Multivessel coronary artery disease: Difference between revisions
| (11 intermediate revisions by the same user not shown) | |||
| Line 17: | Line 17: | ||
===CABG Versus PCI=== | ===CABG Versus PCI=== | ||
Many PCI vs. CABG trials did not have widespread use of | Many [[PCI]] vs. [[CABG]] trials did not have widespread use of [[stent]]s (either [[BMS|bare metal]] or [[DES|drug-eluting]]) or [[GP IIb/IIIa inhibitors]], and <10% of patients who were screened for trials of [[PTCA]] vs. [[CABG]] were actually randomized, and therefore represent a highly select population. With multivessel [[stent]]ing, the target [[lesion]] [[revascularization]] (TLR) rates become cumulative. [[Diabetics]] with both [[retinopathy]] and [[nephropathy]] appear to have very high major adverse [[cardiac]] events (MACE) rates with [[PCI]] (up to 50%). | ||
==Revascularization by Percutaneous Coronary Intervention (PCI)== | ==Revascularization by Percutaneous Coronary Intervention (PCI)== | ||
===PCI Versus Medical Therapy=== | ===PCI Versus Medical Therapy=== | ||
The subgroup analysis of patients with stable, multi-vessel CAD in the COURAGE trial suggested no difference in death and MI rates between PCI- and medically-treated groups. | The subgroup analysis of patients with [[stable CAD|stable]], multi-vessel [[CAD]] in the [[COURAGE]] trial suggested no difference in death and [[MI]] rates between [[PCI]]- and medically-treated groups. | ||
===PCI Versus CABG=== | ===PCI Versus CABG=== | ||
One must be confident in their ability to achieve complete revascularization with PCI when offering it as an alternative to CABG. | One must be confident in their ability to achieve complete [[revascularization]] with [[PCI]] when offering it as an alternative to [[CABG]]. | ||
====Mortality and MI==== | ====Mortality and MI==== | ||
Mortality and nonfatal MI rates are not significantly different between the two strategies. | [[Mortality]] and nonfatal [[MI]] rates are not significantly different between the two strategies. The [[BARI]] trial showed similar 5-year survival among over 1800 patients randomized to an initial strategy of [[PTCA]] or [[CABG]] for multi-vessel [[CAD]], despite the higher rates of “complete [[revascularization]]” in the [[CABG]] arm. This trial preceded the use of [[drug eluting stents]]. | ||
====Recurrent Revascularization==== | ====Recurrent Revascularization==== | ||
There are higher rate of recurrent angina and repeat revascularization after PCI (most trials in low-risk patients with 2-vessel disease and normal [[ | There are higher rate of recurrent [[angina]] and repeat [[revascularization]] after [[PCI]] (most trials in low-risk patients with 2-vessel disease and normal [[left ventricular function]]); this may change in future with the evolution of [[DES|drug eluting stents]]. In [[BARI]], [[CABG]] was associated with a lower rate of repeat [[revascularization]]s. | ||
====Symptom Relief==== | ====Symptom Relief==== | ||
CABG has been associated with a greater relief of | [[CABG]] has been associated with a greater relief of [[angina]]l [[symptom]]s. | ||
====Costs==== | ====Costs==== | ||
*In comparison to CABG, PCI is less invasive, has a shorter hospital stay and [[convalescence]], and has a less expensive initial hospital stay. | *In comparison to [[CABG]], [[PCI]] is less invasive, has a shorter hospital stay and [[convalescence]], and has a less expensive initial hospital stay. However, the cost advantage may be lost over the long-term due to the potential need for repeat [[revascularization]]. | ||
====Trial Results==== | ====Trial Results==== | ||
The ARTS and SYNTAX trials showed higher primary event rates in patients randomized to PCI versus CABG, driven by a higher need for revascularization. | The ARTS and [[SYNTAX]] trials showed higher primary event rates in patients randomized to [[PCI]] versus [[CABG]], driven by a higher need for [[revascularization]]. Rates of hard events, such as death and [[MI]], were similar between the treatment groups. | ||
Several trials (ARTS I, MASS II, ERACI-II, AWESOME) involving [[bare metal stents]] compared to CABG have shown similar survival rates but higher revascularization rates among patients with [[bare | Several trials (ARTS I, MASS II, ERACI-II, AWESOME) involving [[bare metal stents]] compared to [[CABG]] have shown similar survival rates but higher [[revascularization]] rates among patients with [[bare metal stents]] at 5 years. The [[SYNTAX]] trial, a [[randomized control trial|randomized trial]] of multi-vessel or [[left main]] [[CAD]] to [[CABG]] or [[paclitaxel]]-eluting [[stent]]s, showed higher primary adverse event rates in the [[PCI]] group (17.8% vs. 12.4% for [[CABG]]; p=0.002), largely due to an increased rate of repeat [[revascularization]] (13.5% vs. 5.9%, p<0.001). | ||
==Selecting a Therapeutic Strategy or Strategies== | ==Selecting a Therapeutic Strategy or Strategies== | ||
===Optimal Medical Therapy=== | ===Optimal Medical Therapy=== | ||
* Risk factor modification should be undertaken in all patients (smoking cessation, treatment of [[hypertension]], correction of [[dyslipidemia]]). | * Risk factor modification should be undertaken in all patients (smoking cessation, treatment of [[hypertension]], correction of [[dyslipidemia]]). | ||
* Optimal medical therapy should be advised for all patients. | * Optimal medical [[therapy]] should be advised for all patients. Medical [[therapy]] may be an acceptable choice as the sole treatment when [[left ventricle|LV]] [[systolic]] function is normal or mildly depressed, and when the lifestyle is acceptable with medical therapy. | ||
* Revascularization is chosen when unacceptable | * [[Revascularization]] is chosen when unacceptable [[symptom]]s persist despite optimal medical therapy and when [[lesion]]s and risk factors are present for which [[revascularization]] improves [[morbidity]] and [[mortality]] compared with medical therapy (e.g. a [[LMCA|left main]] [[lesion]]). Specifically, [[revasulcarization]] is appropriate for patients with: | ||
**[[Refractory]] | **[[Refractory]] [[symptom]]s | ||
**More extensive or proximal disease | **More extensive or [[proximal]] disease | ||
**Left ventricular dysfunction | **[[Left ventricular dysfunction]] | ||
===Scenarios Favoring CABG Over Medical Therapy to Prolong Survival=== | ===Scenarios Favoring CABG Over Medical Therapy to Prolong Survival=== | ||
* [[Left main]] stenosis >50%, | * [[LMCA|Left main]] [[stenosis]] >50%, | ||
* Multivessel disease and LV systolic dysfunction, | * Multivessel disease and [[left ventricle|LV]] [[systolic dysfunction]], | ||
* 3-vessel disease with proximal LAD [[stenosis]] regardless of [[LV function]], | * 3-vessel disease with [[proximal]] [[LAD]] [[stenosis]] regardless of [[LV function]], | ||
* 2-vessel disease and LV systolic dysfunction (especially with proximal disease and severe angina) | * 2-vessel disease and [[left ventricle|LV]] [[systolic dysfunction]] (especially with [[proximal]] disease and severe [[angina]]) | ||
===Scenarios Favoring PCI over CABG=== | ===Scenarios Favoring PCI over CABG=== | ||
* [[CAD]] anatomy and complexity is suitable for PCI (single or two vessel disease for instance) | * [[CAD]] [[anatomy]] and complexity is suitable for [[PCI]] (single or two [[vessel]] disease for instance) | ||
* Younger patients who will likely require CABG in future in order to delay the inevitable surgery | * Younger patients who will likely require [[CABG]] in future in order to delay the inevitable [[surgery]] | ||
* Limited [[life-expectancy]] | * Limited [[life-expectancy]] | ||
* High operative risk (including: [[cerebrovascular disease]] and severe [[COPD|chronic obstructive pulmonary disease | * High operative risk (including: [[cerebrovascular disease]] and severe [[COPD|chronic obstructive pulmonary disease]]), or other illnesses limiting survival | ||
* Poor graft conduits (no [[IMA]] available or poor vein quality) | * Poor [[graft]] conduits (no [[IMA]] available or poor [[vein]] quality) | ||
* Patient prefers to avoid surgery | * Patient prefers to avoid [[surgery]] | ||
* In | * In non [[diabetic]] patients with 3-vessel disease and poor [[left ventricle|LV]] [[systolic]] function, consider [[PCI]] in select patients with low-risk [[lesion]]s if complete [[revascularization]] can be achieved. The risks and benefits of lifelong [[plavix]] and risks of [[stent thrombosis]] must be discussed extensively with the patient. [[Stent]]ing is not an ideal choice for the patient who is not compliant with medication such as [[thienopyridine]]s. | ||
===Scenarios Favoring CABG over PCI=== | ===Scenarios Favoring CABG over PCI=== | ||
* In the presence of concurrent [[valvular disease]] requiring surgical repair, | * In the presence of concurrent [[valvular disease]] requiring [[surgery|surgical]] repair, | ||
* Complete functional revascularization unlikely to be achieved with PCI, | * Complete functional [[revascularization]] unlikely to be achieved with [[PCI]], | ||
* | * [[Lesion]]s not suitable for [[stent]]ing (low likelihood of success, high risk of [[complication]]s, high risk of [[restenosis]]), | ||
* Patient prefers to limit number of revascularization procedures | * Patient prefers to limit number of [[revascularization]] procedures | ||
* Patients with [[diabetes mellitus]] and multi-vessel disease may benefit more from CABG than PTCA. In general, CABG is recommended because the BARI trial showed improved survival after CABG compared with multivessel PCI if at least one IMA conduit was used, and if there were 4 or more | * Patients with [[diabetes mellitus]] and multi-vessel disease may benefit more from [[CABG]] than [[PTCA]]. In general, [[CABG]] is recommended because the [[BARI]] trial showed improved survival after [[CABG]] compared with multivessel [[PCI]] if at least one [[IMA]] conduit was used, and if there were 4 or more [[lesion]]s, especially in patients with [[left ventricle|LV]] [[systolic dysfunction]]. However, the [[BARI]] trial was done in the pre- [[stent]] era and before the widespread use of [[GP IIb/IIIa inhibitors]]. It should be noted that there was no difference in survival among [[diabetic]] patients treated with [[CABG]] vs. [[PCI]] in 2 nonrandomized trials, including 5-year survival in [[BARI]] registry. Randomized data of [[diabetic]] patients using contemporary [[stent]]s and [[CABG]] techniques are lacking. The [[NHLBI]] sponsored FREEDOM trial is currently enrolling patients with [[diabetes]] and multi-vessel [[CAD]] for a comparison of [[PCI]] and [[CABG]] outcomes. | ||
A collaborative analysis of data from 10 randomized controlled trials (N= 7812) was pooled to compare effectiveness of [[CABG]] with [[PCI]] in view of long term effects on mortality in various clinical subgroups. PCI was done with balloon angioplasty in six trials and with bare | A collaborative analysis of data from 10 [[randomized controlled trial|randomized controlled trials]] (N= 7812) was pooled to compare effectiveness of [[CABG]] with [[PCI]] in view of long term effects on [[mortality]] in various clinical subgroups. [[PCI]] was done with [[balloon angioplasty]] in six trials and with [[bare metal stent]]s in four trials. The results showed that the long term [[mortality]] is similar with [[PCI]] and [[CABG]] in most patient subgroups who had multivessel disease. Based on this, the choice of the procedure should be made depending on patient's preference for other outcomes. [[CABG]] proved a better option in [[diabetic]]s and elderly over 65 years.<ref name="pmid19303634">{{cite journal| author=Hlatky MA, Boothroyd DB, Bravata DM, Boersma E, Booth J, Brooks MM et al.| title=Coronary artery bypass surgery compared with percutaneous coronary interventions for multivessel disease: a collaborative analysis of individual patient data from ten randomised trials. | journal=Lancet | year= 2009 | volume= 373 | issue= 9670 | pages= 1190-7 | pmid=19303634 | doi=10.1016/S0140-6736(09)60552-3 | pmc= | url=http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&tool=sumsearch.org/cite&retmode=ref&cmd=prlinks&id=19303634 }} [http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&tool=sumsearch.org/cite&retmode=ref&cmd=prlinks&id=19620152 Review in: Ann Intern Med. 2009 Jul 21;151(2):JC1-8, JC1-9] </ref> | ||
===Scenarios Favoring a Hybrid of CABG and PCI=== | ===Scenarios Favoring a Hybrid of CABG and PCI=== | ||
* Adjunctive PCI may be performed before after CABG in | * [[Adjunctive PCI]] may be performed before after [[CABG]] in [[lesion]]s that are not amenable to [[coronary artery bypass grafting]]. This may include [[lesion]]s in very [[distal]] [[artery|arteries]], or [[lesion]]s in very small [[vessel]]s | ||
* Due to higher CABG mortality in patients with [[Unstable angina / non ST elevation myocardial infarction|UA/NSTEMI]], a strategy of PCI to the “culprit | * Due to higher [[CABG]] [[mortality]] in patients with [[Unstable angina / non ST elevation myocardial infarction|UA/NSTEMI]], a strategy of [[PCI]] to the “culprit [[artery]]” followed by [[elective]] [[revascularization]] (as needed) of the residual disease may be employed. Identification of the culprit [[artery]] requires localizing [[ECG]], [[echocardiographic]], or [[angiographic]] features ([[coronary artery|coronary]] [[thrombus]], [[ulcer|ulcerative]] [[plaque]], slow flow, a high grade [[stenosis]], or pressure wire technique). | ||
==Technical Considerations in the Performance of Multivessel PCI== | ==Technical Considerations in the Performance of Multivessel PCI== | ||
* One may need to stage the procedure because of contrast load and [[radiation]] dose, as well as procedure time. | * One may need to stage the procedure because of [[contrast]] load and [[radiation]] dose, as well as procedure time. | ||
* Starting with the most challenging lesion in patients for whom CABG is an option, may be advisable to evaluate feasibility of complete revascularization. | * Starting with the most challenging [[lesion]] in patients for whom [[CABG]] is an option, may be advisable to evaluate feasibility of complete [[revascularization]]. | ||
* Assessment of patient’s ability to comply with lifetime dual [[antiplatelet]] therapy is also crucial especially with bifurcation stenting, long lesions and small | * Assessment of patient’s ability to comply with lifetime dual [[antiplatelet]] therapy is also crucial especially with [[bifurcation stenting]], long lesions and small [[vessel]]s, which are common in patients with multivessel disease where risk of [[stent thrombosis]] is highest. | ||
==References== | ==References== | ||
Latest revision as of 19:21, 15 January 2013
Editor-In-Chief: C. Michael Gibson, M.S., M.D. [1]; Associate Editors-In-Chief: Joanna J. Wykrzykowska, MD [2]; Robert Sperling, MD; Brian Bigelow, MD; Roger J. Laham, MD [3]
See also the chapter on Chronic stable angina revascularization
Overview
Multi-vessel coronary artery disease (CAD) is a disease stage in which at least two or three of the epicardial coronary arteries is involved with atherosclerosis of significant severity. Multivessel disease is often associated with a higher burden of comorbidities, left ventricular dysfunction, and cardiovascular risk. In general, PCI is preferred in patients with single or low risk two vessel disease. In contrast, coronary artery bypass graft surgery is recommended in patients with complex two vessel disease, three vessel disease and in some cases of isolated left main disease. Diabetics with left anterior descending coronary artery disease may benefit more from coronary artery bypass surgery. The goal in the treatment of multivessel disease is to reduce angina and heart failure symptoms and to reduce a patient's subsequent risk of adverse cardiovascular events.
Medical Therapy
Medical therapy consists of antianginal and antiplatelet medications as well as high dose statins. All patients, whether they undergo PCI, CABG or not, should receive optimal medical therapies to reduce cardiovascular event-risk and angina. Patients with lower-risk, stable CAD may be effectively treated by medical therapy. Medical therapy has no procedural risk or prolonged convalescence, but the COURAGE Trial[1] showed an increased risk of angina and a decreased quality of life in patients treated with medical therapy. Moreover, 30% of the patients treated with medical therapy eventually needed revascularization. Approximately 2/3 of the study population in COURAGE had multi-vessel CAD, and randomization to an initial strategy of medical therapy resulted in similar rates of death and myocardial infarction (MI) to an initial strategy of PCI.
Revascularization Utilizing Coronary Artery Bypass Grafting (CABG)
When compared to medical therapy and percutaneous coronary intervention, CABG is associated with a lower incidence of recurrent angina and a lower need for repeat revascularization. It reduces late cardiac mortality in diabetic patients who received at least 1 internal mammary (IMA) graft. The rate of revascularization may be comparable in the era of drug eluting stents, but the definitive results of ongoing trials are pending.
CABG Versus Medical Therapy
Recommendations are limited by the quality of data. For instance, older trials of CABG vs. medical therapy had little use of an IMA conduit (which has greater durability) and limited use of ASA, ACE inhibitors, and statins. Several randomized trials of CABG versus medical therapy support the concept of greater absolute benefit associated with CABG with respect to long-term survival in patients with more extensive or proximal CAD, or in patients with impaired left ventricular function. These older data are limited by low usage of internal mammary artery (IMA) grafting, antiplatelet agents, and a high cross-over of the medical treatment arm to CABG. CABG offers survival benefit in patients with left main stenosis, multivessel disease and LV systolic dysfunction, 3-vessel disease with proximal LAD stenosis regardless of left ventricular function, and 2-vessel disease and LV systolic dysfunction, especially with proximal disease and severe angina.
CABG Versus PCI
Many PCI vs. CABG trials did not have widespread use of stents (either bare metal or drug-eluting) or GP IIb/IIIa inhibitors, and <10% of patients who were screened for trials of PTCA vs. CABG were actually randomized, and therefore represent a highly select population. With multivessel stenting, the target lesion revascularization (TLR) rates become cumulative. Diabetics with both retinopathy and nephropathy appear to have very high major adverse cardiac events (MACE) rates with PCI (up to 50%).
Revascularization by Percutaneous Coronary Intervention (PCI)
PCI Versus Medical Therapy
The subgroup analysis of patients with stable, multi-vessel CAD in the COURAGE trial suggested no difference in death and MI rates between PCI- and medically-treated groups.
PCI Versus CABG
One must be confident in their ability to achieve complete revascularization with PCI when offering it as an alternative to CABG.
Mortality and MI
Mortality and nonfatal MI rates are not significantly different between the two strategies. The BARI trial showed similar 5-year survival among over 1800 patients randomized to an initial strategy of PTCA or CABG for multi-vessel CAD, despite the higher rates of “complete revascularization” in the CABG arm. This trial preceded the use of drug eluting stents.
Recurrent Revascularization
There are higher rate of recurrent angina and repeat revascularization after PCI (most trials in low-risk patients with 2-vessel disease and normal left ventricular function); this may change in future with the evolution of drug eluting stents. In BARI, CABG was associated with a lower rate of repeat revascularizations.
Symptom Relief
CABG has been associated with a greater relief of anginal symptoms.
Costs
- In comparison to CABG, PCI is less invasive, has a shorter hospital stay and convalescence, and has a less expensive initial hospital stay. However, the cost advantage may be lost over the long-term due to the potential need for repeat revascularization.
Trial Results
The ARTS and SYNTAX trials showed higher primary event rates in patients randomized to PCI versus CABG, driven by a higher need for revascularization. Rates of hard events, such as death and MI, were similar between the treatment groups.
Several trials (ARTS I, MASS II, ERACI-II, AWESOME) involving bare metal stents compared to CABG have shown similar survival rates but higher revascularization rates among patients with bare metal stents at 5 years. The SYNTAX trial, a randomized trial of multi-vessel or left main CAD to CABG or paclitaxel-eluting stents, showed higher primary adverse event rates in the PCI group (17.8% vs. 12.4% for CABG; p=0.002), largely due to an increased rate of repeat revascularization (13.5% vs. 5.9%, p<0.001).
Selecting a Therapeutic Strategy or Strategies
Optimal Medical Therapy
- Risk factor modification should be undertaken in all patients (smoking cessation, treatment of hypertension, correction of dyslipidemia).
- Optimal medical therapy should be advised for all patients. Medical therapy may be an acceptable choice as the sole treatment when LV systolic function is normal or mildly depressed, and when the lifestyle is acceptable with medical therapy.
- Revascularization is chosen when unacceptable symptoms persist despite optimal medical therapy and when lesions and risk factors are present for which revascularization improves morbidity and mortality compared with medical therapy (e.g. a left main lesion). Specifically, revasulcarization is appropriate for patients with:
- Refractory symptoms
- More extensive or proximal disease
- Left ventricular dysfunction
Scenarios Favoring CABG Over Medical Therapy to Prolong Survival
- Left main stenosis >50%,
- Multivessel disease and LV systolic dysfunction,
- 3-vessel disease with proximal LAD stenosis regardless of LV function,
- 2-vessel disease and LV systolic dysfunction (especially with proximal disease and severe angina)
Scenarios Favoring PCI over CABG
- CAD anatomy and complexity is suitable for PCI (single or two vessel disease for instance)
- Younger patients who will likely require CABG in future in order to delay the inevitable surgery
- Limited life-expectancy
- High operative risk (including: cerebrovascular disease and severe chronic obstructive pulmonary disease), or other illnesses limiting survival
- Poor graft conduits (no IMA available or poor vein quality)
- Patient prefers to avoid surgery
- In non diabetic patients with 3-vessel disease and poor LV systolic function, consider PCI in select patients with low-risk lesions if complete revascularization can be achieved. The risks and benefits of lifelong plavix and risks of stent thrombosis must be discussed extensively with the patient. Stenting is not an ideal choice for the patient who is not compliant with medication such as thienopyridines.
Scenarios Favoring CABG over PCI
- In the presence of concurrent valvular disease requiring surgical repair,
- Complete functional revascularization unlikely to be achieved with PCI,
- Lesions not suitable for stenting (low likelihood of success, high risk of complications, high risk of restenosis),
- Patient prefers to limit number of revascularization procedures
- Patients with diabetes mellitus and multi-vessel disease may benefit more from CABG than PTCA. In general, CABG is recommended because the BARI trial showed improved survival after CABG compared with multivessel PCI if at least one IMA conduit was used, and if there were 4 or more lesions, especially in patients with LV systolic dysfunction. However, the BARI trial was done in the pre- stent era and before the widespread use of GP IIb/IIIa inhibitors. It should be noted that there was no difference in survival among diabetic patients treated with CABG vs. PCI in 2 nonrandomized trials, including 5-year survival in BARI registry. Randomized data of diabetic patients using contemporary stents and CABG techniques are lacking. The NHLBI sponsored FREEDOM trial is currently enrolling patients with diabetes and multi-vessel CAD for a comparison of PCI and CABG outcomes.
A collaborative analysis of data from 10 randomized controlled trials (N= 7812) was pooled to compare effectiveness of CABG with PCI in view of long term effects on mortality in various clinical subgroups. PCI was done with balloon angioplasty in six trials and with bare metal stents in four trials. The results showed that the long term mortality is similar with PCI and CABG in most patient subgroups who had multivessel disease. Based on this, the choice of the procedure should be made depending on patient's preference for other outcomes. CABG proved a better option in diabetics and elderly over 65 years.[2]
Scenarios Favoring a Hybrid of CABG and PCI
- Adjunctive PCI may be performed before after CABG in lesions that are not amenable to coronary artery bypass grafting. This may include lesions in very distal arteries, or lesions in very small vessels
- Due to higher CABG mortality in patients with UA/NSTEMI, a strategy of PCI to the “culprit artery” followed by elective revascularization (as needed) of the residual disease may be employed. Identification of the culprit artery requires localizing ECG, echocardiographic, or angiographic features (coronary thrombus, ulcerative plaque, slow flow, a high grade stenosis, or pressure wire technique).
Technical Considerations in the Performance of Multivessel PCI
- One may need to stage the procedure because of contrast load and radiation dose, as well as procedure time.
- Starting with the most challenging lesion in patients for whom CABG is an option, may be advisable to evaluate feasibility of complete revascularization.
- Assessment of patient’s ability to comply with lifetime dual antiplatelet therapy is also crucial especially with bifurcation stenting, long lesions and small vessels, which are common in patients with multivessel disease where risk of stent thrombosis is highest.
References
- ↑ Boden WE, O'Rourke RA, Teo KK; et al. (2007). "Optimal medical therapy with or without PCI for stable coronary disease". N. Engl. J. Med. 356 (15): 1503–16. doi:10.1056/NEJMoa070829. PMID 17387127. Unknown parameter
|month=ignored (help) - ↑ Hlatky MA, Boothroyd DB, Bravata DM, Boersma E, Booth J, Brooks MM; et al. (2009). "Coronary artery bypass surgery compared with percutaneous coronary interventions for multivessel disease: a collaborative analysis of individual patient data from ten randomised trials". Lancet. 373 (9670): 1190–7. doi:10.1016/S0140-6736(09)60552-3. PMID 19303634. Review in: Ann Intern Med. 2009 Jul 21;151(2):JC1-8, JC1-9