Summary
Background
It is often assumed that differences in the efficacy of treatments between countries (or regions) will be neither negligible nor minor and therefore cannot be overlooked when assessing the potential benefit of treatments in one country (or region) on the basis of trials conducted in another country (or region).
Aim
To assess differences in the results of cardiovascular trials between Europe and North America on the basis of data from an extensive collection of trials.
Methods
A systematic search was conducted of Medline (from the year 2005 to 2008) and the Cochrane Library (from 2000 to 2008) for all meta-analyses of randomized controlled trials aimed at treating and preventing cardiovascular disease. Within each meta-analysis that satisfied given criteria, trial results were compared between Europe and North America with respect to a fatal and/or non-fatal endpoint by forming separate estimates of treatment efficacy for each of these continents.
Results
The literature search found 59 meta-analyses that satisfied all the inclusion criteria. For most meta-analyses, it was the case that relative to the control, the intervention was more favoured in trials conducted in Europe than in North America with regard to both fatal endpoints (28 out of 43 meta-analyses) and non-fatal endpoints (28 out of 40 meta-analyses). However, it was only with regard to non-fatal endpoints that this imbalance turned out to be statistically significant at the 5% level ( P = 0.017). Also, the lack of statistically significant differences in trial results between Europe and North America within individual meta-analyses meant that it was not possible to determine for which types of intervention these intercontinental differences are likely to be more pronounced than others.
Conclusion
There is some evidence to support the theory that, relative to controls, interventions are more favoured in cardiovascular trials conducted in Europe than in North America, when treatment efficacy is measured in terms of a non-fatal endpoint. However, the overall support for systematic differences in cardiovascular trial results between Europe and North America is weak, which may be surprising given the amount of data collected.
Résumé
Background
Il est souvent supposé que les différences dans l’efficacité du traitement entre les pays ou régions ne sont ni négligeables ni mineures, et donc ne devraient pas être négligées lorsque l’on évalue les bénéfices potentiels d’un traitement dans un pays ou une région sur la base d’essais cliniques conduits dans un autre traitement ou région par extrapolation.
Objectifs
Évaluer les différences des résultats des essais cliniques en pathologie cardiovasculaire entre l’Europe et l’Amérique du Nord sur la base des données d’une collection extensive des essais cliniques publiés.
Méthodes
Une recherche systématique a été conduite sur Medline entre 2005 et 2008 et sur la base Cochrane entre 2000 et 2008 pour toutes les méta-analyses des essais contrôlés randomisés, ayant pour objectif de traiter ou de prévenir une pathologie cardiovasculaire. Au sein de chaque méta-analyse satisfaisant les critères définis, les résultats des essais ont été comparés entre l’Europe et l’Amérique du Nord pour ce qui concerne le critère de jugement fatal et/ou non-fatal, en estimant de façon séparée l’efficacité des traitements dans chacun des continents.
Résultats
La recherche effectuée dans la littérature a trouvé 59 méta-analyses satisfaisants les critères d’inclusion de l’étude. Pour la majorité des méta-analyses, il a été vérifié que comparativement aux témoins, l’intervention était plus favorable dans les essais conduits en Europe, plutôt qu’en Amérique du Nord pour ce qui concerne les critères de jugement fatal (28 des 43 méta-analyses) et non-fatal (28 des 40 méta-analyses). Cependant, pour les seuls critères de jugement non-fatal, la différence était statistiquement significative au seuil de 5 % ( p = 0,017). Par ailleurs, l’absence de différence statistiquement significative dans les résultats des essais entre l’Europe et l’Amérique du nord au sein des méta-analyses individuelles a indiqué qu’il n’était pas possible de déterminer pour quel type d’intervention ces différences intercontinentales étaient probablement plus prononcées que d’autres.
Conclusion
Il y a donc certains arguments pour confirmer la théorie que comparativement aux témoins, les interventions sont plus favorables dans les essais cliniques cardiovasculaires conduits en Europe, et non en Amérique du Nord, lorsque l’efficacité du traitement est mesurée en termes de critères de jugement non-fatal. Cependant, la preuve globale pour des différences systématiques entre les essais cardiovasculaires entre l’Europe et l’Amérique du Nord est faible, ce qui est surprenant compte du volume des données collectées.
Background
A question that often arises when attempting to interpret the observed efficacy of treatments within clinical trials, is whether the results of trials conducted in one country can be transferred to other countries or continents. The assumption is often made that differences in the efficacy of treatments between countries (or regions) will be neither negligible nor minor and therefore cannot be overlooked when assessing the potential benefit of treatments in one country (or region) on the basis of trials conducted in another country (or region) . However, much of the justification for this sentiment comes from anecdotal evidence, the analysis of a very limited pool of data or the analysis of trial results that have not been obtained through a full systematic review .
A popular way of statistically assessing heterogeneity in treatment effects within clinical trials is to perform a subgroup analysis . Therefore, one approach to assessing differences in treatment efficacy between countries is to perform a subgroup analysis of a multinational trial with results broken down according to country . However, one difficulty with such a strategy is that the data produced by a single large trial may not be sufficient to reveal such differences, even when they are genuinely quite large. Also, a quick literature search is sufficient to show that such between-country subgroup analyses are rarely published and the authors of the articles concerned may not be willing or able to hand over the required information when it is requested.
An alternative approach would be to compare the results of clinical trials conducted in one country with the results of similar trials conducted in another country, with the matching of trials determined by published systematic reviews; this is the approach that was explored in the present study. In particular, our aim was to assess differences in the results of randomized controlled trials (RCTs) between Europe and North America. This was achieved by categorizing RCTs within published systematic reviews according to the countries in which the trials were conducted. The clinical area that we focused on was cardiovascular disease (CVD); this specialty was chosen due to the rich variety of interventions that are commonly used to treat and prevent this disease (i.e. interventions ranging from the use of drugs, surgery and devices to changes to diet and lifestyle).
Two major obstacles need to be overcome when using systematic reviews to try to gauge the size of international differences in treatment efficacy: these differences are rarely assessed within the reviews themselves; and a comprehensive list of the countries in which each trial within any given review was conducted is rarely provided. Moreover, it is often the case that the country in which a trial took place is not reported even in the article that originally describes the principal findings of the trial. Therefore, without the kind of thorough search for country information that was carried out in the present study, it would be very difficult for a clinician or researcher to form a general picture about international variations in clinical efficacy over a wide range of different interventions.
Methods
A literature search of Medline and the Cochrane Library was conducted for all meta-analyses of RCTs indexed by the MESH term ‘cardiovascular diseases’. Meta-analyses found through this search were included in this study if three criteria were satisfied. First, the meta-analyses had to have been published in the 9 years between January 2000 and December 2008 for the Cochrane Library search and in the 4 years between January 2005 and December 2008 for the Medline search. A longer search period was used for the Cochrane Library as it is a database that generally only contains meta-analyses that are the most up-to-date and of the highest quality and yet, as it is a much smaller database than Medline, it can be thoroughly searched without requiring an excessive amount of resources. The articles containing the meta-analyses had to be written in English.
Second, the patients taking part in the RCTs within each meta-analysis had to be adults. With regard to therapeutic interventions, the intervention had to be aimed at treating a problem with the heart or a blood vessel in the chest cavity. With regard to preventive interventions, the patients concerned had to have CVD or be at high risk of CVD.
Third, for at least one of the outcomes reported, results had to be combined over at least two trials published in or after 1990 that were conducted solely in Europe and at least two trials published in or after 1990 that were conducted solely in North America.
The countries in which trials were conducted were determined on the basis of what was stated in the articles that originally presented the trial results. If such country information was not given, then the country (or countries) where a trial was conducted was determined by contacting the authors directly or through the addresses of the authors themselves, but only if the authors were all based in the same country. North America was defined as the USA and Canada. Countries belonging to Europe were determined according to the definition of Europe used by the statistics division of the United Nations . Trials published before 1990 were excluded in order to focus the study on the measurement of international differences in treatment efficacy that may presently exist, rather than the measurement of such differences in the past. No language restriction was placed on articles that originally presented trial results.
For each meta-analysis satisfying the above criteria, the most commonly reported fatal endpoint and non-fatal endpoint were determined. The rule used to achieve this was that in the meta-analysis concerned, more trials were combined over that fatal (non-fatal) endpoint than any other fatal (non-fatal) endpoint. For example, if for any given meta-analysis there were two non-fatal endpoints and five trials were combined over one of these endpoints while four trials were combined over the other endpoint, then the most commonly reported non-fatal endpoint for that meta-analysis would be defined as the former endpoint. Ties were settled by choosing the endpoint associated with the highest number of patients.
Within each meta-analysis, results for the selected endpoints were separately combined within North America and within Europe to give individual estimates of treatment efficacy both for North America and for Europe with respect to both a fatal and a non-fatal endpoint. However, this was only done if two conditions were satisfied. First, in relation to the given endpoint, a combined estimate of efficacy over all continents was presented as part of the meta-analysis concerned. Second, there were results from at least two trials in Europe and at least two trials in North America for the given endpoint.
The first condition ensured that this study carried out no truly new meta-analyses, but instead produced only new subgroup breakdowns for existing meta-analyses. The second condition allowed results to be combined using a random-effects model . This is more than just a minor technical issue, as combining trial results using a random-effects rather than a fixed-effect model allows heterogeneity between trials conducted in the same continent or region to be taken into account. For example, if a trial conducted in Europe and a trial conducted in North America show vastly different results, it would be impossible to determine without further information whether, for instance, a second trial conducted in Europe would show similar results to the one conducted in North America. Clearly, there is the need for the results of at least two trials conducted in a given region in order to take account of such heterogeneity between trials within that region.
If an endpoint was event based and event rates were available, then trial results were combined over relative risks, even if trial results were reported in terms of odds ratios in the meta-analysis concerned. However, if event rates were not available and the meta-analysis reported trial results in terms of odds ratios, then trial results were combined over odds ratios. In other words, where possible, the relative risk was used to measure treatment efficacy.
It should be noted that although fatal and non-fatal endpoints were allowed to vary over different meta-analyses, the same fatal endpoint (e.g. cardiovascular death within 1 year) or non-fatal endpoint (e.g. non-fatal myocardial infarction within 1 year) was used to compare trials within any given meta-analysis. Therefore, since comparisons of trial results for Europe and North America were only conducted within individual meta-analyses, a large variation over different meta-analyses in the types of fatal and non-fatal endpoints being analysed was of only secondary concern.
Results
The literature search found 221 meta-analyses that satisfied the first and second inclusion criteria. Of these meta-analyses, 162 meta-analyses did not have the required number of trials published in or after 1990 in the relevant geographical areas for the third criterion to be satisfied. Although the exclusion of trials published before 1990 may seem to be quite a restrictive criterion, 179 of the initially identified 221 meta-analyses did not contain any trials published before 1990.
In summary, 59 meta-analyses were found that satisfied all the inclusion criteria of the study. Of these meta-analyses, 47 meta-analyses contained a sufficient number of trials to allow separate estimates of treatment efficacy for the most commonly reported fatal endpoints to be calculated both for Europe and for North America. Details regarding these meta-analyses are given in Tables 1 and 2 . On the other hand, 44 out of the 59 meta-analyses included in the study contained a sufficient number of trials to allow such estimates to be calculated with respect to the most commonly reported non-fatal endpoints. These meta-analyses are listed in Tables 3–5 .
| Authors, year, Reference | Intervention | Control | Endpoint | No. of trials (Europe/N. Amer.) | Relative risk estimate (95% CI) | Best region a ( P value of difference) | |
|---|---|---|---|---|---|---|---|
| Europe | N. Amer. | ||||||
| Hood et al., 2004 | Digitalis for CHF | Placebo | All-cause mortality | 2/3 | 0.55 (0.12, 2.53) | 0.99 (0.93, 1.06) | Europe (0.447) |
| Hooper et al., 2004 | Omega 3 fatty acids for CVD | Placebo, no supplement or usual diet | Cardiovascular mortality | 8/2 | 1.00 (0.73, 1.38) | 0.25 (0.03, 2.24) | N. Amer. (0.219) |
| Rees et al., 2004 | Psychological interventions for CHD | Usual care or no intervention | Cardiovascular mortality | 2/6 | 0.70 (0.32, 1.52) | 0.93 (0.67, 1.28) | Europe (0.505) |
| Amsallem et al., 2005 | Phosphodiesterase III inhibitors for HF | Placebo | Cardiovascular mortality | 2/5 | 1.55 (1.01, 2.37) | 1.03 (0.72, 1.49) | N. Amer. (0.160) |
| Krum et al., 2005 | Beta-blockers without ACE inhibitors | Placebo | All-cause mortality | 2/3 | 0.93 (0.35, 2.47) | 0.72 (0.45, 1.17) | N. Amer. (0.641) |
| Rothberg et al., 2005 | Warfarin plus aspirin after MI or ACS | Aspirin only | All-cause mortality | 3/2 | 0.93 (0.71, 1.22) | 1.63 (0.33, 8.02) | Europe (0.498) |
| Keeley et al., 2006 | Facilitated PCI for ST-segment elevation MI | Primary PCI | Short-term death (all cause) | 8/2 | 1.07 (0.57, 1.99) | 0.64 (0.08, 5.10) | N. Amer. (0.644) |
| Bravata et al., 2007 | PCI b | CABG surgery b | All-cause mortality | 7/3 | 0.61 (0.33, 1.13) | 0.74 (0.41, 1.34) | Not applic. (0.652) |
| Dentali et al., 2007 | Combined aspirin-oral anticoagulant therapy | Oral anticoagulant therapy alone | All-cause mortality | 4/2 | 1.13 (0.88, 1.45) | 0.61 (0.15, 2.41) | N. Amer. (0.382) |
| Ezekowitz et al., 2007 | Implantable cardioverter defibrillators for LVSD | Usual care | All-cause mortality | 2/3 | 0.82 (0.62, 1.08) | 0.75 (0.63, 0.89) | N. Amer. (0.584) |
| Lafuente-Lafuente et al., 2007 | Antiarrhythmics after cardioversion of AF | Placebo, drugs for rate control or no treatment | All-cause mortality | 6/4 | 1.00 (0.78, 1.29) | 1.43 (0.76, 2.70) | Europe (0.301) |
| Wijeysundera et al., 2007 | Rescue PCI after failed fibrinolytic therapy for MI | Conservative therapy | All-cause mortality | 2/2 | 0.68 (0.38, 1.22) | 0.80 (0.04, 14.5) | Europe (0.918) |
| De Luca et al., 2008 | Early IIb/IIIa inhibitors for primary PCI | Late IIb/IIIa inhibitors | All-cause mortality | 6/2 | 0.64 (0.33, 1.23) | 1.19 (0.38, 3.67) | Europe (0.357) |
| Schomig et al., 2008 | PCI-based invasive strategy for stable CAD | Usual medical treatment | Cardiovascular mortality | 5/2 | 0.51 (0.26, 1.00) | 0.92 (0.53, 1.59) | Europe (0.187) |
| Whitlock et al., 2008 | Prophylactic steroids for cardiopulmonary bypass | Placebo or standard care | In-hospital mortality (all cause) | 4/6 | 0.67 (0.22, 2.09) | 0.65 (0.23, 1.86) | N. Amer. (0.967) |
| Wijeysundera et al., 2008 | Early invasive strategy after fibrinolytic therapy for MI | Ischaemia-guided management | All-cause mortality | 3/2 | 0.58 (0.37, 0.94) | 0.60 (0.16, 2.22) | Europe (0.976) |
| Bosch and Marrugat, 2001 | Glycoprotein IIb/IIIa inhibitors for PCI | Placebo | Death within 30 days (all cause) | 7/8 | 0.60 (0.32, 1.11) | 0.72 (0.50, 1.06) | Europe (0.598) |
| Andraws et al., 2005 | Antichlamydial antibiotic therapy for CAD | Placebo | All-cause mortality | 4/2 | 1.11 (0.70, 1.76) | 1.08 (0.86, 1.35) | N. Amer. (0.898) |
| Clark et al., 2005 | Secondary prevention programmes for CAD | Usual care | All-cause mortality | 17/10 | 0.81 (0.70, 0.94) | 1.08 (0.83, 1.40) | Europe (0.066) |
| Gabriel et al., 2005 | Hormone replacement therapy for preventing CVD | Placebo | Cardiovascular mortality | 2/4 | 0.70 (0.41, 1.20) | 1.11 (0.88, 1.39) | Europe (0.130) |
| Holland et al., 2005 | Multidisciplinary interventions for HF | Usual care | All-cause mortality | 13/10 | 0.80 (0.61, 1.06) | 0.82 (0.66, 1.01) | Europe (0.922) |
| Mehta et al., 2005 | Routine invasive strategies for ACS b | Selective invasive strategies b | All-cause mortality | 2/3 | 0.55 (0.17, 1.83) | 1.18 (0.93, 1.51) | Not applic. (0.221) |
| Roccaforte et al., 2005 | Disease management programmes for HF | Usual care | All-cause mortality | 12/13 | 0.85 (0.65, 1.11) | 0.84 (0.67, 1.05) | N. Amer. (0.967) |
| Taylor et al., 2005 | Case management intervention for HF | Usual care | All-cause mortality | 4/5 | 1.12 (0.85, 1.47) | 0.75 (0.53, 1.05) | N. Amer. (0.074) |
| Abdulla et al., 2006 | CRT for LVSD | No CRT | All-cause mortality | 5/3 | 0.69 (0.54, 0.87) | 0.82 (0.66, 1.03) | Europe (0.287) |
| Collet et al., 2006 | Systematic early PCI after fibrinolytic therapy | Ischaemia-guided PCI | All-cause mortality | 2/2 | 0.52 (0.27, 1.00) | 0.92 (0.24, 3.61) | Europe (0.460) |
| Dong et al., 2006 | Thrombolytic therapy for PE | Heparin or placebo plus heparin | All-cause mortality | 2/3 | 1.57 (0.45, 5.45) | 0.91 (0.16, 5.32) | N. Amer. (0.620) |
| Gohler et al., 2006 | Disease management programmes for CHF | Standard care | All-cause mortality | 12/15 | 0.79 (0.59, 1.06) | 0.86 (0.71, 1.04) | Europe (0.653) |
| Roiron et al., 2006 | Drug-eluting stents | Bare-metal stents | All-cause mortality | 3/5 | 0.47 (0.06, 3.64) | 0.98 (0.52, 1.84) | Europe (0.506) |
| Sedrakyan et al., 2006 | Off-pump CABG surgery | On-pump CABG surgery | All-cause mortality | 9/3 | 0.92 (0.44, 1.93) | 1.21 (0.41, 3.58) | Europe (0.677) |
| Clark et al., 2007 | Secondary prevention programmes for CHD | Usual care | All-cause mortality | 9/7 | 0.81 (0.70, 0.94) | 0.99 (0.71, 1.38) | Europe (0.278) |
| Clark et al., 2007 | Remote monitoring programmes for chronic HF | Usual care | All-cause mortality | 3/9 | 0.63 (0.43, 0.92) | 0.78 (0.60, 1.00) | Europe (0.363) |
| Kastrati et al., 2007 | Sirolimus-eluting stents | Bare-metal stents | All-cause mortality | 5/2 | 1.00 (0.69, 1.44) | 0.95 (0.65, 1.39) | N. Amer. (0.837) |
| Labinaz et al., 2007 | Glycoprotein IIb/IIIa inhibitors for PCI | Placebo or standard care | Death within 30 days (all cause) | 4/7 | 0.44 (0.21, 0.89) | 0.72 (0.50, 1.04) | Europe (0.221) |
| McAlister et al., 2007 | CRT for LVSD | No CRT | All-cause mortality | 5/4 | 0.72 (0.57, 0.90) | 0.83 (0.67, 1.03) | Europe (0.349) |
| Ramakrishnan, 2007 | Thrombolytic therapy for submassive PE | Placebo | Short-term death (all cause) | 2/2 | 1.57 (0.45, 5.45) | 0.70 (0.08, 6.44) | N. Amer. (0.534) |
| Roberts et al., 2007 | Statin monotherapy for CVD in older adults | Placebo | All-cause mortality | 5/6 | 0.89 (0.84, 0.94) | 0.97 (0.43, 2.18) | Europe (0.828) |
| Ward et al., 2007 | Statins for CHD | Placebo | CHD mortality | 12/5 | 0.81 (0.74, 0.89) | 0.82 (0.64, 1.05) | Europe (0.900) |
| Moller et al., 2008 | Off-pump CABG surgery | On-pump CABG surgery | All-cause mortality | 14/4 | 0.96 (0.60, 1.53) | 1.55 (0.72, 3.36) | Europe (0.290) |
| Qayyum et al., 2008 | Routine invasive strategy for ACS b | Selective invasive strategy b | All-cause mortality | 4/3 | 0.74 (0.47, 1.16) | 1.18 (0.93, 1.51) | Not applic. (0.0728) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
