An electronic literature search was performed to identify all relevant studies published in PubMed, EMBASE, and Cochrane databases in the English language from Jan. 1, 2000 to Sep. 13, 2014, using the terms “atherosclerosis” and “cholesterol blood level”. Trials were included using the criteria as: 1) randomized controlled or prospective, blinded end-points trials, and its primary end point was CAP change detected by IVUS; 2) report of LDL-C levels at baseline and follow-up; 3) data on the volume of CAP at baseline and follow-up, and volume of CAP was calculated as vessel volume minus lumen volume; Exclusion criteria were: 1) only CAP area or volume index or percent atheroma volume were detected; 2) the levels of LDL-C at baseline or follow-up were not provided; and 3) target plaques were unstable.

Two investigators independently reviewed all potentially eligible studies and collected data on patient and study characteristics, and any disagreement was resolved by consensus. The primary end point of this study was the volume change of CAP detected by IVUS. Quality assessments of trials were evaluated with Jadad quality scale.

Volume changes of CAP from baseline to follow-up were analyzed using standardized mean differences (SMD).

Volume changes of plaque in every arm were used for pooled analysis. The trials were firstly grouped into group Western and Asian according to the location of the trials. Then, according to the levels and the reducing percentage of LDL-C at follow-up, the arms were grouped to following groups: ≤70, >70 ≤ 100 HP, >70 ≤ 100 MP, >70 ≤ 100 LP, >100 mg/dL; and <0, ≥0 < 30, ≥30 < 40, ≥40 < 50, ≥50% respectively [3], to investigate the effect of different levels of LDL-C at follow up on CAPs. According to statins, the arms were grouped to: rosuvastatin, atorvastatin, pitavastatin, simvastatin, fluvastatin and pravastatin groups, to investigate the effect of different statins on CAPs. The volume of CAP at follow up was compared with that at baseline to evaluate effect of LDL-C levels on regression of CAP.

Heterogeneity across trials (arms) was assessed via a standard χ ² test with significance being set at p < 0.10 and also assessed by means of I ² statistic with significance being set at I ²  > 50%. Pooled analyses were calculated using fixed-effect models, whereas random-effect models were applied in case of significant heterogeneity across trials (arms). Sensitivity analyses (exclusion of one study at one time) were performed to determine the stability of the results. Publication bias was assessed using the Egger regression asymmetry test. Statistical analyses were performed using STATA software 12.0 (StataCorp, College Station, Texas).

All continuous variables were expressed as mean ± SD, and continuous variables were compared between the Western and Asian groups using Student’s t test (SigmaStat 3.5). A P value <0.05 was considered to be statistically significant.

The flow of selecting studies for the meta-analysis was shown in Figure 1. Briefly, of the initial 673 articles, one hundred and twenty-two of abstracts were reviewed, resulting in exclusion of 102 articles, and 20 articles were reviewed in full text, resulting in exclusion of 10 trials and inclusion of 18 additional trials cited in the 20 articles. Twenty two RCTs [4‐25] and six blinded end-points trial [26‐31] were carefully evaluated, and eight trials [4, 8, 9, 18, 19, 21, 27, 31] were excluded because of specific the index of plaque or lack of some data. Sixteen RCT (ESTABLISH [11], REVERSAL [10], A-PLUS [5], ACTIVATE [6], ILLUSTRATE [7], JAPAN-ACS [20], REACH [14], SATURN [16], ARTMAP [17], ERASE [23], STRADIVARIUS [24], PERISCOPE [25], and trials by Yokoyama M [12], by Kawasaki M [13], by Hong MK [15], and Tani S [22]) and four blinded end-points trial (ASTEROID [26], COSMOS [29], trial by Jensen LO [28] and trial by Nasu K [30]) were finally analyzed.

The characteristics of the included trials were as same as in the study [3] and shown in Table 1. Briefly, among the 20 trials, 10 trials are completed in European, America and Australia [10, 5‐7, 16, 23‐26, 28], 10 in Asia [20, 11‐15, 17, 22, 29, 30], and there were 15 trials assessing statins (statin vs. usual care in 6 trials [11‐14, 22, 30]; intensive statin vs. moderate statin treatment in 5 trials [10, 15‐17, 20]; follow up vs baseline in 3 trial [26, 28, 29], before acute coronary syndrome (ACS) vs after ACS in one trial [23]), 2 trials assessing enzyme acyl–coenzyme A: cholesterol acyltransferase (ACAT) inhibition [5, 6], one trial assessing cholesteryl ester transfer protein (CETP) inhibitor torcetrapib [7], one trial assessing a decreasing obesity drug: rimonabant [24], and one trial assessing glucose-lowering agents [25]. Overall, 5910 patients with coronary heart disease (CHD) underwent serial IVUS examination for evaluating regression of CAP. Follow-up periods ranged from 2 to 24 months. The levels of LDL-C of each arm at baseline and follow-up were shown in Table 2.Risk of bias of included studies, evaluated through Cochrane’s methods, showed an overall acceptable quality of selected trials (Figures 2 and 3).

For Western, meta-analysis indicated that LDL-C lowering in group ≤70 mg/dL could lead to regression of CAP, but LDL-C lowering in group >70 ≤ 100 HP, >70 ≤ 100 MP, >70 ≤ 100 LP and >100 mg/dL could not (Figure 4, Table 3).

In group ≤70 mg/dL (including three arms) with mean 23.1 months of follow up, the volumes of CAP (160.6 mm³) at follow up were significantly decreased, compared with the volumes (171.4 mm³) at baseline [SMD −0.156 mm³, 95% CI (confidence interval) -0.248 ~ −0.064, p = 0.001]. There was no significant heterogeneity among arms (χ ² for heterogeneity = 0.33, p =0.886, I² = 0%).

Sensitivity analyses suggested that LDL-C lowering in group ≤70 mg/dL could lead to regression of CAP with reduction of the CAP volume ranged from −0.139 mm³ (SMD, 95% CI: −0.257 ~ −0.021) when the arm of 2006 ASTEROID Ros was omitted to −0.175 mm³ (SMD, 95% CI: −0.317 ~ −0.034) when the arm of 2011 SATURN Ros was omitted. No publication bias was found, the values of p by Egger’s test was 0.789.

LDL-C lowering in group ≤70 mg/dL and >70 ≤ 100 HP could lead to regression of CAP, but LDL-C lowering in group >100 mg/dL could not (Figure 5, Table 3).

In group ≤70 mg/dL (including four arms) with mean 6.9 months of follow up and group >70 ≤ 100HP mg/dL (including eight arms) with mean 11.0 months of follow up, the volumes of CAP (179.9, 87.5 mm³ respectively) at follow up were significantly decreased, compared with the volumes (192.2, 96.4 mm³ respectively) at baseline [SMD −0.157 mm³, 95% CI −0.307 ~ −0.008, p = 0.039; SMD −0.211 mm³, 95% CI −0.331 ~ −0.092, p = 0.001; respectively]. There was no significant heterogeneity among arms (χ ² for heterogeneity = 0.24, p =0.955, I² = 0% for group ≤70 mg/dL; χ ² for heterogeneity = 2.68, p =0.913, I² = 0% for group >70 ≤ 100HP mg/dL).

Sensitivity analyses suggested that LDL-C lowering in group >70 ≤ 100 HP mg/dL could lead to regression of CAP with reduction of the CAP volume ranged from −0.177 mm³ (SMD, 95% CI: −0.314 ~ −0.040) when the arm of 2009 JAPAN-ACS Ato was omitted to −0.231 mm³ (SMD, 95% CI: −0.368 ~ −0.094) when the arm of 2009 COSMOS Ros was omitted; but that LDL-C lowering in group ≤ 70 mg/dL could not significantly lead to regression of CAP with reduction of the CAP volume when the arm of 2012 ARTMAP Ros or 2012 ARTMAP Ato was omitted (Table 3).

For Western, meta-analysis showed that LDL-C lowering in group ≥40 < 50, ≥50% could lead to regression of CAP, but LDL-C lowering in group <0, ≥0 < 30% and ≥30 < 40 could not (Figure 6, Table 3).

In group ≥40 < 50% (including four arms) with mean 22.6 months of follow up, the volumes of CAP (143.1 mm³) at follow up were significantly decreased, compared with the volumes (148.8 mm³) at baseline (SMD −0.095 mm³, 95% CI −0.171 ~ −0.019, p = 0.014). There was no significant heterogeneity among arms (χ ² for heterogeneity = 1.64, P = 0.651, I² = 0%).

Sensitivity analyses showed that LDL-C lowering in group ≥40 < 50 could still lead to regression of CAP with reduction of the plaque volume ranged from −0.065 mm³ (95% CI −0.163 ~ 0.032) when the arm of 2011 SATURN Ros was omitted to −0.116 mm³ (SMD, 95% CI −0.201 ~ −0.032) when 2004 REVERSAL Ato was omitted. Publication bias analysis suggested the values of p by Egger’s test were 0.804.

In group group <0, ≥0 < 30% and ≥30 < 40, meta-analysis were showed in Table 3.

LDL-C lowering in group ≥30 < 40, ≥40 < 50% could lead to regression of CAP, but LDL-C lowering in group ≥0 < 30% could not (Figure 7, Table 3).

In group ≥30 < 40% (including nine arms) with mean 10.9 months of follow up, and group ≥40 < 50% (including four arms) with mean 6.9 months of follow up, the volumes of CAP (90.0, 179.9 mm³ respectively) at follow up were significantly decreased, compared with the volumes (98.6, 192.2 mm³ respectively) at baseline (SMD −0.206 mm³, 95% CI −0.324 ~ −0.088, p = 0.001; SMD −0.157 mm³, 95% CI −0.307 ~ −0.008, p = 0.039; respectively). There was no significant heterogeneity among arms (χ ² for heterogeneity = 2.91, P = 0.840, I² = 0%; χ ² for heterogeneity = 0.33, p =0.955, I² = 0%; for group ≥30 < 40, and group ≥40 < 50 respectively).

Sensitivity analyses showed that LDL-C lowering in group ≥30 < 40% could still lead to regression of CAP with reduction of the plaque volume ranged from −0.172 mm³ (95% CI −0.306 ~ −0.038) when the arm of 2009 JAPAN-ACS Ato was omitted to −0.223 mm³ (SMD, 95% CI −0.357 ~ −0.089) when 2009 COSMOS Ros was omitted. Publication bias analysis suggested that bias was significant with 0.004 of p value by Egger’s test.

Mean levels of LDL-C at baseline and follow up, mean reducing percentage of LDL-C in each group were showed in Table 4.

For Western, atorvastatin, rosuvastatin, pravastatin and simvastatin were used in trials to investigate the effects of LDL-C lowering on CAP. Meta-analysis indicated that LDL-C lowering by rosuvastatin could lead to regression of CAP, but LDL-C lowering by atorvastatin, pravastatin, and simvastatin could not (Figure 8, Table 5).

LDL-C lowering by rosuvastatin (mean 40.0 mg daily for mean 24 months) could significantly decrease the volumes of CAP at follow up, compared with the volumes at baseline (SMD −0.158 mm³, 95% CI: −0.253 ~ −0.064, p = 0.001). There was no significant heterogeneity among arms (χ ² for heterogeneity = 0.18, p =0.672, I² = 0%).

Sensitivity analyses suggested that lowering LDL-C by rosuvastatin could lead to regression of CAP with reduction of the plaque volume ranged from −0.142 mm³ (SMD, 95% CI: −0.263 ~ −0.020) when the arm of 2006 ASTEROID Ros was omitted to −0.183 mm³ (SMD, 95% CI: −0.332 ~ −0.035) when the arm of 2011 SATURN Ros was omitted. But publication bias was found, the values of p by Egger’s test was 0.000 (Table 5).

For Asian, atorvastatin, rosuvastatin, pitavastatin, pravastatin, fluvastatin and simvastatin were used in trials to investigate the effects of LDL-C lowering on CAP. Meta-analysis indicated that LDL-C lowering by rosuvastatin, atorvastatin could lead to regression of CAP, but LDL-C lowering by pitavastatin, pravastatin, fluvastatin and simvastatin could not (Figure 9, Table 5).

LDL-C lowering by rosuvastatin (mean 14.1 mg daily for mean 10.3 months), atorvastatin (mean 18.9 mg daily for mean 7.8 months) could significantly decrease the volumes of CAP at follow up, compared with the volumes at baseline (SMD −0.172 mm³, 95% CI: −0.331 ~ −0.012, p = 0.035; SMD −0.185, 95% CI: −0.330 ~ −0.040, p = 0.013; respectively). There was no significant heterogeneity among arms (χ ² for heterogeneity = 0.17, p =0.917, I² = 0% for rosuvastatin; χ ² for heterogeneity = 1.94, p =0.858, I² = 0% for atorvastatin).

Sensitivity analyses suggested that lowering LDL-C by rosuvastatin could not significantly lead to regression of CAP when the arm of 2012 ARTMAP Ros or 2009 COSMOS Ros was omitted. Also, Lowering LDL-C by atorvastatin could not significantly lead to regression of CAP when the arm of 2009 JAPAN-ACS Ato was omitted. No publication bias was found, the values of p by Egger’s test for rosuvastatin and atorvastatin group were 0.660, 0.456 respectively (Table 5).

Intensity of lowering LDL-C by different statins was shown in Table 6. Rosuvastatin and atorvastatin could reduce LDL-C by more than 40%.

The meta analysis showed that rosuvastatin and atorvastatin can regress CAP (Table 5). LDL-C levels, intensity of lowering LDL-C by rosuvastatin and atorvastatin, its dosage and duration were compared between Western and Asian (Table 7). Intensity of lowering LDL-C by rosuvastatin and atorvastatin in Western group were similar to that in Asian group, but the dosages of rosuvastatin and atorvastatin in Asian group were significantly lower than those in Western group, and the duration of statins administration in Asian group were significantly shorter than those in Western, as showed in Table 7.

For Western including American, Canadian, German, French, English, Australian and Dane [10, 5‐7, 16, 23‐26, 28], the meta-analysis (Table 3) in subgroup ≤70 mg and ≥40 < 50% of Western indicated that LDL-C level lowering to <69.3 mg or reducing by > 45% for 22.6 months of follow up (Table 4) could lead to regression of CAP, but the meta-analysis (Table 3) in subgroup >70 ≤ 100 HP mg of Western showed that LDL-C level lowering to 73.2 mg or reducing by 43.6% for 21.7 months of follow up (Table 4) was not enough for regressing CAP.

For Asian including Japanese and Korean [20, 11‐15, 17, 22, 29, 30], the meta-analysis in subgroup ≤70 mg and ≥40 < 50% of Asian indicated that LDL-C level lowering to 57.0 mg or reducing by 47.2% for 6.9 months of follow up could lead to regression of CAP, but sensitivity analyses showed that LDL-C lowering in this two subgroup could not significantly lead to regression of CAP when the arm of 2012 ARTMAP Ros or 2012 ARTMAP Ato was omitted (Table 3). The meta-analysis in subgroup ≥ 30 < 40% of Asian indicated that LDL-C level lowering to 84.6 mg or reducing by 36.0% for 10.9 months of follow up could also lead to regression of CAP, but publication bias was significant. The meta-analysis in subgroup >70 ≤ 100HP mg of Asian with good sensitivity and no publication bias indicated that LDL-C level lowering to 84.0 mg or reducing by 36.1% for 11 months of follow up with could lead to regression of CAP (Table 3).

Taken all the results of meta-analysis together, for Western, it was recommended that LDL-C level might be reduced by >45% or to a target level < 69 mg/dL for regressing CAP; for Asian, LDL-C level might be reduced by >36% or to a target level < 84 mg/dL.

Whether statins has different effect on Westerns and Asians remains to be settled.

The study by Lee E et al. [34] and MEGA Study [35] suggested statins have different effects on Westerns and Asians. In 2005, Lee E et al. [34] prospectively examined the pharmacokinetics of rosuvastatin in White and Asian individuals living in Singapore, and reported that plasma exposure to rosuvastatin and its metabolites was significantly higher in Chinese, Malay, and Asian-Indian subjects compared with Western subjects living in the same environment. But the mechanisms underlying ethnic differences in rosuvastatin disposition remain to be unearthed [36]. MEGA Study [35] indicated that a small dose of pravastatin that was half the dose administered to western patients, reduced LDL-C by 19-22% (which is lower than that reductions of 23–35% in western patients), but could substantially reduce the risk of coronary heart disease in Japanese.

But two meta-analysis did not demonstrate the difference of rosuvastatin and atorvastatin on Westerns and Asians. The meta-analysis including the 36 trials of pharmacodynamics of rosuvastatin in Western and Asian hypercholesterolemia patients did not confirm that there was significant difference in the exposure-response relationship for LDL-C reduction between Westerners and Asians [33].The meta-analysis including 22 pharmacokinetic studies also demonstrated no differences in the systemic exposure to atorvastatin between Asian and Caucasian subjects [32].

Our meta-analysis revealed that there were difference of rosuvastatin and atorvastatin in lowering LDL-C and regressing CAP between Westerns and Asians. The meta-analysis of rosuvastatin including 2 trials with 869 Western patients indicated that 40 mg of rosuvastatin daily for 24 months with reducing LDL-C by 49.9% could regress CAP. But the meta-analysis of rosuvastatin including 3 trials with 304 Asian patients showed that 14.1 mg of rosuvastatin daily for 10.3 months with reducing LDL-C by 44.0% could also regress CAP though the result of sensitivity analyses is not as good as that in Western (Table 5). The meta-analysis of atorvastatin including 2 trials with 772 Western patients showed that 80 mg of atorvastatin daily for 22 months with reducing LDL-C by 43.0% could not significantly regress CAP. But the meta-analysis of atorvastatin including 6 trials with 366 Asian patients demonstrated that 18.9 mg of atorvastatin daily for 7.8 months with reducing LDL-C by 40.7% could significantly regress CAP though the result of sensitivity analyses is not as good as that expected (Table 5).

Comparison between Western and Asian in using rosuvastatin and atorvastatin indicated that the dosages of rosuvastatin and atorvastatin in Asian group were significantly lower than those in Western (Table 7).

Based on this meta-analysis, reducing LDL-C by >40% in Westerns need atorvastatin 80 mg or rosuvastatin 40 mg, but in Asians need only atorvastatin 18.9 mg or rosuvastatin 14.1 mg. For regressing CAP, 40 mg of rosuvastatin might be daily administrated in Western for 24 months; 14.1 mg of rosuvastatin or 18.9 mg of atorvastatin might be daily administrated in Asian for 10.7 or 7.8 months respectively.

As with the meta-analysis [3], this study has some limitations. There might be publication bias, difference of the method detected and follow up duration. But those differences in measurements and plaque selection did not affect the change of the target plaque with LDL-C levels. So, it has little effect on homogeneous of studies, and on the relationship between CAP change and LDL-C level. But the trials of single statin on LDL-C and CAP of specific population (for example, 2 trials about atorvastatin on Western with 727 participants or 6 on Asian with 366 in Table 5) were limited, the effect of statin on specific population remains to be investigated. The duration of follow up between Western and Asian was different (Table 4, 6 and 7), and treatment duration might have some effect on CAP regression. But the trials from Asian and Western were respectively meta-analysed in this study. Therefore, the difference in follow-up duration between Asian and Western did not influence the results of the meta-analysis. The CAP regression in short period of statins therapy in Asian suggested that the CAPs in Asian were easily regressed by statins.

This meta-analysis did not investigate the effect of reduction of LDL-C on adverse cardiovascular events because all participants of the included trial must be alive at follow up. But in the Extended-ESTABLISH study, the incidence of adverse cardiovascular events in statin group with CAP regression were reduced to half that seen in the control group [37]. In the Extended JAPAN-ACS study [38], there was no significantly different association of incidence of adverse cardiovascular events with the CAP regression extent, but that greater external elastic membrane volume regression (<−6.56%) had a significantly lower incidence of cumulative events than the lesser suggested the importance of CAP regression in reducing adverse cardiovascular events. A meta-analysis [39] included 7864 CAD patients showed that rates of plaque volume regression were significantly associated with the incidence of MI or revascularization.