-zero studies should influence the outcome metric. Bayesian methods present an alternative way to deal with datasets including sparse data, and Bayesian analyses are now easily implemented using appropriate software. However, Bayesian methods were used infrequently within the included reviews; with the development of guidelines to support Bayesian analyses, it may be that such methods would become more popular.

Many reviews in this sample included observational studies, in some cases this was the only type of primary study included, whilst in other reviews both observational studies and trials were included. Observational studies may offer advantages over trials, such as a longer period of follow-up. Inclusion of observational studies also increases the number of studies and individuals within a meta-analysis where both trials and observational studies are available, thus increasing power. However, this advantage may be counterbalanced by concerns regarding bias in observational studies. The reviews included here showed a variety of approaches to mixed study design; this is an area where guidelines would assist in combining all available data whilst addressing issues of different study design.

As mentioned in the introduction, a previous systematic review of reviews and meta-analyses of primary studies of adverse effects of a drug intervention has been conducted [4], including a total of 43 references, all published in 2006. Of these, 15% assessed quality of primary studies, compared to 42.2% of references in our review, and only 24 performed a meta-analysis. As seen in our review, there was some poor reporting of the methods used for pooling data, but 83% did report the method used for pooling data and exploring heterogeneity. With regard to funding source, 23% (of the 43 reviews) had pharmaceutical funding, compared to the 16.3% of references in our review that had commercial funding. Hence, there is potentially some suggestion of systematic differences between reviews concerning only drugs and those including other interventions, and between reviews that contain a meta-analysis and those that do not (although such observations could be confounded by the wider time range we considered).

Development of guidelines

Within the field of meta-analysis for adverse events, the concept of definitive guidelines is possibly too prescriptive; the diverse nature of medical interventions, the potential adverse outcomes and the ways they may be measured, and the formats in which data may be available preclude the use of standardised methods. However, general guidelines for approaching specific methodological issues (applying to other outcomes beyond adverse events) may be more feasible and useful. We hope this review sets the context in which future research and guidelines into the conduct of adverse event meta-analyses can be placed.

There is perhaps a need for more research before informed guidelines could be drafted. Many of the unique issues relating to adverse event meta-analyses are due to the typically sparse event data available for such analyses. Such sparse data presents unique challenges, as highlighted by this review. As well as specific challenges, such as dealing with zero events in arms of studies, there are also potential concerns relating to the use of broadly accepted meta-analysis methods in a sparse data context due to the potentially very low power such methods may have. For example, simulation has shown that tests for heterogeneity have very low power in sparse data situations [191, 192]. Similar issues are likely to exist for the use of meta-regression, tests for publication bias [193] and other methodologies. Rather than trying to use cutting-edge advanced meta-analysis methodology, in a sparse data context, it may often be wiser to restrict focus to simpler methods (e.g. fixed effect models) and be realistic about the potentially limited conclusions that can be drawn from the data. Since the primary aim of many adverse effect meta-analyses is to establish the existence of an elevated risk of an event due to a particular intervention, use of simple methods is, perhaps, consistent with this aim. Information on the performance of different meta-analytic estimators for sparse data and the use and avoidance of correction factors has been considered at length elsewhere [191, 192] and this is one domain in which knowledge on which methods to use and which to avoid is available.

Hierarchical models to address issues related to combination of different study types have been developed [194], and more recent developments on the use of methods to adjust studies in meta-analysis to account for bias appear to be promising [195, 196]. However, further exploration into the use of such methods in an adverse events context is warranted.

Development of Bayesian methods is clearly an area where further research would be both timely and beneficial, especially in the light of many of the difficulties surrounding meta-analysis of adverse events data, which Bayesian methods may be able to address, such as inclusion of primary studies with sparse events without the need for continuity corrections using Markov Chain Monte Carlo (MCMC) methods [197200]. However, difficulties exist in ensuring that all prior distributions are plausibly vague, when not based on external information – a challenge given how little data is sometimes available [201, 202].

Specifically to adverse events, an area where further research would be valuable is the consideration of drug class effects, allowing information on multiple drugs of the same class to be combined while acknowledging the potential differences in effects across drugs. This would be an area where hierarchical (and potentially Bayesian) models would be particularly useful [203]. Another clinical aspect related to adverse events issues is that there may be several indications for a certain intervention; patients with different conditions may be at varying risk of adverse events, despite receiving the same intervention. Whilst it is desirable to combine all available data to increase power, any meta-analysis should be able to adjust for the differences in indication for the intervention; as this scenario is similar to combination of data regarding individual drugs of the same class, hierarchical modelling may be a means to achieve this.

Conclusion

Conducting meta-analyses where the outcome is an unintended or adverse event presents a range of potential difficulties, and requires careful consideration of the statistical issues, as well as an awareness of the clinical context. This review has demonstrated that a diversity of approaches have been employed when conducting such meta-analyses. Hence, standardised guidelines may be beneficial in this area, even if, due to the range of clinical situations and availability and format of data, they are necessarily of a general nature. This is especially true since a meta-analysis may present the only feasible method to estimate potential risks, due to the often infrequent occurrence of adverse events within an individual trial or observational study.

Declarations

Acknowledgements

FCW performed this study whilst being supported by a studentship from the Department of Health Sciences, University of Leicester. KRA is partially supported by a National Institute for Health Research (NIHR) Senior Investigator Award (NI-SI-0508-10061). There were no sources of external funding for this study. The authors would like to thank the reviewers for their useful and supportive comments.

Authors’ Affiliations

(1)
Peninsula College of Medicine and Dentistry, St Luke’s Campus, University of Exeter
(2)
Department of Health Sciences, Adrian Building, University of Leicester
(3)
Centre for Reviews and Dissemination (CRD), University of York

References

  1. Loke YK, Price D, Herxheimer A, Cochrane Adverse Effects Methods Group: Systematic reviews of adverse effects: framework for a structured approach. BMC Medical Research Methodology. 2007, 7: 32-10.1186/1471-2288-7-32.PubMedPubMed CentralGoogle Scholar
  2. Loke YK, Price D, Herxheimer A: Adverse effects. Cochrane Handbook for Systematic Reviews of Interventions. Edited by: Higgins JPT, Green S. 2008, John Wiley and Sons, Chichester, UK, 433-449.Google Scholar
  3. Sutton A, Cooper N, Lambert P, Jones D, Abrams K, Sweeting M: Meta-analysis of rare and adverse event data. Expert Review of Pharmacoeconomics and Outcomes Research. 2002, 2 (4): 367-10.1586/14737167.2.4.367.PubMedGoogle Scholar
  4. Cornelius VR, Perrio MJ, Shakir SAW, Smith LA: Systematic reviews of adverse effects of drug interventions: a survey of their conduct and reporting quality. Pharmacoepidemiology and Drug Safety. 2009, 18 (12): 1223-10.1002/pds.1844.PubMedGoogle Scholar
  5. Derry S, Loke YK, Aronson JK: Incomplete evidence: the inadequacy of databases in tracing published adverse drug reactions in clinical trials. BMC Medical Research Methodology. 2001, 1: 7-10.1186/1471-2288-1-7.PubMedPubMed CentralGoogle Scholar
  6. Golder S, McIntosh HM, Duffy S, Glanville J: Developing efficient search strategies to identify reports of adverse effects in MEDLINE and EMBASE. Health Information and Libraries Journal. 2006, 23 (1): 3-10.1111/j.1471-1842.2006.00634.x.PubMedGoogle Scholar
  7. Golder S, McIntosh HM, Loke Y: Identifying systematic reviews of the adverse effects of health care interventions. BMC Medical Research Methodology. 2006, 6: 22-10.1186/1471-2288-6-22.PubMedPubMed CentralGoogle Scholar
  8. Golder S, Loke Y, McIntosh HM: Room for improvement? A survey of the methods used in systematic reviews of adverse effects. BMC Medical Research Methodology. 2006, 6: 3-10.1186/1471-2288-6-3.PubMedPubMed CentralGoogle Scholar
  9. Golder S, Loke Y, McIntosh HM: Poor reporting and inadequate searches were apparent in systematic reviews of adverse effects. Journal of Clinical Epidemiology. 2008, 61 (5): 440-10.1016/j.jclinepi.2007.06.005.PubMedGoogle Scholar
  10. Alfirevic Z, Sundberg K, Brigham S: Amniocentesis and chorionic villus sampling for prenatal diagnosis. Cochrane Database of Systematic Reviews. 2003, Issue 3(Art. No.: CD003252) [http://dx.doi.org/10.1002/14651858.CD003252],Google Scholar
  11. Allen DB, Mullen M, Mullen B: A meta-analysis of the effect of oral and inhaled corticosteroids on growth. Journal of Allergy and Clinical Immunology. 1994, 93 (6): 967-10.1016/S0091-6749(94)70043-5.PubMedGoogle Scholar
  12. Allison DB, Mentore JL, Heo M, Chandler LP, Cappelleri JC, Infante MC, Weiden PJ: Antipsychotic-induced weight gain: a comprehensive research synthesis. American Journal of Psychiatry. 1999, 156 (11): 1686-PubMedGoogle Scholar
  13. Anderson JW, Kendall CWC, Jenkins DJA: Importance of weight management in Type 2 diabetes: review with meta-analysis of clinical studies. Journal of the American College of Nutrition. 2003, 22 (5): 331-PubMedGoogle Scholar
  14. Anderson-Hanley C, Sherman ML, Riggs R, Agocha VB, Compas BE: Neuropsychological effects of treatments for adults with cancer: a meta-analysis and review of the literature. Journal of the International Neuropsychological Society. 2003, 9 (7): 967-PubMedGoogle Scholar
  15. Ashcroft DM, Chapman SR, Clark WK, Millson DS: Upper gastroduodenal ulceration in arthritis patients treated with celecoxib. Annals of Pharmacotherapy. 2001, 35 (7–8): 829-PubMedGoogle Scholar
  16. Ashraf E, Cooper S, Kellstein D, Jayawardena S: Safety profile of nonprescription ibuprofen in the elderly osteoarthritis patient: a meta-analysis. Inflammopharmacology. 2001, 9 (1–2): 35-Google Scholar
  17. Banks E: Hormone replacement therapy and the sensitivity and specificity of breast cancer screening: a review. Journal of Medical Screening. 2001, 8 (1): 29-10.1136/jms.8.1.29.PubMedGoogle Scholar
  18. Barbui C, Saraceno B: Low-dose neuroleptic therapy and extrapyramidal side effects in schizophrenia: an effect size analysis. European Psychiatry. 1996, 11 (8): 412-10.1016/S0924-9338(97)82580-1.PubMedGoogle Scholar
  19. Barone JE, Tucker JB, Rassias D, Corvo PR: Routine perioperative pulmonary artery catheterization has no effect on rate of complications in vascular surgery: a meta-analysis. American Surgeon. 2001, 67 (7): 674-PubMedGoogle Scholar
  20. Ben-David S, Einarson T, Ben-David Y, Nulman I, Pastuszak A, Koren G: The safety of nitrofurantoin during the first trimester of pregnancy: meta-analysis. Fundamental and Clinical Pharmacology. 1995, 9 (5): 503-10.1111/j.1472-8206.1995.tb00525.x.PubMedGoogle Scholar
  21. Bender BG, Berning S, Dudden R, Milgrom H, Tran ZV: Sedation and performance impairment of diphenhydramine and second-generation antihistamines: a meta-analysis. Journal of Allergy and Clinical Immunology. 2003, 111 (4): 770-10.1067/mai.2003.1408.PubMedGoogle Scholar
  22. Berglundh T, Persson L, Klinge B: A systematic review of the incidence of biological and technical complications in implant dentistry reported in prospective longitudinal studies of at least 5 years. Journal of Clinical Periodontology. 2002, 29 (Supplement 3): 197-PubMedGoogle Scholar
  23. Bernal-Delgado E, Latour-Perez J, Pradas-Arnal F, Gomez-Lopez LI: The association between vasectomy and prostate cancer: a systematic review of the literature. Fertility and Sterility. 1998, 70 (2): 191-10.1016/S0015-0282(98)00142-3.PubMedGoogle Scholar
  24. Brown JS, Sawaya G, Thom DH, Grady D: Hysterectomy and urinary incontinence: a systematic review. Lancet. 2000, 356 (9229): 535-10.1016/S0140-6736(00)02577-0.PubMedGoogle Scholar
  25. Brown S, Small R, Faber B, Krastev A, Davis P: Early postnatal discharge from hospital for healthy mothers and term infants. Cochrane Database of Systematic Reviews. 2002, Issue 3(Art. No.: CD002958) [http://dx.doi.org/10.1002/14651858.CD002958],Google Scholar
  26. Brumback BA, Holmes LB, Ryan LM: Adverse effects of chorionic villus sampling: a meta-analysis. Statistics in Medicine. 1999, 18 (16): 2163-10.1002/(SICI)1097-0258(19990830)18:16<2163::AID-SIM180>3.0.CO;2-H.PubMedGoogle Scholar
  27. Cannegieter SC, Rosendaal FR, Briet E: Thromboembolic and bleeding complications in patients with mechanical heart valve prostheses. Circulation. 1994, 89 (2): 635-10.1161/01.CIR.89.2.635.PubMedGoogle Scholar
  28. Caraballo PJ, Gabriel SE, Castro MR, Atkinson EJ, Melton LJ: Changes in bone density after exposure to oral anticoagulants: a meta-analysis. Osteoporosis International. 1999, 9 (5): 441-10.1007/s001980050169.PubMedGoogle Scholar
  29. Cardwell ME, Siviter G, Smith AF: Non-steroidal anti-inflammatory drugs and perioperative bleeding in paediatric tonsillectomy. Cochrane Database of Systematic Reviews. 2005, Issue 2(Art. No.: CD003591) [http://dx.doi.org/10.1002/14651858.CD003591.pub2],Google Scholar
  30. Caro-Paton T, Carvajal A, Martin de Diego I, Martin-Arias LH, Alvarez Requejo A, Rodriguez Pinilla E: Is metronidazole teratogenic: a meta-analysis. British Journal of Clinical Pharmacology. 1997, 44 (2): 179-PubMedPubMed CentralGoogle Scholar
  31. Chan WS, Ray J, Wai EK, Ginsburg S, Hannah ME, Corey PN, Ginsberg JS: Risk of stroke in women exposed to low-dose oral contraceptives: a critical evaluation of the evidence. Archives of Internal Medicine. 2004, 164 (7): 741-10.1001/archinte.164.7.741.PubMedGoogle Scholar
  32. Chang CH, Chen KY, Lai MY, Chan KA: Meta-analysis: ribavirin-induced haemolytic anaemia in patients with chronic hepatitis C. Alimentary Pharmacology and Therapeutics. 2002, 16 (9): 1623-10.1046/j.1365-2036.2002.01326.x.PubMedGoogle Scholar
  33. Chapron C, Fauconnier A, Goffinet F, Breart G, Dubuisson JB: Laparoscopic surgery is not inherently dangerous for patients presenting with benign gynaecologic pathology: results of a meta-analysis. Human Reproduction. 2002, 17 (5): 1334-10.1093/humrep/17.5.1334.PubMedGoogle Scholar
  34. Choi PT, Galinski SE, Takeuchi L, Lucas S, Tamayo C, Jadad AR: PDPH is a common complication of neuraxial blockade in parturients: a meta-analysis of obstetrical studies. Canadian Journal of Anaesthesia. 2003, 50 (5): 460-10.1007/BF03021057.PubMedGoogle Scholar
  35. Col NF, Hirota LK, Orr RK, Erban JK, Wong JB, Lau J: Hormone replacement therapy after breast cancer: a systematic review and quantitative assessment of risk. Journal of Clinical Oncology. 2001, 19 (8): 2357-PubMedGoogle Scholar
  36. Collaborative Group on Hormonal Factors in Breast Cancer: Breast cancer and hormonal contraceptives: collaborative reanalysis of individual data on 53 297 women with breast cancer and 100 239 women without breast cancer from 54 epidemiological studies. Lancet. 1996, 347 (9017): 1713-Google Scholar
  37. Correll CU, Leucht S, Kane JM: Lower risk for tardive dyskinesia associated with second-generation antipsychotics: a systematic review of 1-year studies. American Journal of Psychiatry. 2004, 161 (3): 414-10.1176/appi.ajp.161.3.414.PubMedGoogle Scholar
  38. Coughlin SS, Giustozzi A, Smith SJ, Lee NC: A meta-analysis of estrogen replacement therapy and risk of epithelial ovarian cancer. Journal of Clinical Epidemiology. 2000, 53 (4): 367-10.1016/S0895-4356(99)00179-1.PubMedGoogle Scholar
  39. Cutler C, Giri S, Jeyapalan S, Paniagua D, Viswanathan A, Antin JH: Acute and chronic graft-versus-host disease after allogeneic peripheral-blood stem-cell and bone marrow transplantation: a meta-analysis. Journal of Clinical Oncology. 2001, 19 (16): 3685-PubMedGoogle Scholar
  40. Derry S, Loke YK: Risk of gastrointestinal haemorrhage with long term use of aspirin: meta-analysis. BMJ. 2000, 321 (7270): 1183-10.1136/bmj.321.7270.1183.PubMedPubMed CentralGoogle Scholar
  41. Dezfulian C, Lavelle J, Nallamothu BK, Kaufman SR, Saint S: Rates of infection for single-lumen versus multilumen central venous catheters: a meta-analysis. Critical Care Medicine. 2003, 31 (9): 2385-10.1097/01.CCM.0000084843.31852.01.PubMedGoogle Scholar
  42. DiMatteo MR, Morton SC, Lepper HS, Damush TM, Carney MF, Pearson M, Kahn KL: Cesarean childbirth and psychosocial outcomes: a meta-analysis. Health Psychology. 1996, 15 (4): 303-PubMedGoogle Scholar
  43. Dolovich LR, Addis A, Vaillancourt RJ, Power BJ, Koren G, Einarson TR: Benzodiazepine use in pregnancy and major malformations of oral cleft: meta-analysis of cohort and case-control studies. BMJ. 1998, 317 (7162): 839-10.1136/bmj.317.7162.839.PubMedPubMed CentralGoogle Scholar
  44. Dong EW, Connelly JE, Borden SP, Yorzyk W, Passov DG, Kupelnick B, Luo DH, Ross SD: A systematic review and meta-analysis of the incidence of cancer in randomized, controlled trials of verapamil. Pharmacotherapy. 1997, 17 (6): 1210-PubMedGoogle Scholar
  45. Doren M, Nilsson JA, Johnell O: Effects of specific post-menopausal hormone therapies on bone mineral density in post-menopausal women: a meta-analysis. Human Reproduction. 2003, 18 (8): 1737-10.1093/humrep/deg315.PubMedGoogle Scholar
  46. Douketis JD, Ginsberg JS, Holbrook A, Crowther M, Duku EK, Burrows RF: A reevaluation of the risk for venous thromboembolism with the use of oral contraceptives and hormone replacement therapy. Archives of Internal Medicine. 1997, 157 (14): 1522-10.1001/archinte.1997.00440350022002.PubMedGoogle Scholar
  47. Doyle LW, Davis PG: Postnatal corticosteroids in preterm infants: systematic review of effects on mortality and motor function. Journal of Paediatrics and Child Health. 2000, 36 (2): 101-10.1046/j.1440-1754.2000.00481.x.PubMedGoogle Scholar
  48. Duffy G, Neal KR: Differences in post-operative infection rates between patients receiving autologous and allogeneic blood transfusion: a meta-analysis of published randomized and nonrandomized studies. Transfusion Medicine. 1996, 6 (4): 325-10.1111/j.1365-3148.1996.tb00091.x.PubMedGoogle Scholar
  49. Egger M, Davey Smith G, Stettler C, Diem P: Risk of adverse effects of intensified treatment in insulin-dependent diabetes mellitus: a meta-analysis. Diabetic Medicine. 1997, 14 (11): 919-10.1002/(SICI)1096-9136(199711)14:11<919::AID-DIA456>3.0.CO;2-A.PubMedGoogle Scholar
  50. Eikelboom JW, Mehta SR, Pogue J, Yusuf S: Safety outcomes in meta-analyses of phase 2 vs phase 3 randomized trials: intracranial hemorrhage in trials of bolus thrombolytic therapy. JAMA. 2001, 285 (4): 444-10.1001/jama.285.4.444.PubMedGoogle Scholar
  51. Etminan M, Samii A, Takkouche B, Rochon PA: Increased risk of somnolence with the new dopamine agonists in patients with Parkinson’s disease: a meta-analysis of randomised controlled trials. Drug Safety. 2001, 24 (11): 863-10.2165/00002018-200124110-00007.PubMedGoogle Scholar
  52. Etminan M, Gill S, Samii A: Comparison of the risk of adverse events with pramipexole and ropinirole in patients with Parkinson’s disease: a meta-analysis. Drug Safety. 2003, 26 (6): 439-10.2165/00002018-200326060-00005.PubMedGoogle Scholar
  53. Etminan M, Gill S, Samii A: Effect of non-steroidal anti-inflammatory drugs on risk of Alzheimer’s disease: systematic review and meta-analysis of observational studies. BMJ. 2003, 327 (7407): 128-10.1136/bmj.327.7407.128.PubMedPubMed CentralGoogle Scholar
  54. Fernandez E, La Vecchia, Balducci A, Chatenoud L, Franceschi S, Negri E: Oral contraceptives and colorectal cancer risk: a meta-analysis. British Journal of Cancer. 2001, 84 (5): 722-10.1054/bjoc.2000.1622.PubMedPubMed CentralGoogle Scholar
  55. Furberg CD, Psaty BM, Meyer JV: Nifedipine: dose-related increase in mortality in patients with coronary heart disease. Circulation. 1995, 92 (5): 1326-10.1161/01.CIR.92.5.1326.PubMedGoogle Scholar
  56. Gabriel Sánchez R, Carmona L, Roque M, Sánchez Gomez LM, Bonfill X: Hormone replacement therapy for preventing cardiovascular disease in post-menopausal women. Cochrane Database of Systematic Reviews. 2005, Issue 2(Art. No.: CD002229) [http://dx.doi.org/10.1002/14651858.CD002229.pub2],Google Scholar
  57. Garcia Rodriguez LA: Nonsteroidal antiinflammatory drugs, ulcers and risk: a collaborative meta-analysis. Seminars in Arthritis and Rheumatism. 1997, 26 (6 Supplement 1): 16-PubMedGoogle Scholar
  58. Garcia Rodriguez LA, Hernandez-Diaz S, de Abajo FJ: Association between aspirin and upper gastrointestinal complications: systematic review of epidemiologic studies. British Journal of Clinical Pharmacology. 2001, 52 (5): 563-10.1046/j.0306-5251.2001.01476.x.PubMedGoogle Scholar
  59. Garg PP, Kerlikowske K, Subak L, Grady D: Hormone replacement therapy and the risk of epithelial ovarian carcinoma: a meta-analysis. Obstetrics and Gynecology. 1998, 92 (3): 472-10.1016/S0029-7844(98)00139-2.PubMedGoogle Scholar
  60. Gillum LA, Mamidipudi SK, Claiborne Johnston S: Ischemic stroke risk with oral contraceptives: a meta-analysis. JAMA. 2000, 284 (1): 72-10.1001/jama.284.1.72.PubMedGoogle Scholar
  61. Gordon PV, Young ML, Marshall DD: Focal small bowel perforation: an adverse effect of early postnatal dexamethasone therapy in extremely low birth weight infants. Journal of Perinatology. 2001, 21 (3): 156-10.1038/sj.jp.7200520.PubMedGoogle Scholar
  62. Grady D, Gebretsadik T, Kerlikowske K, Ernster V, Petitti D: Hormone replacement therapy and endometrial cancer risk: a meta-analysis. Obstetrics and Gynecology. 1995, 85 (2): 304-10.1016/0029-7844(94)00383-O.PubMedGoogle Scholar
  63. Graham GD: Tissue plasminogen activator for acute ischemic stroke in clinical practice: a meta-analysis of safety data. Stroke. 2003, 34 (12): 2847-10.1161/01.STR.0000101752.23813.C3.PubMedGoogle Scholar
  64. Greenland S, Satterfield MH, Lanes SF: A meta-analysis to assess the incidence of adverse effects associated with the transdermal nicotine patch. Drug Safety. 1998, 18 (4): 297-10.2165/00002018-199818040-00005.PubMedGoogle Scholar
  65. Grodstein F, Newcomb PA, Stampfer MJ: Postmenopausal hormone therapy and the risk of colorectal cancer: a review and meta-analysis. American Journal of Medicine. 1999, 106 (5): 574-10.1016/S0002-9343(99)00063-7.PubMedGoogle Scholar
  66. Grossman E, Messerli FH: Effect of calcium antagonists on plasma norepinephrine levels, heart rate, and blood pressure. American Journal of Cardiology. 1997, 80 (11): 1453-10.1016/S0002-9149(97)00722-4.PubMedGoogle Scholar
  67. Guise JM, Berlin M, McDonagh M, Osterweil P, Chan B, Helfand M: Safety of vaginal birth after cesarean: a systematic review. Obstetrics and Gynecology. 2004, 103 (3): 420-10.1097/01.AOG.0000116259.41678.f1.PubMedGoogle Scholar
  68. Gupta P, Sachdev HP: Safety of oral use of nimesulide in children: systematic review of randomized controlled trials. Indian Pediatrics. 2003, 40 (6): 518-PubMedGoogle Scholar
  69. Hart RG, Benavente O, Pearce LA: Increased risk of intracranial hemorrhage when aspirin is combined with warfarin: a meta-analysis and hypothesis. Cerebrovascular Diseases. 1999, 9 (4): 215-10.1159/000015958.PubMedGoogle Scholar
  70. Hauth JC, Goldenberg RL, Parker CR, Cutter GR, Cliver SP: Low-dose aspirin: lack of association with an increase in abruptio placentae or perinatal mortality. Obstetrics and Gynecology. 1995, 85 (6): 1055-10.1016/0029-7844(95)00042-P.PubMedGoogle Scholar
  71. He J, Whelton PK, Vu B, Klag MJ: Aspirin and risk of hemorrhagic stroke: a meta-analysis of randomized controlled trials. JAMA. 1998, 280 (22): 1930-10.1001/jama.280.22.1930.PubMedGoogle Scholar
  72. Hébert-Croteau N: A meta-analysis of hormone replacement therapy and colon cancer in women. Cancer Epidemiology, Biomarkers & Prevention. 1998, 7 (8): 653-Google Scholar
  73. Heisel O, Heisel R, Balshaw R, Keown P: New onset diabetes mellitus in patients receiving calcineurin inhibitors: a systematic review and meta-analysis. American Journal of Transplantation. 2004, 4 (4): 583-10.1046/j.1600-6143.2003.00372.x.PubMedGoogle Scholar
  74. Henk JM: Controlled trials of synchronous chemotherapy with radiotherapy in head and neck cancer: overview of radiation morbidity. Clinical Oncology. 1997, 9 (5): 308-10.1016/S0936-6555(05)80064-0.PubMedGoogle Scholar
  75. Hennessy S, Berlin JA, Kinman JL, Margolis DJ, Marcus SM, Strom BL: Risk of venous thromboembolism from oral contraceptives containing gestodene and desogestrel versus levonorgestrel: a meta-analysis and formal sensitivity analysis. Contraception. 2001, 64 (2): 125-10.1016/S0010-7824(01)00234-7.PubMedGoogle Scholar
  76. Henry D, McGettigan P: Epidemiology overview of gastrointestinal and renal toxicity of NSAIDs. International Journal of Clinical Practice. 2003, 135 (Supplement): 43-PubMedGoogle Scholar
  77. Hoes AW, Grobbee DE, Peet TM, Lubsen J: Do non-potassium sparing diuretics increase the risk of sudden cardiac death in hypertensive patients: recent evidence. Drugs. 1994, 47 (5): 711-10.2165/00003495-199447050-00002.PubMedGoogle Scholar
  78. Huang JQ, Sridhar S, Hunt RH: Gastrointestinal safety profile of nabumetone: a meta-analysis. American Journal of Medicine. 1999, 107 (6A): 55S-PubMedGoogle Scholar
  79. Humphrey LL, Takano LMA, Chan BKS: Postmenopausal hormone replacement therapy and cardiovascular disease [Systematic evidence review no. 10]. 2002, Agency for Healthcare Research and Quality, Rockville, MDGoogle Scholar
  80. Hwang BF, Jaakkola JJ: Water chlorination and birth defects: a systematic review and meta-analysis. Archives of Environmental Health. 2003, 58 (2): 83-10.3200/AEOH.58.2.83-91.PubMedGoogle Scholar
  81. Impicciatore P, Choonara I, Clarkson A, Provasi D, Pandolfini C, Bonati M: Incidence of adverse drug reactions in paediatric in/out-patients: a systematic review and meta-analysis of prospective studies. British Journal of Clinical Pharmacology. 2001, 52 (1): 77-10.1046/j.0306-5251.2001.01407.x.PubMedPubMed CentralGoogle Scholar
  82. Ismail AI, Bandekar RR: Fluoride supplements and fluorosis: a meta-analysis. Community Dentistry and Oral Epidemiology. 1999, 27 (1): 48-10.1111/j.1600-0528.1999.tb01991.x.PubMedGoogle Scholar
  83. Ivanov R, Allen J, Calvin JE: The incidence of major morbidity in critically ill patients managed with pulmonary artery catheters: a meta-analysis. Critical Care Medicine. 2000, 28 (3): 615-10.1097/00003246-200003000-00002.PubMedGoogle Scholar
  84. Janowsky EC, Kupper LL, Hulka BS: Meta-analysis of the relation between silicone breast implants and the risk of connective tissue diseases. New England Journal of Medicine. 2000, 342 (11): 781-10.1056/NEJM200003163421105.PubMedGoogle Scholar
  85. Johnson AG, Nguyen TV, Day RO: Do nonsteroidal anti-inflammatory drugs affect blood pressure: a meta-analysis. Annals of Internal Medicine. 1994, 121 (4): 289-PubMedGoogle Scholar
  86. Johnston SC, Colford JJM, Gress DR: Oral contraceptives and the risk of subarachnoid hemorrhage: a meta-analysis. Neurology. 1998, 51 (2): 411-10.1212/WNL.51.2.411.PubMedGoogle Scholar
  87. Jolles BM, Bogoch ER: Posterior versus lateral surgical approach for total hip arthroplasty in adults with osteoarthritis. Cochrane Database of Systematic Reviews. 2006, Issue 3(Art. No.: CD003828) [http://dx.doi.org/10.1002/14651858.CD003828.pub3],Google Scholar
  88. Jones G, Riley M, Couper D, Dwyer T: Water fluoridation, bone mass and fracture: a quantitative overview of the literature. Australian and New Zealand Journal of Public Health. 1999, 23 (1): 34-10.1111/j.1467-842X.1999.tb01202.x.PubMedGoogle Scholar
  89. Jonker-Pool: Sexual functioning after treatment for testicular cancer: review and meta-analysis of 36 empirical studies between 1975–2000. Archives of Sexual Behavior. 2001, 30 (1): 55-10.1023/A:1026468707362.PubMedGoogle Scholar
  90. Kellstein DE, Waksman JA, Furey SA, Binstok G, Cooper SA: The safety profile of nonprescription ibuprofen in multiple-dose use: a meta-analysis. Journal of Clinical Pharmacology. 1999, 39 (5): 520-PubMedGoogle Scholar
  91. Kemmeren JM, Algra A, Grobbee DE: Third generation oral contraceptives and risk of venous thrombosis: meta-analysis. BMJ. 2001, 323 (7305): 131-10.1136/bmj.323.7305.131.PubMedPubMed CentralGoogle Scholar
  92. Khader YS, Rice J, John L, Abueita O: Oral contraceptives use and the risk of myocardial infarction: a meta-analysis. Contraception. 2003, 68 (1): 11-10.1016/S0010-7824(03)00073-8.PubMedGoogle Scholar
  93. Ko DT, Hebert PR, Coffey CS, Sedrakyan A, Curtis JP, Krumholz HM: Beta-blocker therapy and symptoms of depression, fatigue, and sexual dysfunction. JAMA. 2002, 288 (3): 351-10.1001/jama.288.3.351.PubMedGoogle Scholar
  94. Ko DT, Hebert PR, Coffey CS, Curtis JP, Foody JM, Sedrakyan A, Krumholz HM: Adverse effects of beta-blocker therapy for patients with heart failure: a quality overview of randomized trials. Archives of Internal Medicine. 2004, 164 (13): 1389-10.1001/archinte.164.13.1389.PubMedGoogle Scholar
  95. Kongnyuy EJ, Norman RJ, Flight IHK, Rees MCP: Oestrogen and progestogen hormone replacement therapy for peri-menopausal and post-menopausal women: weight and body fat distribution. Cochrane Database of Systematic Reviews. 1999, Issue 3(Art. No.: CD001018) [http://dx.doi.org/10.1002/14651858.CD001018],Google Scholar
  96. Kozer E, Nikfar S, Costei A, Boskovic R, Nulman I, Koren G: Aspirin consumption during the first trimester of pregnancy and congenital anomalies: a meta-analysis. American Journal of Obstetrics and Gynecology. 2002, 187 (6): 1623-10.1067/mob.2002.127376.PubMedGoogle Scholar
  97. Lazarou J, Pomeranz BH, Corey PN: Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospective studies. JAMA. 1998, 279 (15): 1200-10.1001/jama.279.15.1200.PubMedGoogle Scholar
  98. LeBlanc ES, Janowsky J, Chan BK, Nelson HD: Hormone replacement therapy and cognition: systematic review and meta-analysis. JAMA. 2001, 285 (11): 1489-10.1001/jama.285.11.1489.PubMedGoogle Scholar
  99. Lee TM, Yip JT, Jones-Gotman M: Memory deficits after resection from left or right anterior temporal lobe in humans: a meta-analytic review. Epilepsia. 2002, 43 (3): 283-10.1046/j.1528-1157.2002.09901.x.PubMedGoogle Scholar
  100. Lee A, Cooper MC, Craig JC, Knight JF, Keneally JP: Effects of nonsteroidal anti-inflammatory drugs on postoperative renal function in adults with normal renal function. Cochrane Database of Systematic Reviews. 2004, Issue 2(Art No.: CD002765) [http://dx.doi.org/10.1002/14651858.CD002765],Google Scholar
  101. Leipzig RM, Cumming RG, Tinetti ME: Drugs and falls in older people: a systematic review and meta-analysis. I. Psychotropic drugs. Journal of the American Geriatrics Society. 1999, 47 (1): 30-PubMedGoogle Scholar
  102. Leipzig RM, Cumming RG, Tinetti ME: Drugs and falls in older people: a systematic review and meta-analysis. II. Cardiac and analgesic drugs. Journal of the American Geriatrics Society. 1999, 47 (1): 40-PubMedGoogle Scholar
  103. Lethaby A, Suckling J, Barlow D, Farquhar CM, Jepson RG, Roberts H: Hormone replacement therapy in postmenopausal women: endometrial hyperplasia and irregular bleeding. Cochrane Database of Systematic Reviews. 2004, Issue 3(Art. No.: CD000402) [http://dx.doi.org/10.1002/14651858.CD000402],Google Scholar
  104. Lipworth BJ: Systemic adverse effects of inhaled corticosteroid therapy: a systematic review and meta-analysis. Archives of Internal Medicine. 1999, 159 (9): 941-10.1001/archinte.159.9.941.PubMedGoogle Scholar
  105. Liu EH, Sia AT: Rates of Caesarean section and instrumental vaginal delivery in nulliparous women after low concentration epidural infusions or opioid analgesia: systematic review. BMJ. 2004, 328 (7453): 1410-10.1136/bmj.38097.590810.7C.PubMedPubMed CentralGoogle Scholar
  106. Loke YK, Derry S, Pritchard-Copley A: Appetite suppressants and valvular heart disease: a systematic review. BMC Clinical Pharmacology. 2002, 2: 6-PubMedPubMed CentralGoogle Scholar
  107. MacLennan SC, MacLennan AH, Ryan P: Colorectal cancer and oestrogen replacement therapy: a meta-analysis of epidemiological studies. Medical Journal of Australia. 1995, 162 (9): 491-PubMedGoogle Scholar
  108. Marcolongo R, Frediani B, Biasi G, Minari C, Barreca C: A meta-analysis of the tolerability of amtolmetin guacil, a novel, effective nonsteroidal anti-inflammatory drug, compared with established agents. Clinical Drug Investigation. 1999, 17 (2): 89-10.2165/00044011-199917020-00002.Google Scholar
  109. Mardirosoff C, Dumont L, Boulvain M, Tramer MR: Fetal bradycardia due to intrathecal opioids for labour analgesia: a systematic review. BJOG: an International Journal of Obstetrics and Gynaecology. 2002, 109 (3): 274-10.1111/j.1471-0528.2002.01380.x.Google Scholar
  110. McAlister FA, Clark HD, Wells PS, Laupacis A: Perioperative allogeneic blood transfusion does not cause adverse sequelae in patients with cancer: a meta-analysis of unconfounded studies. British Journal of Surgery. 1998, 85 (2): 171-10.1046/j.1365-2168.1998.00698.x.PubMedGoogle Scholar
  111. Mehta SR, Eikelboom JW, Yusuf S: Risk of intracranial haemorrhage with bolus versus infusion thrombolytic therapy: a meta-analysis. Lancet. 2000, 356 (9228): 449-10.1016/S0140-6736(00)02552-6.PubMedGoogle Scholar
  112. Miller J, Chan BK, Nelson H: Hormone replacement therapy and risk of venous thromboembolism. [Systematic evidence review no. 11]. 2002, Agency for Healthcare Research and Quality, Rockville, MDGoogle Scholar
  113. Moiniche S, Romsing J, Dahl JB, Tramer MR: Nonsteroidal antiinflammatory drugs and the risk of operative site bleeding after tonsillectomy: a quantitative systematic review. Anesthesia and Analgesia. 2003, 96 (1): 68-PubMedGoogle Scholar
  114. Mol BW, Ankum WM, Bossuyt PM, Van der Veen: Contraception and the risk of ectopic pregnancy: a meta-analysis. Contraception. 1995, 52 (6): 337-10.1016/0010-7824(95)00221-9.PubMedGoogle Scholar
  115. Muldoon MF, Manuck SB, Mendelsohn AB, Kaplan JR, Belle SH: Cholesterol reduction and non-illness mortality: meta-analysis of randomised clinical trials. BMJ. 2000, 322 (7277): 11-Google Scholar
  116. Nalysnyk L, Fahrbach K, Reynolds MW, Zhao SZ, Ross S: Adverse events in coronary artery bypass graft (CABG) trials: a systematic review and analysis. Heart. 2003, 89 (7): 767-10.1136/heart.89.7.767.PubMedPubMed CentralGoogle Scholar
  117. Nanda K, Bastian LA, Hasselblad V, Simel DL: Hormone replacement therapy and the risk of colorectal cancer: a meta-analysis. Obstetrics and Gynecology. 1999, 93 (5 Part 2 Supplement S): 880-PubMedGoogle Scholar
  118. Nazareth I, Lewin J, King M: Sexual dysfunction after treatment for testicular cancer: a systematic review. Journal of Psychosomatic Research. 2001, 51 (6): 735-10.1016/S0022-3999(01)00282-3.PubMedGoogle Scholar
  119. O’Connell D, Robertson J, Henry D, Gillespie W: A systematic review of the skeletal effects of estrogen therapy in postmenopausal women. II. An assessment of treatment effects. Climacteric. 1998, 1 (2): 112-10.3109/13697139809085526.PubMedGoogle Scholar
  120. Ofman JJ, MacLean CH, Straus WL, Morton SC, Berger ML, Roth EA, Shekelle P: A metaanalysis of severe upper gastrointestinal complications of nonsteroidal antiinflammatory drugs. Journal of Rheumatology. 2002, 29 (4): 804-PubMedGoogle Scholar
  121. Ofman JJ, Maclean CH, Straus WL, Morton SC, Berger ML, Roth EA, Shekelle PG: Meta-analysis of dyspepsia and nonsteroidal antiinflammatory drugs. Arthritis and Rheumatism. 2003, 49 (4): 508-10.1002/art.11192.PubMedGoogle Scholar
  122. Oger E, Scarabin PY: Assessment of the risk for venous thromboembolism among users of hormone replacement therapy. Drugs and Aging. 1999, 14 (1): 55-10.2165/00002512-199914010-00004.PubMedGoogle Scholar
  123. Olsen O: Meta-analysis of the safety of home birth. Birth. 1997, 24 (1): 4-10.1111/j.1523-536X.1997.tb00330.x.PubMedGoogle Scholar
  124. Park-Wyllie L, Mazzotta P, Pastuszak A, Moretti ME, Beique L, Hunnisett L, Friesen MH, Jacobson S, Kasapinovic S, Chang D, Diav-Citrin O, Chitayat D, Nulman I, Einarson TR, Koren G: Birth defects after maternal exposure to corticosteroids: prospective cohort study and meta-analysis of epidemiological studies. Teratology. 2000, 62 (6): 385-10.1002/1096-9926(200012)62:6<385::AID-TERA5>3.0.CO;2-Z.PubMedGoogle Scholar
  125. Pfahlberg A, Hassan K, Wille L, Lausen B, Gefeller O: Systematic review of case-control studies: oral contraceptives show no effect on melanoma risk. Public Health Reviews. 1997, 25 (3–4): 309-PubMedGoogle Scholar
  126. Pladevall-Vila M, Delclos GL, Varas C, Guyer H, Brugues-Tarradellas J, Anglada-Arisa A: Controversy of oral contraceptives and risk of rheumatoid arthritis: meta-analysis of conflicting studies and review of conflicting meta-analyses with special emphasis on analysis of heterogeneity. American Journal of Epidemiology. 1996, 144 (1): 1-10.1093/oxfordjournals.aje.a008846.PubMedGoogle Scholar
  127. Raaijmakers E, Engelen AM: Is sensorineural hearing loss a possible side effect of nasopharyngeal and parotid irradiation: a systematic review of the literature. Radiotherapy and Oncology. 2002, 65 (1): 1-10.1016/S0167-8140(02)00211-6.PubMedGoogle Scholar
  128. Ray JG, O’Brien TE, Chan WS: Preconception care and the risk of congenital anomalies in the offspring of women with diabetes mellitus: a meta-analysis. QJM: An International Journal of Medicine. 2001, 94 (8): 435-10.1093/qjmed/94.8.435.Google Scholar
  129. Reid FD, Mercer PM, Harrison M, Bates T: Cholecystectomy as a risk factor for colorectal cancer: a meta-analysis. Scandinavian Journal of Gastroenterology. 1996, 31 (2): 160-10.3109/00365529609031981.PubMedGoogle Scholar
  130. Reynolds F, Sharma SK, Seed PT: Analgesia in labour and fetal acid-base balance: a meta-analysis comparing epidural with systemic opioid analgesia. BJOG: an International Journal of Obstetrics and Gynaecology. 2002, 109 (12): 1344-Google Scholar
  131. Robinson JW, Dufour MS, Fung TS: Erectile functioning of men treated for prostate carcinoma. Cancer. 1997, 79 (3): 538-10.1002/(SICI)1097-0142(19970201)79:3<538::AID-CNCR16>3.0.CO;2-4.PubMedGoogle Scholar
  132. Ross SD, Kupelnick B, Kumashiro M, Arellano FM, Mohanty N, Allen IE: Risk of serious adverse events in hypertensive patients receiving isradipine: a meta-analysis. Journal of Human Hypertension. 1997, 11 (11): 743-10.1038/sj.jhh.1000532.PubMedGoogle Scholar
  133. Ross SD, Akhras KS, Zhang S, Rozinsky M, Nalysnyk L: Discontinuation of antihypertensive drugs due to adverse events: a systematic review and meta-analysis. Pharmacotherapy. 2001, 21 (8): 940-10.1592/phco.21.11.940.34520.PubMedGoogle Scholar
  134. Rothwell PM, Slattery J, Warlow CP: A systematic comparison of the risks of stroke and death due to carotid endarterectomy for symptomatic and asymptomatic stenosis. Stroke. 1996, 27 (2): 266-10.1161/01.STR.27.2.266.PubMedGoogle Scholar
  135. Rothwell PM, Slattery J, Warlow CP: A systematic review of the risks of stroke and death due to endarterectomy for symptomatic carotid stenosis. Stroke. 1996, 27 (2): 260-10.1161/01.STR.27.2.260.PubMedGoogle Scholar
  136. Ruesch S, Walder B, Tramer MR: Complications of central venous catheters. Internal jugular versus subclavian access: a systematic review. Critical Care Medicine. 2002, 30 (2): 454-10.1097/00003246-200202000-00031.PubMedGoogle Scholar
  137. Rutschmann OT, McCrory DC, Matchar DB: Immunization and MS: a summary of published evidence and recommendations. Neurology. 2002, 59 (12): 1837-10.1212/WNL.59.12.1837.PubMedGoogle Scholar
  138. Sachdev M, Miller WC, Ryan T, Jolis JG: Effect of fenfluramine-derivative diet pills on cardiac valves: a meta-analysis of observational studies. American Heart Journal. 2002, 144 (6): 1065-10.1067/mhj.2002.126733.PubMedGoogle Scholar
  139. Safdar N, Said A, Gangnon RE, Maki DG: Risk of hemolytic uremic syndrome after antibiotic treatment of Escherichia coli 0157:H7 enteritis: a meta-analysis. JAMA. 2002, 288 (8): 996-10.1001/jama.288.8.996.PubMedGoogle Scholar
  140. Sagsveen M, Farmer JE, Prentice A, Breeze A: Gonadotrophin-releasing hormone analogues for endometriosis: bone mineral density. Cochrane Database of Systematic Reviews. 2003, Issue 4(Art. No.: CD001297) [http://dx.doi.org/10.1002/14651858.CD001297],Google Scholar
  141. Sakai H, Hayashi K, Origasa H, Kusunoki T: An application of meta-analysis techniques in the evaluation of adverse experiences with antihypertensive agents. Pharmacoepidemiology and Drug Safety. 1999, 8 (3): 169-10.1002/(SICI)1099-1557(199905/06)8:3<169::AID-PDS416>3.0.CO;2-F.PubMedGoogle Scholar
  142. Salpeter SR, Ormiston TM, Salpeter EE, Wood-Baker R: Cardioselective beta-blockers for reversible airway disease. Cochrane Database of Systematic Reviews. 2002, Issue 4(Art. No.: CD002992) [http://dx.doi.org/10.1002/14651858.CD002992],Google Scholar
  143. Salpeter SR, Greyber E, Pasternak GA, Salpeter EE: Risk of fatal and nonfatal lactic acidosis with metformin use in type 2 diabetes mellitus. Archives of Internal Medicine. 2003, 163 (21): 2594-10.1001/archinte.163.21.2594.PubMedGoogle Scholar
  144. Salpeter SR, Walsh JM, Greyber E, Ormiston TM, Salpeter EE: Mortality associated with hormone replacement therapy in younger and older women: a meta-analysis. Journal of General Internal Medicine. 2004, 19 (7): 791-10.1111/j.1525-1497.2004.30281.x.PubMedPubMed CentralGoogle Scholar
  145. Salpeter SR, Ormiston TM, Salpeter EE: Cardioselective beta-blockers for chronic obstructive pulmonary disease. Cochrane Database of Systematic Reviews. 2005, Issue 4(Art. No.: CD003566) [http://dx.doi.org/10.1002/14651858.CD003566.pub2],Google Scholar
  146. Salpeter SR, Greyber E, Pasternak GA, Salpeter EE: Risk of fatal and nonfatal lactic acidosis with metformin use in type 2 diabetes mellitus. Cochrane Database of Systematic Reviews. 2006, Issue 1(Art. No.: CD002967) [http://dx.doi.org/10.1002/14651858.CD002967.pub2],Google Scholar
  147. Schaumberg DA, Dana MR, Christen WG, Glynn RJ: A systematic overview of the incidence of posterior capsule opacification. Ophthalmology. 1998, 105 (7): 1213-10.1016/S0161-6420(98)97023-3.PubMedGoogle Scholar
  148. Schoenfeld P: Gastrointestinal safety profile of meloxicam: a meta-analysis and systematic review of randomized controlled trials. American Journal of Medicine. 1999, 107 (6A): 48S-PubMedGoogle Scholar
  149. Serebruany VL, Malinin AI, Eisert RM, Sane DC: Risk of bleeding complications with antiplatelet agents: meta-analysis of 338,191 patients enrolled in 50 randomized controlled trials. American Journal of Hematology. 2004, 75 (1): 40-10.1002/ajh.10451.PubMedGoogle Scholar
  150. Seto A, Einarson T, Koren G: Pregnancy outcome following first trimester exposure to antihistamines: meta-analysis. American Journal of Perinatology. 1997, 14 (3): 119-10.1055/s-2007-994110.PubMedGoogle Scholar
  151. Smith JS, Green J, Berrington de Gonzalez A, Appleby P, Peto J, Plummer M, Franceschi S, Beral V: Cervical cancer and use of hormonal contraceptives: a systematic review. Lancet. 2003, 361 (9364): 1159-10.1016/S0140-6736(03)12949-2.PubMedGoogle Scholar
  152. Sneyd JR, Carr A, Byrom WD, Bilski AJ: A meta-analysis of nausea and vomiting following maintenance of anaesthesia with propofol or inhalational agents. European Journal of Anaesthesiology. 1998, 15 (4): 433-PubMedGoogle Scholar
  153. Stason WB, Schmid CH, Niedzwiecki D, Whiting GW, Caubet JF, Luo D, Ross SD, Chalmers TC: Safety of nifedipine in patients with hypertension: a meta-analysis. Hypertension. 1997, 30 (1 Part 1): 7-PubMedGoogle Scholar
  154. Stason WB, Schmid CH, Niedzwiecki D, Whiting GW, Caubet JF, Cory D, Luo D, Ross SD, Chalmers TC: Safety of nifedipine in angina pectoris: a meta-analysis. Hypertension. 1999, 33 (1): 24-10.1161/01.HYP.33.1.24.PubMedGoogle Scholar
  155. Torgerson DJ, Bell-Syer SEM: Hormone replacement therapy and prevention of nonvertebral fractures: a meta-analysis of randomized trials. JAMA. 2001, 285 (22): 2891-10.1001/jama.285.22.2891.PubMedGoogle Scholar
  156. Torgerson DJ, Bell-Syer SEM: Hormone replacement therapy and prevention of vertebral fractures: a meta-analysis of randomised trials. BMC Musculoskeletal Disorders. 2001, 2: 7-10.1186/1471-2474-2-7.PubMedPubMed CentralGoogle Scholar
  157. Torvaldsen S, Roberts CL, Bell JC, Raynes-Greenow CH: Discontinuation of epidural analgesia late in labour for reducing the adverse delivery outcomes associated with epidural analgesia. Cochrane Database of Systematic Reviews. 2004, Issue 4(Art. No.: CD004457) [http://dx.doi.org/10.1002/14651858.CD004457.pub2],Google Scholar
  158. Tramer MR, Moore RA, Reynolds DJ, McQuay HJ: Quantitative estimation of rare adverse events which follow a biological progression: a new model applied to chronic NSAID use. Pain. 2000, 85 (1–2): 169-PubMedGoogle Scholar
  159. Trindade E, Menon D, Topfer LA, Coloma C: Adverse effects associated with selective serotonin reuptake inhibitors and tricyclic antidepressants: a meta-analysis. Canadian Medical Association Journal. 1998, 159 (10): 1245-PubMedPubMed CentralGoogle Scholar
  160. Uzzan B, Campos J, Cucherat M, Nony P, Boissel JP, Perret GY: Effects on bone mass of long term treatment with thyroid hormones: a meta-analysis. Journal of Clinical Endocrinology and Metabolism. 1996, 81 (12): 4278-10.1210/jc.81.12.4278.PubMedGoogle Scholar
  161. Vamvakas EC: Perioperative blood transfusion and cancer recurrence: meta-analysis for explanation. Transfusion. 1995, 35 (9): 760-10.1046/j.1537-2995.1995.35996029162.x.PubMedGoogle Scholar
  162. Vamvakas EC: Meta-analysis of randomized controlled trials investigating the risk of postoperative infection in association with white blood cell-containing allogeneic blood transfusion: the effects of the type of transfused red blood cell product and surgical setting. Transfusion Medicine Reviews. 2002, 16 (4): 304-10.1053/tmrv.2002.35209.PubMedGoogle Scholar
  163. Vamvakas EC: WBC-containing allogeneic blood transfusion and mortality: a meta-analysis of randomized controlled trials. Transfusion. 2003, 43 (7): 963-10.1046/j.1537-2995.2003.00426.x.PubMedGoogle Scholar
  164. Vercauteren SB, Bosmans JL, Elseviers MM, Verpooten GA, De Broe: A meta-analysis and morphological review of cyclosporine-induced nephrotoxicity in auto-immune diseases. Kidney International. 1998, 54 (2): 536-10.1046/j.1523-1755.1998.00017.x.PubMedGoogle Scholar
  165. Verhoeven AC, Boers M: Limited bone loss due to corticosteroids: a systematic review of prospective studies in rheumatoid arthritis and other diseases. Journal of Rheumatology. 1997, 24 (8): 1495-PubMedGoogle Scholar
  166. von Dadelszen P, Ornstein MP, Bull SB, Logan AG, Koren G, Magee LA: Fall in mean arterial pressure and fetal growth restriction in pregnancy hypertension: a meta-analysis. Lancet. 2000, 355 (9198): 87-10.1016/S0140-6736(98)08049-0.PubMedGoogle Scholar
  167. Vorperian VR, Havighurst TC, Miller S, January CT: Adverse effects of low dose amiodarone: a meta-analysis. Journal of the American College of Cardiology. 1997, 30 (3): 791-10.1016/S0735-1097(97)00220-9.PubMedGoogle Scholar
  168. Warltier DCE, Marret E, Flahault A, Samama CM, Bonnet F: Effects of postoperative, nonsteroidal, antiinflammatory drugs on bleeding risk after tonsillectomy: meta-analysis of randomized, controlled trials. Anesthesiology. 2003, 98 (6): 1497-10.1097/00000542-200306000-00030.Google Scholar
  169. Waterman EJ, Magee LA, Lim KI, Skoll A, Rurak D, von Dadelszen P: Do commonly used oral antihypertensives alter fetal or neonatal heart rate characteristics: a systematic review. Hypertension in Pregnancy. 2004, 23 (2): 155-10.1081/PRG-120028291.PubMedGoogle Scholar
  170. Wehling M: Meta-analysis of flecainide safety in patients with supraventricular arrhythmias. Arzneimittel-Forschung Drug Research. 2002, 52 (7): 507-PubMedGoogle Scholar
  171. Wiffen P, Gill M, Edwards J, Moore A: Adverse drug reactions in hospital patients: a systematic review of the prospective and retrospective studies. Bandolier Extra. 2002, 101: 4-Google Scholar
  172. Wilson K, Mottram P: A comparison of side effects of selective serotonin reuptake inhibitors and tricyclic antidepressants in older depressed patients: a meta-analysis. International Journal of Geriatric Psychiatry. 2004, 19 (8): 754-10.1002/gps.1156.PubMedGoogle Scholar
  173. Xiong X, Buekens P, Wollast E: IUD use and the risk of ectopic pregnancy: a meta-analysis of case-control studies. Contraception. 1995, 52 (1): 23-10.1016/0010-7824(95)00120-Y.PubMedGoogle Scholar
  174. Yaffe K, Sawaya G, Lieberburg I, Grady D: Estrogen therapy in postmenopausal women: effects on cognitive function and dementia. JAMA. 1998, 279 (9): 688-10.1001/jama.279.9.688.PubMedGoogle Scholar
  175. Zaric DDZ, Christiansen CCC, Pace NL, Punjasawadwong Y: Transient neurologic symptoms (TNS) following spinal anaesthesia with lidocaine versus other local anaesthetics. Cochrane Database of Systematic Reviews. 2005, Issue 4(Art. No.: CD003006) [http://dx.doi.org/10.1002/14651858.CD003006.pub2],Google Scholar
  176. Mantel N, Haenszel W: Statistical aspects of the analysis of data from retrospective studies of disease. Journal of the National Cancer Institute. 1959, 22 (4): 719-PubMedGoogle Scholar
  177. Yusuf S, Peto R, Lewis J, Collins R, Sleight P: Beta blockade during and after myocardial infarction: an overview of the randomized trials. Progress in Cardiovascular Diseases. 1985, 27 (5): 353-Google Scholar
  178. DerSimonian R, Laird N: Meta-analysis in clinical trials. Controlled clinical trials. 1986, 7 (3): 177-10.1016/0197-2456(86)90046-2.PubMedGoogle Scholar
  179. Egger M, Davey Smith G, Schneider M, Minder C: Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997, 315 (7109): 629-10.1136/bmj.315.7109.629.PubMedPubMed CentralGoogle Scholar
  180. Begg CB, Mazumdar M: Operating characteristics of a rank correlation test for publication bias. Biometrics. 1994, 50 (4): 1088-10.2307/2533446.PubMedGoogle Scholar
  181. Duval S, Tweedie R: Trim and fill: A simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000, 56 (2): 455-10.1111/j.0006-341X.2000.00455.x.PubMedGoogle Scholar
  182. Higgins JPT, Thompson SG: Quantifying heterogeneity in a meta-analysis. Statistics in Medicine. 2002, 21 (11): 1539-10.1002/sim.1186.PubMedGoogle Scholar
  183. Higgins JPT, Thompson SG, Deeks JJ, Altman DG: Measuring inconsistency in meta-analyses. BMJ. 2003, 327 (7414): 557-10.1136/bmj.327.7414.557.PubMedPubMed CentralGoogle Scholar
  184. Takkouche B, Cadarso-Suárez C, Spiegelman D: Evaluation of old and new tests of heterogeneity in epidemiologic meta-analysis. American Journal of Epidemiology. 1999, 150 (2): 206-10.1093/oxfordjournals.aje.a009981.PubMedGoogle Scholar
  185. Berlin JA, Laird NM, Sacks HS, Chalmers TC: A comparison of statistical methods for combining event rates from clinical trials. Statistics in Medicine. 1989, 8 (2): 141-10.1002/sim.4780080202.PubMedGoogle Scholar
  186. Nissen S, Wolski K: Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes. New England Journal of Medicine. 2007, 356 (24): 2457-10.1056/NEJMoa072761.PubMedGoogle Scholar
  187. Diamond GA, Box L, Kaul S: Uncertain effects of rosiglitazone on the risk for myocardial infarction and cardiovascular death. Annals of Internal Medicine. 2007, 147 (8): 578-PubMedGoogle Scholar
  188. Bracken MB: Rosiglitazone and cardiovascular risk. New England Journal of Medicine. 2007, 357 (9): 937-PubMedGoogle Scholar
  189. Cobitz A, Zambanini A, Sowell M, Heise M, Louridas B, McMorn S, Semigran M, Koch G: A retrospective evaluation of congestive heart failure and myocardial ischemia events in 14 237 patients with type 2 diabetes mellitus enrolled in 42 short-term, double-blind, randomized clinical studies with rosiglitazone. Pharmacoepidemiology and Drug Safety. 2008, 17 (8): 769-10.1002/pds.1615.PubMedGoogle Scholar
  190. Nissen SE, Wolski K: Rosiglitazone revisited: an updated meta-analysis of risk for myocardial infarction and cardiovascular mortality. Archives of Internal Medicine. 2010, 170 (14): 1191-10.1001/archinternmed.2010.207.PubMedGoogle Scholar
  191. Sweeting MJ, Sutton AJ, Lambert PC: What to add to nothing? Use and avoidance of continuity corrections in meta-analysis of sparse data. Statistics in Medicine. 2004, 23 (9): 1351-10.1002/sim.1761.PubMedGoogle Scholar
  192. Bradburn MJ, Deeks JJ, Berlin JA, Localio AR: Much ado about nothing: a comparison of the performance of meta-analytical methods with rare events. Statistics in Medicine. 2007, 26 (1): 53-10.1002/sim.2528.PubMedGoogle Scholar
  193. Ioannidis JPA, Trikalinos TA: The appropriateness of asymmetry tests for publication bias in meta-analyses: a large survey. Canadian Medical Association Journal. 2007, 176 (8): 1091-10.1503/cmaj.060410.PubMedPubMed CentralGoogle Scholar
  194. Prevost TC, Abrams KR, Jones DR: Hierarchical models in generalized synthesis of evidence: an example based on studies of breast cancer screening. Statistics in Medicine. 2000, 19 (24): 3359-10.1002/1097-0258(20001230)19:24<3359::AID-SIM710>3.0.CO;2-N.PubMedGoogle Scholar
  195. Turner RM, Spiegelhalter DJ, Smith GCS, Thompson SG: Bias modelling in evidence synthesis. Journal of the Royal Statistical Society Series A - Statistics in Society. 2009, 172 (1): 21-Google Scholar
  196. Welton NJ, Ades AE, Carlin JB, Altman DG, Sterne JAC: Models for potentially biased evidence in meta-analysis using empirically based priors. Journal of the Royal Statistical Society Series A - Statistics in Society. 2009, 172 (1): 119-Google Scholar
  197. Carlin JB: Meta-analysis for 2 × 2 tables - a Bayesian approach. Statistics in Medicine. 1992, 11 (2): 141-10.1002/sim.4780110202.PubMedGoogle Scholar
  198. Ashby D, Hutton JL, McGee MA: Simple Bayesian analyses for case-control studies in cancer epidemiology. Statistician. 1993, 42 (4): 385-10.2307/2348472.Google Scholar
  199. Ashby D, Hutton JL: Bayesian epidemiology. 1996, , New York, NY, USAGoogle Scholar
  200. Sutton AJ, Abrams KR: Bayesian methods in meta-analysis and evidence synthesis. Statistical Methods in Medical Research. 2001, 10 (4): 277-10.1191/096228001678227794.PubMedGoogle Scholar
  201. Spiegelhalter D, Abrams K, Myles J: Bayesian approaches to clinical trials and healthcare evaluation. 2004, John Wiley & Sons, Chichester, UKGoogle Scholar
  202. Lambert PC, Sutton AJ, Burton PR, Abrams KR, Jones DR: How vague is vague? A simulation study of the impact of the use of vague prior distributions in MCMC using WinBUGS. Statistics in Medicine. 2005, 24 (15): 2401-10.1002/sim.2112.PubMedGoogle Scholar
  203. Warren FC: An exploration of evidence synthesis methods for adverse events. PhD thesis. 2010, University of Leicester, Department of Health SciencesGoogle Scholar
  204. Pre-publication history

    1. The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2288/12/64/prepub

Copyright

© Warren et al.; licensee BioMed Central Ltd. 2012

This article is published under license to 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.

Advertisement