Statins Reduce Abdominal Aortic Aneurysm Growth, Rupture, and Perioperative Mortality: A Systematic Review and Meta‐Analysis

We conducted a systematic review of the medical literature according to the 2009 Preferred Reporting Items for Systematic Reviews and Meta‐Analyses statement guidelines.³⁴ Our search strategy was devised in consultation with an experienced cardiovascular librarian. We searched the Cochrane Central Register of Controlled Clinical Trials (inception to May 2018), OVID versions of MEDLINE and MEDLINE Daily including e‐publications and in‐progress and nonindexed citations (inception to June 15, 2018), and EMBASE including EMBASE Classic (inception to June 15, 2018) for studies examining the effects of statins on AAA outcomes using a combination of medical subject heading terms and keywords for AAA and statins, including a list of generic statin names in order to maximize our search sensitivity. We limited our search to human studies and did not apply any language limitations. We also hand‐searched the reference lists of any reviews identified by our strategy to maximize sensitivity and reduce publication bias. A detailed search strategy for all sources can be found in Data S1 and S2.

Study titles and abstracts were screened for relevance by 2 independent reviewers (K.S. and M.S.), and a third author (M.A.H.) resolved any discrepancies. We included all randomized and observational studies that assessed the effects of statin therapy on AAA growth, rupture, or elective perioperative mortality in nonruptured patients with AAA. We defined AAAs as aortas measuring >3 cm in diameter by any imaging modality. We attempted to limit our study population to adults with degenerative AAAs. Only studies reporting data on patients with a mean/median age >50 years old were included if AAA etiology was not specifically stated because the incidence of degenerative AAAs is known to begin after this age cutoff. We did not select for specific statins or dosing regimens, and no specific minimum duration of drug therapy was required. Studies reporting aneurysm growth as an outcome were included if at least 2 measurements of aneurysm size using any single imaging modality were used. We excluded case reports, case series (<30 patients), all reviews, and any other studies that did not meet our inclusion criteria.

Data were collected by 2 independent abstractors (K.S. and M.S.) using a prepiloted standardized electronic data collection form, and a third author (M.A.H.) adjudicated resolution of any discrepancies. Collected variables included study authors, year of publication, design, sample size, outcomes reported, cohort demographics, and covariates as well as study effect size measurements. Authors were contacted by e‐mail correspondence for additional covariate data if not grouped by statin versus control in the original publication. Study quality and bias were assessed using the Newcastle‐Ottawa Scale.³⁵

We conducted a meta‐analysis of the effects of statins on AAA growth, rupture, and elective perioperative mortality using published summary data and any additional summary data provided by authors. We calculated pooled mean differences and 95% confidence intervals (95% CIs) for AAA growth, and the effects of statins on AAA rupture and perioperative mortality were summarized using odds ratios (ORs) and 95% CIs. A subgroup analysis was conducted to assess the effects of statins on AAA growth by baseline AAA diameter (<4 cm versus ≥4 cm) because aneurysm growth rate is known to increase with aneurysm size, thereby implying a greater potential growth reduction benefit for larger aneurysms. In addition, perioperative mortality is known to differ considerably between open surgical repair and EVAR. Consequently, we analyzed the effect of statins on perioperative mortality following elective AAA repair by repair approach (open surgical repair versus EVAR). Because the approach to statin therapy has changed as understanding of high‐ versus moderate‐ and low‐intensity treatment regimens has expanded, we also analyzed the effect of statins on AAA growth and 30‐day mortality by year of study publication. We used 2007 inclusive as the cutoff because the majority of statin regimen–comparing randomized trials were published before this year.³⁶ Finally, we planned to conduct sensitivity analyses for each outcome, excluding any studies with a high probability of bias (Newcastle‐Ottawa Scale score ≤5).

All analyses were conducted in Review Manager 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Demark) based on random‐effects models, using adjusted data where available. We used the inverse‐variance method for AAA growth and the generic inverse variance method for rupture and 30‐day mortality to make use of adjusted‐effect size estimates provided by included studies. Where unavailable, crude ORs were calculated using provided event rates. Standard deviations for meta‐analysis of growth outcomes were calculated according to methods within the Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.2 if they were not reported or could not be obtained.³⁷ The presence of publication and reporting bias was assessed visually for each outcome using funnel plots. Heterogeneity was assessed using the I² statistic, and significance was set at P<0.05. Our study protocol was registered in the International Prospective Register of Systematic Reviews before study commencement (CRD42017056480). The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.

We identified 1225 publications using our search strategy, and a total of 911 were retained after removal of duplicates. Of the 911 articles screened, 854 were excluded and 57 passed screening. An additional 10 articles were identified by hand search of reviews identified by the search strategy, resulting in 67 articles for full‐text review. Following full‐text review, 45 articles were excluded for failing to report 1 of the outcomes of interest (n=12), for not comparing statins against a control group (n=10), for not studying elective AAA patients (n=10), for being abstracts without associated publications (n=7), for being clinical trial registrations without associated publication (n=2), for using duplicate data (n=3), and for being available only as a non–English language report (n=1). A total of 22 articles were included in our systematic review, and data from 21 of these were used for meta‐analysis. Our search is summarized in the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses flow diagram in Figure 1.

Our list of included studies comprised observational studies published between 2004 and 2018. One of these studies was a case‐control study, 5 were prospective cohort studies, and the remainder were retrospective cohort studies. Twelve of the included studies reported AAA growth, 5 reported or provided additional rupture rate data, and 8 reported or provided additional perioperative mortality data. The sample sizes of the included studies varied considerably for each outcome. Growth study sample sizes for the comparison of interest ranged from 121 to 2026 patients, and the follow‐up ranged from 1 to 5 years. Rupture study sample sizes ranged from 121 to 7168 patients, and follow‐up ranged from 1.9 to 3.6 years, although 2 studies did not report the length of follow‐up. Finally, the sample size for the perioperative mortality studies ranged from 400 to 34 822 patients, and all of these studies reported 30‐day mortality or a composite outcome including 30‐day mortality. These characteristics are summarized in Table 1.²⁰

The overall age of patients across the studies ranged from 65 to 74 years of age, and this age distribution was similar for each of the outcomes studied. The mean ages of patients in the statin and control groups did not differ by more than 3 years in each individual study. Sex distribution varied considerably among studies, with 67% to 99% of patients being male across studies. Similarly, some individual studies demonstrated considerable differences in sex distribution between the exposure and control groups (78% versus 88% for Kertai et al⁴⁵). Baseline AAA diameter also varied considerably across all studies and across studies reporting the same outcome. The ranges of baseline AAA diameters were 33 to 68 mm across all included studies, 33 to 46 mm for growth studies, 39 to 68 mm for rupture studies, and 46 to 59 for perioperative mortality studies. The distributions of other important covariates including hypertension, coronary artery disease, hyperlipidemia, smoking history, and diabetes mellitus also varied considerably among studies. Hypertension demonstrated a prevalence of 20% to 90% across all studies, with similar but higher frequencies in statin groups. Similarly, coronary artery disease was present in 16% to 74% of patients and generally more prevalent among statin users. The prevalence of hyperlipidemia was reported by 4 studies and ranged from 31% to 78%, with higher rates among statin users. Table 2 summarizes all available covariate data for the included studies.

Study quality assessment revealed an overall low (Newcastle‐Ottawa Scale score ≥8) to moderate (Newcastle‐Ottawa Scale score 6‐7) risk of bias. The case‐control study was at low risk of bias, with deduction of 1 point for poor case definition. Thirteen of the 20 cohort studies were at low risk of bias, 6 were at moderate risk of bias, and 1 was at high risk of bias. A common issue affecting quality among these studies was poor cohort comparability due to inadequate statistical control for important confounding variables (n=7). Other issues included inadequate description of the outcome assessment method (n=1), length of follow‐up (n=2), and follow‐up completeness/adequacy (n=2). A summary of study quality and risk of bias assessment is presented in Table 3.

Data from 12 of the 13 studies reporting growth outcomes were usable for meta‐analysis. This included a total of 7614 patients, 2589 (34%) in the statins group compared with 5025 (66%) in the control group. Meta‐analysis demonstrated that statin use was associated with a lower AAA growth rate (mean difference −0.82 mm/y, 95% CI −1.32, −0.33, P=0.001, I²=86%) (Figure 2). Subgroup analysis by baseline AAA diameter demonstrated a greater association between statin use and AAA growth rate reduction in larger aneurysms (mean difference −1.13 mm/y, 95% CI −2.09, −0.17, P=0.02, I²=90% versus mean difference −0.52 mm/y, 95% CI −0.98, −0.06, P=0.03, I²=69%, for ≥4 cm versus <4 cm AAA, respectively) (Figures 3 and 4). The Badger et al³⁸ study was not included in the meta‐analysis because annual growth rate data could not be obtained from the authors. However, their results mirrored those of our meta‐analysis in that statin patients demonstrated lower mean annual AAA growth than controls (4.5% versus 7.5%, P=0.005). Analysis of the effect of study publication year on AAA growth rate demonstrated no significant association between statin therapy and AAA growth rate for studies published up to and including 2007 (mean difference −2.72 mm/y, 95% CI −6.00, 0.56, P=0.10, I²=90%), whereas studies published after 2007 maintained a smaller but still significant association in favor of statin therapy (mean difference −0.61 mm/y, 95% CI −1.09, −0.13, P=0.01, I²=85%). Exclusion of the Sukhija et al study⁵⁵ from meta‐analysis because of low methodological quality reduced the statin effect on AAA growth rate reduction (mean difference −0.24 mm/y, 95% CI −0.40, −0.08, P=0.004, I²=85%), but the lower effect size maintained statistical significance. All of the growth rate comparisons were affected by significant heterogeneity. A funnel plot of studies investigating the effect of statins on AAA growth rate demonstrated evidence of publication bias (Figure 5). However, the missing studies appeared to be of low quality and symmetrically distributed around the pooled effect estimate.

The 4 studies that reported AAA rupture rate included 9327 patients, 1486 (16%) of whom were treated with statins compared with 7841 (84%) controls. The crude rupture rate was 34% (501/1486) for statin patients and 47% (3686/7841) for control patients. Meta‐analysis including available adjusted ORs demonstrated an association between statins and lower rates of AAA rupture (OR 0.63, 95% CI 0.51, 0.78, P<0.0001, I²=27%) (Figure 6). Heterogeneity for this comparison was low, allowing the conduct of a fixed‐effects analysis that did not appreciably change the effect of statin use on AAA rupture (OR 0.67, 95% CI 0.58, 0.77, P<0.00001). The funnel plot for this outcome demonstrated probable publication bias of methodologically low‐quality studies likely to demonstrate a smaller or no effect of statins on AAA rupture risk reduction (Figure 7).

A total of 65 340 patients were analyzed among the 8 perioperative mortality studies. Patients taking statins comprised 59% (38 235/65 340) of this cohort, and controls accounted for 41% (27 105/65 340) of the cohort. The crude 30‐day mortality rate was 1.8% (698/38 235) for the statin group and 2.6% (706/27 105) for the control group. Pooling of crude and available adjusted data demonstrated a lower perioperative mortality associated with statin use compared with controls (OR 0.64, 95% CI 0.48, 0.84, P=0.002, I²=50%) (Figure 8). Subgroup analysis by repair approach demonstrated a greater potential effect in the open‐repair subgroup (OR 0.54, 95% CI 0.30, 1.00, P=0.05, I²=71%), whereas patients undergoing elective EVAR had a smaller association with lower mortality from statin use (OR 0.63, 95% CI 0.50, 0.79, P<0.0001, I²=0%) (Figures 9 and 10). Finally, subgroup analysis by publication year revealed statin therapy to be associated with a reduction in 30‐day mortality irrespective of publication year, although studies published up to and including 2007 demonstrated a greater effect (OR 0.38, 95% CI 0.21, 0.69, P=0.002, I²=28%), and those published afterward demonstrated a more conservative effect (OR 0.77, 95% CI 0.64, 0.93, P=0.008, I²=19%). The combined elective‐repair and open‐repair subgroup comparisons suffered from significant and moderate heterogeneity, respectively. In contrast, the heterogeneity for the elective‐EVAR subgroup was 0%, and both groups for analysis by year demonstrated low heterogeneity. Visual inspection of the funnel plot for the combined elective AAA group demonstrated likely publication bias of low‐ to moderate‐quality negative studies, with a lower likelihood of missing low‐quality studies demonstrating a protective effect of statins for 30‐day mortality following elective AAA repair (Figure 11).

The effect of statin pharmacotherapy on AAA‐related outcomes has been the subject of several previous systematic reviews dating back to 2008.²¹ Seven of these review articles conducted meta‐analyses. Two of them evaluated the effect of statin therapy on long‐term survival following AAA repair,²¹ 5 evaluated the effect of statins on AAA growth,²⁶ 1 considered the effects of cardiovascular drugs including statins on AAA rupture,³³ and 1 assessed 30‐day mortality following elective AAA repair.²⁸ Most recently, Takagi et al demonstrated that statin therapy reduced AAA growth by a standardized mean difference of 0.42 mm/y (95% CI 0.19, 0.65, P<0.001) in their 2012 systematic review and meta‐analysis of 11 studies involving 4647 patients.²⁶ The results of our study are consistent with those of this and other meta‐analyses of this outcome. Regarding the effect of statin therapy on AAA rupture rate, the meta‐analysis of individual patient data by Sweeting et al did not demonstrate any statistically significant effect of statins or lipid lowering drugs on AAA rupture rate.³³ However, the authors noted low event rates among the included studies as well as poor reporting of cardiovascular drug use. Twine et al are responsible for the only other assessment of the effect of statin therapy on perioperative mortality following elective AAA repair.²⁸ They assessed 912 patients from 2 studies and demonstrated no beneficial effect of statin therapy on perioperative outcomes following AAA repair (OR 0.22, 95% CI 0.02, 2.90, P=0.06). However, their analysis was limited by the paucity of perioperative mortality data available at the time and was confounded by an inability to separate open and EVAR repairs. Since their study, 4 additional reports including ≈30 times more patients have been published that allowed us to arrive at our results. Furthermore, these studies reported proportions of repair approaches, allowing us to conduct our subgroup analyses.

The exact mechanism by which statins may attenuate AAA growth is unknown, but several theories exist. The prevailing theory is that statins reduce intramural aortic MMP expression.⁵⁹ Overexpression of these collagenases is thought to be an integral pathophysiologic mechanism for AAA development and progression. Culture and tissue studies have demonstrated higher levels of MMPs 1, 2, 8, 9, and 13, and lower levels of their corresponding inhibitors in aneurysmal cells and tissues and at sites of aneurysm rupture.⁶² Furthermore, correlation between degree of inflammation and MMP expression has been shown, and immunochemistry studies have localized MMP production to macrophages and B cells in the adventitia of aneurysmal aortas.⁶⁵ At least in part, statin‐mediated MMP expression attenuation is thought to be derived from the anti‐inflammatory properties of statuns.²⁰

Aneurysm diameter and growth rate are well‐established risk factors for AAA rupture, lending validity to the observed association between statin use and AAA rupture rate.⁶⁸ Furthermore, even though a submillimeter annual growth rate reduction is of questionable clinical significance, the association with a 37% lower rupture rate in the statin arm of our meta‐analysis highlights the potential clinical importance of this effect. This small annual growth rate effect may accumulate over time to produce the observed effect on rupture rate. However, it is difficult to confidently make this assertion due to limited and incomplete reporting of follow‐up in the rupture studies. An alternative explanation for the lower rupture rate among statin users may involve additional AAA stabilization by the attenuation of underlying inflammation and associated tissue friability.

The substantial perioperative mortality rate reduction that we observed in statin users compared with controls is also noteworthy. The effect that we observed in our systematic review might be attributed to the reduction of perioperative coronary events and associated complications as opposed to direct aneurysm‐related causes of mortality. Multiple randomized trials have demonstrated statin‐mediated reductions of cardiovascular events and mortality in primary and secondary coronary disease prevention settings.⁷³ Furthermore, studies have also shown substantial statin‐mediated reductions in mortality, cardiovascular events, and adverse limb events among peripheral arterial disease patients.⁷⁷ Atherosclerosis is a common underlying pathophysiologic phenomenon in coronary, peripheral, and aneurysmal disease, and multiple studies have demonstrated a high prevalence of severe coronary disease and/or peripheral artery disease among patients with AAA.⁷⁸ Consequently, many of these patients already have indications for statin therapy according to existing American Heart Association/American College of Cardiology guidelines.⁸² Furthermore, the most recent Canadian Cardiovascular Society dyslipidemia guidelines consider AAA a statin‐indicated condition irrespective of the presence of other cardiovascular risk factors.⁸⁴ Unfortunately, current evidence suggests that many vascular surgery patients are not adequately risk factor controlled despite knowledge of the common underlying pathophysiological mechanisms for coronary and other vascular diseases.⁸¹ As a result, a substantial missed opportunity may exist for the reduction of mortality following both open and EVAR repair of AAA.

Despite the face validity and consistency of our results with existing literature, our findings should be interpreted in light of several limitations. First, our review consisted only of observational studies. As a result, potential for propagation of information and confounding bias exists. Second, many of the comparisons that we conducted were affected by moderate to significant heterogeneity, likely reflecting the variance in covariate definitions and therefore covariate distributions among studies. The wide variation in baseline aneurysm diameter in the growth and rupture studies is likely to have significantly contributed to this heterogeneity because aneurysm growth demonstrates a progressive exponential increase in growth rate with increasing diameter. Unfortunately, due to incomplete reporting of these covariates and inability to obtain covariate information grouped by our comparison of interest, a thorough assessment of the sources of heterogeneity was not possible. Third, the relative absence of covariate data also made it difficult to precisely assess whether the pooled population represented typical patients with AAA and to what degree statins were indicated in these patients. Similarly, we were unable to account for the effects of different statins, dosages, and durations of treatment. We attempted to do so in our comparison of older and more recent studies, assuming that newer studies included greater proportions of patients on high‐intensity statin regimens in response to emerging evidence and guidelines. The demonstrated potential association between statin use and AAA growth rate reduction among newer studies supported this assumption. In contrast, the presence of a smaller effect of statins on perioperative mortality following AAA repair among newer studies refuted it. Therefore, our findings do not allow us to suggest specific statins, doses, or minimum duration of therapy required to achieve the suggested benefits on each AAA outcome. Finally, our study may not be an accurate assessment of all the available research. We attempted to assess the presence of publication bias for each of our main comparisons, but in many cases, the funnel plots were difficult to interpret. The limited number of studies as well as moderate to high heterogeneity precluded objective publication bias assessment using Egger and other tests and made visual interpretation of the funnel plots difficult.

Despite these limitations, conduct of a large randomized trial to confirm our findings is unlikely, and the present observational data are likely the best assessment possible of the associations between statin therapy and AAA outcomes. Due to the high prevalence of atherosclerotic disease in nonaortic vascular beds in patients with AAA and the resulting existence of indications for statin therapy by current standards, randomization of patients to statin versus placebo groups is fraught with ethical challenges. Furthermore, because statins are now generic, pharmaceutical industry funding is likely to be very difficult to secure for such a study. Finally, significant challenges exist with respect to measurement of AAA‐related outcomes. Growth is commonly measured using abdominal ultrasonography due to its availability, low cost, and absent radiation exposure. However, this modality is affected by technical expertise, body habitus, bowel gas, and a number of other factors that account for the 1.6‐ to 4.4‐mm intraobserver and up to 10‐mm interobserver variability for abdominal ultrasonography of AAA.⁸⁸ Ultimately, all AAA treatments aim to reduce or prevent rupture. However, fewer than 2% of AAAs rupture below 5 cm, substantially increasing the sample size required for a randomized trial to demonstrate an effect if the standard of care is to be followed and aneurysms repaired at 5 cm for women, and 5.5 cm for men.