Methods

Articles were identified by a literature search of the MEDLINE database (1966 to December 1, 2010), English-language journals, a manual search of the authors’ reference files, and reference lists of original articles, reviews, and meta-analyses using methods previously described to identify randomized controlled trials. Two abstractors independently extracted information using a standardized protocol and reporting form. Disagreements were resolved by consensus. For inclusion in meta-analyses, a study must have met these criteria: (1) studies were designed to evaluate the effect of diet, statins, niacin, fibrates, bile acid sequestrants, or surgery compared to an active or placebo control; (2) random blinded allocation of study participants to a treatment or a control group; and (3) apoB, total cholesterol, LDL cholesterol, and HDL cholesterol were measured ≥1 time after baseline. If measured non-HDL cholesterol was not available, non-HDL cholesterol was calculated from total cholesterol minus HDL cholesterol. Measured and calculated non-HDL cholesterol levels were within 1 mg/dl for every study in which non-HDL cholesterol was measured. These analyses used lipid values reported in publications reporting apoB levels; they may differ from values reported in the main results publication. Interval for lipid measurement was not fixed, and in some cases values could represent the average during the trial. (4) Trials of ≥2-year duration were included to provide a stable estimate of relative risk decrease. (5) Nonfatal myocardial infarctions and coronary heart disease deaths were summed for the “hard” coronary heart disease end point; nonfatal strokes and fatal strokes were used for the stroke end point. For statin trials we used updated coronary heart disease, stroke, and major vascular (first major coronary heart disease, stroke, cardiovascular disease death, and coronary revascularization) event rates from the 2010 individual meta-analysis reported by the Cholesterol Treatment Trialists (with the exception of the Stroke Prevention by Aggressive Reduction of Cholesterol Levels [SPARCL] trial). In nonstatin trials we used event rates from the main results publication. Analyses of total cardiovascular events included major cardiovascular events (including nonfatal myocardial infarction, stroke, cardiovascular disease death, arterial revascularizations, and unstable angina for previous niacin trials; no stroke outcomes were reported for 2 of 4 fibrate trials and 4 of 4 niacin trials; Table 1 ); coronary heart disease and stroke events were added if a composite cardiovascular disease measurement was not available. (6) Most of the study population did not have serious noncardiovascular co-morbidities (e.g., heart or renal failure, organ transplantation) that might bias estimates of cardiovascular events owing to competing causes of mortality.

Decreases in lipid and apoB levels were calculated as the difference (milligrams per deciliter) between the 2 treatment groups at the same study interval. Because relative risks and hazard ratios were not consistently reported across studies, the study-specific crude relative risk and associated SEs were estimated from the published total number of subjects and incident cases in the treatment and control groups for a primary analysis of the composite cardiovascular disease end point and secondary analyses of coronary heart disease and stroke separately. A random-effects model was used for the meta-analyses. Natural log-transformed relative risk was modeled as a linear function of the study-specific mean difference in apoB, LDL cholesterol, and non-HDL cholesterol between the 2 treatment groups and mean length of follow-up. For comparative models, treatment-by-apoB interaction was added to allow the effect of apoB decrease to vary between the 2 types of treatments. Gaussian errors were specified as a combination of within- and between-study variations. Study-specific SEs for estimated relative risks were used to account for within-study variation. Between-study variation was estimated in the analysis. Bayesian methods were used to fit random-effects meta-analysis models. A previous odds of unity was assumed to indicate no previous preference for the null or alternative hypothesis. Vague previous specifications were used for all regression parameters.

Models were fit with the WinBUGS and the methods of Chib were used to compute the Bayes factor (BF) K to compare the association between magnitude of apoB decrease and coronary heart disease, stroke, or cardiovascular disease risk decrease. A K value of around 1 provides evidence for either (neither) model. A K value >1 indicates that model 1 is more strongly supported by the data than model 2, although the interpretation depends on the scale of K. For example, a BF 1 to 3 provides slight support for model 1, and the strength of evidence is “barely worth mentioning.” In contrast, a BF 3 to 15 provides “substantial” support for model 1. Similarly, a K value <1 indicates that model 2 is more strongly supported by the data than model 1. In addition, deviance information criteria (DICs) were used to help with model selection. Roughly, differences in DICs of >10 might definitely rule out the model with the higher DIC; however, if the difference in DIC is small, e.g., <5, then DIC alone may not provide strong evidence for model selection. Cochrane Q test was used to test for heterogeneity using the Meta R 2.10.

Results

Analysis included 25 trials (n = 131,134). For statin trials, 12 met inclusion criteria (n = 101,957); all used statins as monotherapy and 7 were placebo controlled ( Table 1 ) (LaRosa JC, Personal communication, 2010). For fibrate trials, 4 met inclusion criteria (n = 12,908) and all used a fibrate as monotherapy compared to placebo: 1 used gemfibrozil (n = 2,531), 2 used fenofibrate (n = 10,213), and 1 used bezafibrate (n = 164). One large trial and 4 small niacin trials met all inclusion criteria (n = 3,819). Two trials compared simvastatin–ezetimibe to placebo (n = 11,143), 1 trial compared intensive LDL cholesterol and blood pressure lowering to standard therapy (n = 499), and 1 trial evaluated ileal bypass surgery (n = 838). Mean follow-up for the trials was 4.3 years.

Statin trials reported 15- to 68-mg/dl decreases in LDL cholesterol, 25- to 72-mg/dl decreases in non-HDL cholesterol, and 10- to 39-mg/dl decreases in apoB between the 2 treatment groups ( Table 2 ); in these trials, 4% to 46% decreases in cardiovascular disease risk, 8% to 52% decreases in coronary heart disease risk, and 13% increases to 48% decreases in relative risk of stroke were observed. Fibrate trials reported 0- to 24-mg/dl decreases in LDL cholesterol, 10- to 29-mg/dl decreases in non-HDL cholesterol, and 5- to 16-mg/dl decreases in apoB between the 2 treatment groups; in these trials, 10% to 64% decreases in cardiovascular disease risk, 11% to 49% decreases in coronary heart disease risk, and 10% to 24% decreases in stroke risk were observed. Niacin and niacin combination trials reported 4- to 56-mg/dl decreases in LDL cholesterol, 11- to 60-mg/dl decreases in non-HDL cholesterol, and 8- to 33-mg/dl decreases in apoB; in these trials an 86% decrease to >200% increase in coronary heart disease risk was observed. Simvastatin–ezetimibe trials reported 33- to 70-mg/dl decreases in LDL cholesterol, 42- to 85-mg/dl decreases in non-HDL cholesterol, and 23- to 57-mg/dl decreases in apoB; in these trials, 14% to 22% decreases in cardiovascular disease risk, 7% to 23% decreases in coronary heart disease risk, and 17% decrease to 12% increase in stroke risk were observed. The remaining 2 trials reported 32- to 67-mg/dl decreases in LDL cholesterol, 36- to 79-mg/dl decreases in non-HDL cholesterol, and 22- to 79-mg/dl decreases in apoB; in these trials, 24% to 32% decreases in cardiovascular disease, 7% to 23% decreases in coronary heart disease, and 67% to 76% decreases in stroke were observed.

Table 2

Absolute difference in on-treatment lipid and apolipoprotein B levels between treatment groups and relative risk decrease in cardiovascular disease ^⁎ , coronary heart disease ^† , and stroke (study references same as Table 1 )

Study Acronym	On-Trial Lipid Measurement (years)	Mean Difference (mg/dl) Between Pl/Less Intense and More Intense Treatment						CVD		CHD		Stroke
		TC↓	HDL Cholesterol↑	TG↓	LDL Cholesterol↓	Non-HDL Cholesterol↓	apoB↓	Crude RR Decrease	95% CI	Crude RR Decrease	95% CI	Crude RR Decrease	95% CI
Statins
4S	5.4	68	4	22	68	72	28	30%	24–36	34%	26–42	26%	−4 to 48
LIPID	1	39	2	12	37	41	32	19%	13–25	26%	15–35	17%	−1 to 32
AF/TexCAPS	1	44	1	20	41	45	28	29%	12–42	38%	16–54	18%	−67 to 59
HPS	3	46	1	35	38	47	29	26%	21–30	30%	23–36	24%	14–32
CARDS	1	56	0	11	48	56	30	35%	15–50	32%	3–53	47%	10–69
IDEAL	0.33	25	−1	21	20	24	19	15%	8–21	12%	0–22	13%	−8 to 30
TNT	5	25	0	27	24	25	23	23%	17–29	22%	9–33	24%	4–40
SPARCL	2.5	61	1	33	56	62	32	18%	6–28	32%	11–49	15%	1–27
PROVE-IT	1	31	−2	31	33	41	24	12%	1–22	16%	−4 to 32	−9%	−103 to 41
JUPITER	1	55	2	20	55	57	39	46%	32–57	52%	29–68	48%	21–65
SEARCH	1	17	1	16	15	25	10	4%	−2 to 10	8%	−1 to 16	9%	−8 to 23
Post-CABG	4	41	1	12	39	42	17	21%	−4 to 40	12%	−36 to 44	−13%	−119 to 42
Fibrates
VA-HIT	1	9	2	55	0	10	5	22%	10–32	20%	6–32	24%	−7 to 45
FIELD	2	20	0	32	6	12	6	10%	0.1–19	11%	−5 to 24	10%	−12 to 27
DAIS	1.7	22	1	55	11	22	10	33%	2–56	24%	−78 to 67	NA	NA
SENDCAP	3	20	5	49	24	29	16	64%	13–85	49%	−172 to 90	NA	NA
Niacin
AIM-HIGH	1	NA	5	34	4	11	8	−2%	−18 to 13	−4%	−35 to 19	−78%	−233 to 5
FATS	1.25	44	15	138	33	59	31	80%	16–95	NA	NA	NA	NA
CLAS II	4	45	15	24	56	60	33	31%	−16 to 59	52%	−57 to 85	NA	NA
HATS	1.5	49	6	70	41	55	31	89%	17–990	86%	−11 to 98	NA	NA
HARP	1.5	56	−6	43	53	43	29	22%	−75 to 66	−220%	−2,867 to 65	NA	NA
Simvastatin + ezetimibe
SHARP	2.5	41	1	24	33	42	23	14%	6–21	7%	−11 to 23	17%	−1 to 31
SEAS	4	NA	NA	NA	70	85	57	22%	5–36	23%	−5 to 44	−12%	−83 to 31
Other
SANDS	3	37	0	−23	32	36	22	24%	−101 to 71	7%	−138 to 52	67%	−212 to 97
POSCH	5	68	4	NA	74	72	33	33%	16–45	23%	17–49	76%	−89 to 55

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Usefulness of Albuminuria as Predictor for Coronary Artery Stenosis, Regardless of Estimated Glomerular Filtration Rate, in Patients With Type 2 Diabetes Mellitus
2. Relation of Statin Nonadherence and Treatment Intensification
3. Long-Term Outcome of Patients With Moderate and Severe Prosthetic Aortic Valve Regurgitation After Transcatheter Aortic Valve Implantation
4. Relation Between Carotid Intima–Media Thickness and Left Ventricular Mass in Type 1 Diabetes Mellitus (from the Epidemiology of Diabetes Interventions and Complications [EDIC] Study)

Stay updated, free articles. Join our Telegram channel

Join