**3. Results and discussion**

A total of 721 articles were initially identified after electronic search from January 1, 2004, to January 1, 2015, was restricted to animal studies using usual PubMed filters. Three duplicated articles were subtracted, and 719 unique publications were screened by title and abstract review. In this continuum, articles dealing with non-New Zealand rabbit strains and studying imaging end points like MRI, CT scan, and radiology have been excluded (see Figure 1). The remaining articles were submitted to detail inspection, and articles that focused on inflamma‐ tory and histopathologic end points of atherosclerosis without reporting lipid and lipoprotein profiles of studied animals were excluded. In total, 32 eligible studies were included in the systematic review and meta-analysis involving 1104 rabbit subjects [10–41]. We just considered control and cholesterol-fed (treated; model) rabbits for meta-analytic assessment, and other animal groups that received other interventions (drug, nutrients, etc.) have been excluded from the data set. The study identification process is shown in Figure 1.

Of the included RCT, 30 studies reported on outcomes of triacylglycerols (TGs; *n* = 638 animal subjects), 32 studies reported on total cholesterol (TC; *n* = 677 animal subjects), 26 studies reported on LDL-C (*n* = 595 animal subjects), 30 studies reported on high-density lipoprotein cholesterol (HDL-C; *n* = 571 animal subjects), 4 studies reported on LDL-C/HDL-C (*n* = 98 animal subjects), 1 study reported on LDL-C/TC (*n* = 8 animal subjects), 2 studies reported on oxidized LDL-C (ox-LDL-C; *n* = 32 animal subjects), and 3 studies reported on TC/HDL-C (*n* = 46 animal subjects).

Age in the start of studies varied from 1.86 to 9 months old with an average value of 4.37 ± 2.52 months old in 9 studies, and it was not reported in the majority of studies. The duration of nutritional intervention went from 4 to 56 weeks with an average value of 8.90 ± 7.26 weeks in all studies (Figure 2). The majority of studies (56%) reported ≤10 months, while 37.5% of studies reported 10–20 weeks of nutritional intervention to translate a rabbit model of atherosclerosis. The amount of dietary cholesterol surplus that utilized to induce atherosclerosis varied from 0.34% to 4.00% with an average value of 0.98 ± 0.67% in all studies (Figure 3). In this context, 31% of all eligible studies used 0.5% cholesterol in feed, while 46% of studies used 1% choles‐ terol in their feed.

**2.2. Statistical analysis**

6 Lipoproteins - From Bench to Bedside

accordingly.

**3. Results and discussion**

= 46 animal subjects).

Effect sizes are indices that measure the magnitude of the differences between two groups. For each comparison, individual RCT data for each outcome measure and combined measure were pooled to calculated standardized mean difference (SMD) effect size considering *P* < 0.05 significant level [42] using the Comprehensive Meta-Analysis ver.2.2.064, a software package

To show substantial heterogeneity among studies, we report fixed-, random-, and mixedeffects meta-analysis [43]. In this continuum, random-effects meta-analysis takes into account the precision of discrete studies and the variation among studies and weights of each study

We conducted two subgroup analyses to explore association of dietary cholesterol and duration of cholesterol intake with outcome measures. For each subgroup category, overall net change estimates were calculated using fixed-effects, random-effects, and mixed-effects models, and the heterogeneity of estimates was assessed. We conducted a meta-regression analysis to further examine the effects of cholesterol intake as an explanatory factor on outcome variables using random-effects meta-regression (unrestricted maximum likelihood (UREML)).

A total of 721 articles were initially identified after electronic search from January 1, 2004, to January 1, 2015, was restricted to animal studies using usual PubMed filters. Three duplicated articles were subtracted, and 719 unique publications were screened by title and abstract review. In this continuum, articles dealing with non-New Zealand rabbit strains and studying imaging end points like MRI, CT scan, and radiology have been excluded (see Figure 1). The remaining articles were submitted to detail inspection, and articles that focused on inflamma‐ tory and histopathologic end points of atherosclerosis without reporting lipid and lipoprotein profiles of studied animals were excluded. In total, 32 eligible studies were included in the systematic review and meta-analysis involving 1104 rabbit subjects [10–41]. We just considered control and cholesterol-fed (treated; model) rabbits for meta-analytic assessment, and other animal groups that received other interventions (drug, nutrients, etc.) have been excluded from

Of the included RCT, 30 studies reported on outcomes of triacylglycerols (TGs; *n* = 638 animal subjects), 32 studies reported on total cholesterol (TC; *n* = 677 animal subjects), 26 studies reported on LDL-C (*n* = 595 animal subjects), 30 studies reported on high-density lipoprotein cholesterol (HDL-C; *n* = 571 animal subjects), 4 studies reported on LDL-C/HDL-C (*n* = 98 animal subjects), 1 study reported on LDL-C/TC (*n* = 8 animal subjects), 2 studies reported on oxidized LDL-C (ox-LDL-C; *n* = 32 animal subjects), and 3 studies reported on TC/HDL-C (*n*

statistics [44].

developed by Biostat (http://www.meta-analysis.com/; Englewood, NJ 08631 USA).

The heterogeneity of studies was quantified using chi-square test, *Q*, and *I*<sup>2</sup>

Begg's [45] and Egger's tests [46] were employed to identify publication bias.

the data set. The study identification process is shown in Figure 1.

The approximate analysis and the major ingredients of rations were not usually reported in studies. However, in 40.6% of studies, a range of fat (1% to 10%) with an average value of (5.03 ± 2.80%) from diverse sources like lard has been reported. The 5% fat content has been more frequently (30.76%) utilized among studies that reported the fat content of diets. However, it is not precisely clear from studies that the fat added to the ration is the whole content of dietary fat or surplus fat. Experimental diets must be formulated according to the rabbit requirements of NRC [47] because all macronutrients (carbohydrate, fat, and protein), micronutrients (minerals and vitamins), and other diet-specific bioactive compounds may modulate lipid metabolism and cytokine milieu that lead to atherosclerosis or prevent atherosclerosis. In addition, the cholesterol content of diets needs to be determined since 0.34% cholesterol in diet can trigger dyslipidemia in rabbits, depending on the duration of treatment. The lack of dietary formulations and the approximate analysis are two major shortcomings of studies that report cholesterol-induced atherosclerosis in rabbits.

No consensus for initial weights of rabbits were observed among eligible studies as well as on the changes of weight during experiment and the final weight of the studied animal at the end of dietary intervention. The initial weight of rabbits employed to translate atherosclerosis changed from 1.5 to 3.25 kg with an average value of 2.45 ± 0.59 kg. Final weights (3.29 ± 0.21 kg) were reported only in four studies, in which their rabbits were fed 1% cholesterol for an average duration of 2.17 ± 0.84 months of nutritional intervention. By pooling data from 32 RCTs of 1104 rabbit subjects, the current meta-analysis documented lipid and lipoprotein alterations associated with increased intake of cholesterol. However, dietary formulations, duration of cholesterol feeding, initial and final weights of animals or their weight changes during study, and concise age of animals were not reported in most of these studies and would cause heterogeneity.

The effect of dietary cholesterol inclusion on the combined outcome of lipid and lipoprotein profiles and indices of rabbits in a random-effect model as an analysis model for the metaanalyses was 5.618 (95% confidence interval (CI): 4.592, 6.644; *P* = 0.0001). In this way, SMD effects of dietary cholesterol inclusion were 2.424 mg/dl (95% CI: 1.531, 3.318 mg/dl; *P* = 0.000009) for HDL-C, 7.646 mg/dl (6.266, 9.026; *P* = 0.000000000) for LDL-C, 5.216 (95% CI: 2.155, 8.278; *P* = 0.001) for LDL-C/HDL-C, 4.658 ng/ml (95% CI: 3.321, 5.995 ng/ml; *P* = 0.000000000) for ox-LDL-C, 8.643 mg/dl (95% CI: 7.258, 10.028 mg/dl; *P* = 0.00000000) for TC, 4.392 (95% CI: 2.317, 6.466; *P* = 0.00003) for TC/HDL-C, and 3.173 mg/dl (95% CI: 2.451, 3.896 mg/dl; *P* = 0.000000000) for TGs. The value of *I* <sup>2</sup> was 89.387%, indicating that nearly all of the variation is real and sturdily supporting the subgroup analysis and/or the meta-regression. By convention, an *I* <sup>2</sup> value >75% indicates significant between-study heterogeneity [4]. Subgroup analyses according to whether lipid and lipoprotein changes occurred following the stratifi‐ cation of studies according to the duration of cholesterol feeding in cholesterol-fed rabbits. In this way, based on a mixed-effects analysis model, SMD effects of the duration of cholesterol inclusion was 2.788 (95% CI: 2.333, 3.244; *P* = 0.000009). According to the *Q* test (*Q* = 112.206, *P* < 0.001), a significant level of heterogeneity has been concluded in the subgroup analysis of studies based on the duration of cholesterol feeding, which reflects the diversity of study design among the 28 comparisons.

Figure 2. The duration of dietary cholesterol surplus used in a rabbit model of atherosclerosis. **Figure 2.** The duration of dietary cholesterol surplus used in a rabbit model of atherosclerosis.

**Study**

Figure 3. The dietary cholesterol surplus used in a rabbit model of atherosclerosis.

Figure 3. The dietary cholesterol surplus used in a rabbit model of atherosclerosis.

In this continuum, the subgroup analysis based on duration (week) of cholesterol feeding

in a mixed-effects analysis showed combined SMD effects, that is, 2.788 (95% CI: 2.333, 3.244;

*P* = 0.000; *Q* = 112.206; df = 14). Furthermore, SMD effects were 1.414 mg/dl (95% CI: 1.049,

1.780; *P* = 0.000000000; *Q* = 166.780; df = 14) for HDL-C, 4.3888 mg/dl (95% CI: 3.779,

4.996; *P* = 0.000000000; *Q* = 138.362; df = 13) for LDL-C, 3.367 (95% CI: 2.686, 4.048; *P* =

0.00000; *Q* = 35.845; df = 3) for LDL-C/HDL-C, 4.658 ng/ml (95% CI: 3.321, 5.995 ng/ml; *P* =

0.000000000; *Q* = 0.163; df = 1) for ox-LDL-C, 3.942 mg/dl (95% CI: 3.408, 4.475 mg/dl; *P* =

0.00000000; *Q* = 184.421; df = 17) for TC, 3.725 (95% CI: 2.733, 4.718; *P* = 0.000; *Q* = 6.663;

df = 2) for TC/HDL-C, and 2.161 mg/dl (95% CI: 1.792, 2.529 mg/dl; *P* = 0.000000000; *Q* =

79.221; df = 15) for TGs. The duration of cholesterol intake was an important determinant of the

In this continuum, the subgroup analysis based on duration (week) of cholesterol feeding

in a mixed-effects analysis showed combined SMD effects, that is, 2.788 (95% CI: 2.333, 3.244;

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 **Figure 3.** The dietary cholesterol surplus used in a rabbit model of atherosclerosis.

0

In this continuum, the subgroup analysis based on duration (week) of cholesterol feeding in a mixed-effects analysis showed combined SMD effects, that is, 2.788 (95% CI: 2.333, 3.244; *P* = 0.000; *Q* = 112.206; df = 14). Furthermore, SMD effects were 1.414 mg/dl (95% CI: 1.049, 1.780; *P* = 0.000000000; *Q* = 166.780; df = 14) for HDL-C, 4.3888 mg/dl (95% CI: 3.779, 4.996; *P* = 0.000000000; *Q* = 138.362; df = 13) for LDL-C, 3.367 (95% CI: 2.686, 4.048; *P* = 0.00000; *Q* = 35.845; df = 3) for LDL-C/HDL-C, 4.658 ng/ml (95% CI: 3.321, 5.995 ng/ml; *P* = 0.000000000; *Q* = 0.163; df = 1) for ox-LDL-C, 3.942 mg/dl (95% CI: 3.408, 4.475 mg/dl; *P* = 0.00000000; *Q* = 184.421; df = 17) for TC, 3.725 (95% CI: 2.733, 4.718; *P* = 0.000; *Q* = 6.663; df = 2) for TC/HDL-C, and 2.161 mg/dl (95% CI: 1.792, 2.529 mg/dl; *P* = 0.000000000; *Q* = 79.221; df = 15) for TGs. The duration of cholesterol intake was an important determinant of the heterogeneity in studies with regard to lipid and lipoprotein outcomes. Based on *Q* values of outcomes following the subgroup analysis and their comparison with the chi-square table, the high level of heterogeneity has been detected for all outcomes except TC/HDL-C as an atherogenic index. This strange result may be due to the low sample size (*n* = 3) of the TC/HDL-C index. Therefore, the duration of cholesterol intake may not play an essential role in inducing dyslipidemic atherogenesis in cholesterol-fed rabbit models.

variation is real and sturdily supporting the subgroup analysis and/or the meta-regression. By

analyses according to whether lipid and lipoprotein changes occurred following the stratifi‐ cation of studies according to the duration of cholesterol feeding in cholesterol-fed rabbits. In this way, based on a mixed-effects analysis model, SMD effects of the duration of cholesterol inclusion was 2.788 (95% CI: 2.333, 3.244; *P* = 0.000009). According to the *Q* test (*Q* = 112.206, *P* < 0.001), a significant level of heterogeneity has been concluded in the subgroup analysis of studies based on the duration of cholesterol feeding, which reflects the diversity of study

Duration

Cholesterol%

**Figure 2.** The duration of dietary cholesterol surplus used in a rabbit model of atherosclerosis.

Cholesterol%

<sup>2</sup> value >75% indicates significant between-study heterogeneity [4]. Subgroup

Figure 2. The duration of dietary cholesterol surplus used in a rabbit model of atherosclerosis.

Figure 2. The duration of dietary cholesterol surplus used in a rabbit model of atherosclerosis.

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

**Study**

Figure 3. The dietary cholesterol surplus used in a rabbit model of atherosclerosis.

Figure 3. The dietary cholesterol surplus used in a rabbit model of atherosclerosis.

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

**Study**

**Study**

In this continuum, the subgroup analysis based on duration (week) of cholesterol feeding

in a mixed-effects analysis showed combined SMD effects, that is, 2.788 (95% CI: 2.333, 3.244;

*P* = 0.000; *Q* = 112.206; df = 14). Furthermore, SMD effects were 1.414 mg/dl (95% CI: 1.049,

1.780; *P* = 0.000000000; *Q* = 166.780; df = 14) for HDL-C, 4.3888 mg/dl (95% CI: 3.779,

4.996; *P* = 0.000000000; *Q* = 138.362; df = 13) for LDL-C, 3.367 (95% CI: 2.686, 4.048; *P* =

0.00000; *Q* = 35.845; df = 3) for LDL-C/HDL-C, 4.658 ng/ml (95% CI: 3.321, 5.995 ng/ml; *P* =

0.000000000; *Q* = 0.163; df = 1) for ox-LDL-C, 3.942 mg/dl (95% CI: 3.408, 4.475 mg/dl; *P* =

0.00000000; *Q* = 184.421; df = 17) for TC, 3.725 (95% CI: 2.733, 4.718; *P* = 0.000; *Q* = 6.663;

df = 2) for TC/HDL-C, and 2.161 mg/dl (95% CI: 1.792, 2.529 mg/dl; *P* = 0.000000000; *Q* =

79.221; df = 15) for TGs. The duration of cholesterol intake was an important determinant of the

In this continuum, the subgroup analysis based on duration (week) of cholesterol feeding

in a mixed-effects analysis showed combined SMD effects, that is, 2.788 (95% CI: 2.333, 3.244;

convention, an *I*

8 Lipoproteins - From Bench to Bedside

design among the 28 comparisons.

**Duration of dietary intervention** 

**(weeks)**

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

**Figure 3.** The dietary cholesterol surplus used in a rabbit model of atherosclerosis.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

**Cholesterol percentage in feed**

**Cholesterol percentage in feed**

Subgroup analyses according to whether lipid and lipoprotein changes occurred following the stratification of studies according to the amount of cholesterol feeding in cholesterol-fed rabbits. In this continuum, the subgroup analysis based on dietary cholesterol inclusion (%) in of cholesterol-fed rabbits in a mixed-effects analysis showed SMD effects 5.538 (95% CI: 4.613, 6.463; *P* = 0.000; *Q* = 31.622;df = 6). Further, the subgroup analysis based on the amount of cholesterol feeding in a mixed-effects analysis showed SMD effects 1.743 mg/dl (95% CI: 1.145, 2.341; *P* = 0.000000000; *Q* = 77.709; df = 6) for HDL-C, 7.409 mg/dl (95% CI: 6.173, 8.646; *P* = 0.000000000; *Q* = 138.362; df = 4) for LDL-C, 3.208 (95% CI: 1.680, 4.736; *P* = 0.000000000; *Q* = 0.967; df = 1) for LDL-C/HDL-C, 4.658 ng/ml (95% CI: 3.321, 5.995 ng/ml; *P* = 0.000000000; *Q* = not determined; df = 0) for ox-LDL-C, 8.543 mg/dl (95% CI: 7.309, 9.776 mg/dl; *P* = 0.00000000; *Q* = 41.307; df = 6) for TC, 4.255 (95% CI: 2.852, 5.658; *P* = 0.000; *Q* = 3.691;df = 1) for TC/HDL-C, and 3.107 mg/dl (95% CI: 2.491, 3.723 mg/dl; *P* = 0.000000000; *Q* = 23.675; df = 6) for TGs.

As the amount of dietary cholesterol surplus was considered as a moderator, the *P* values of *Q*<sup>R</sup> in mixed-model regression were significant for TGs, TC, LDL-C/HDL-C, LDL-C, HDL-C, and combined lipid and lipid profile (*P* < 0.01; Table 1 and Figure 4), while the *P* values of *Q*<sup>R</sup> in mixed-model regression were nonsignificant for TC/HDL-C (*P* = 0.31731; Table 1). Randomeffects meta-regression (UREML) was conducted for change in all parameters and indices using the cholesterol moderator (data not shown). The ANOVA table of fixed- and mixedeffect models for combined lipid and lipoprotein profile regression is shown in Table 1. Accordingly, *Q*<sup>M</sup> is significant. Then at least one of the regression coefficients between dietary cholesterol inclusion and lipid and lipoprotein profiles and indices is different from zero. In this sense, the funnel plot (Figure 5) and the results of Egger's tests showed a significant publication bias (data not shown). The subgroup analysis based on the amount of dietary cholesterol surplus showed heterogeneity in all outcomes except LDL-C/HDL-C (*n* = 2) and TC/HDL-C (*n* = 2) after comparing their *Q* values using the chi-square table. Therefore, cholesterol content may not be a pivotal determinant in inducing dyslipidemia in cholesterolfed rabbit models, but it raised all lipid and lipoprotein profile and indices in cholesterol-fed rabbits. Moreover, no consensus for atherogenic diets was observed based on their cholesterol and fat contents and the types of their fat, including pure cholesterol, lard, egg yolk powder, coconut oil, and/or safflower oil as well as different nutritional interventions.


*Q*M—model sum of squares compared to chi-square distribution with *p*– 1 df (*p* is number of predictors in the model); *Q*<sup>R</sup> —residual sum of squares compared to chi-square distribution with *k*– *p*– 1 df (*k* is the number of studies) [48].

**Table 1.** Fixed- and mixed-effect models—analysis of variance (ANOVA) table for combined lipid and lipoprotein profile regression.

The resulting funnel graphs and the associated statistics based on Egger et al. [46] revealed a significant asymmetry and publication bias among studies. Therefore, publication or other sources of bias could be relevant in evaluating an overall effect of the selected nutritional interventions in lipid and lipoprotein outcomes in the cholesterol-fed rabbit model of (pre)atherosclerosis.

**Figure 4.** Fixed-effect model—regression of dietary cholesterol surplus on standardized mean difference of lipid and lipid profiles and indices in a rabbit model of atherosclerosis.

**Figure 5.** Funnel plot for a meta-analysis of association between dietary cholesterol inclusion and combined effects of lipid and lipoprotein profiles and indices in a rabbit model of atherosclerosis.
