The impact of statins on psychological wellbeing: a systematic review and meta-analysis

Articles included in the review were RCTs meeting the following inclusion criteria: (1) ≥1 placebo/control condition; (2) random assignment to treatment condition of a statin agent (simvastatin, atorvastatin, fluvastatin, lovastatin, mevastatin, pitavastatin, pravastatin, rosuvastatin, cerivastatin) alone/in conjunction with a dietary program; (3) documentation of psychological outcomes (depression, distress, mood) as a primary/secondary endpoint; (4) minimum 2 weeks of treatment (period over which change in mood/depressed state is monitored); and (5) a defined study sample detailing medical status (risk of CVD/hypercholesterolemia, hypercholesterolemia, history of CVD, healthy). Two independent searches that applied this strategy were conducted by the leading authors (AO'N and LS). Results of the searches were then compared and inter-rater reliability assessed. Where consensus for the inclusion of a study was not reached, the third author was consulted.

To avoid coder bias, data from included articles were extracted by a single investigator and collated into a table for review by the author group. The extraction table included study author, title, journal, drug type and dosage, population, design, assessment points, psychological measure and key findings. Trial quality was assessed according to an inventory guided by Cochrane Collaboration recommendations on evaluating validity. These criteria included: randomization; allocation concealment; blinding; completeness of follow-up; and intention-to-treat analysis. Due to the different measures used to assess depression and mood across studies, estimates were calculated using standardized mean differences (SMD). A random effects model was employed to account for differences in trial designs (for example, crossover, parallel) [16]. Where standard deviations were not reported in the publication (CIs or standard error were instead presented), standard equations for conversion were used to calculate SMD [17]. Where medians and interquartile ranges (IQRs) were reported, medians were substituted for means as recommended by Hozo et al., who suggest that when sample size exceeds 25, the median is the best estimator [18]. IQRs were transformed using methods derived from Hozo et al. Where a study [19] presented only a between-group change score, normative data for non-psychiatric adult populations were obtained for the depression instrument used in that study (Brief Symptom Inventory - Depression and Anxiety subscale) in the absence of baseline data [20]. In cases where approximations were provided, sensitivity analyses were conducted to determine their influence on overall effect.

Where more than one instrument was used to measure mood state, the instrument most comparable with those used in other included trials was selected. Similarly, where more than one medication arm was included in a trial, the statin agent deemed by the author group to be most comparable with the others included in the review was selected for inclusion in the primary meta-analysis. We re-ran our analysis, imputing data from the second statin arm in each of the multi-arm studies, and it had no influence on results (data not shown). To address heterogeneity among characteristics of study samples (hypercholesterolemia, no hypercholesterolemia), sensitivity analyses were conducted.

The primary meta-analysis measured the effects of statin use on psychological wellbeing (depression and mood) of individuals with or without hypercholesterolemia compared with a placebo, with follow-up periods ranging from 4 weeks to 4 years. Meta-analyses were performed using SMD for continuous outcome measures (based on post-post means), with 95% CIs, using Review-Manager 5 software (RevMan Computer Program. Version 5.1. Copenhagen: The Nordic Cochrane Centre. The Cochrane Collaboration, 2011). We applied Hedges' adjusted g to adjust for small sample bias. We explored heterogeneity by calculating the I2 statistic and where high I2 values were observed, sensitivity analyses was conducted to exclude studies with potential bias or characteristics that were not comparable. A priori subgroup analysis estimated the effects of statins on psychological outcomes by length of treatment period, type of statin medication, and psychological outcome (for the purpose of this paper and these analyses, the authors made a distinction between mood and depression in the latter subanalysis for comparability purposes).

Key characteristics are displayed in Table 1. The trials included in the review were published from 1994 to 2006. Studies were conducted in the UK (n = 3) [23‐25], Scandinavia (n = 1) [19], Australia/New Zealand (n = 1) [26], and the US (n = 2) [27, 28]. All studies defined participants in their samples as those with hypercholesterolemia or elevated serum cholesterol levels with the exception of two, which comprised those with past coronary disease (n = 1) [26], or medically healthy individuals (n = 1) [24].

Five studies reported depression outcomes at follow-up using Hamilton Depression Rating Scale (clinician administered) [28], Brief Symptom Inventory (self-report) [19], Hospital Anxiety and Depression Scale (HADS) (self-report) [24], General Health Questionnaire - Depression and Anxiety (self-report) [26] and Center for Epidemiology Studies Depression Scale (self-report) [27]. The remaining studies reported mood outcomes using versions of the Profiles of Mood States [23, 25]. Four studies measured anxiety using the General Health Questionnaire [26], HADS [24], Brief Symptom Inventory [19], and the Profiles of Mood States [23].

Details of treatment length and dosage are presented in Table 1. Lovastatin (n = 2) [25, 28], simvastatin (n = 4) [19, 23, 24, 27] and pravastatin (n = 3) [24‐26] were the statin agents evaluated in the seven studies. Two studies included a three-arm design evaluating two statin arms and one placebo arm [24, 25]. The remaining five studies comprised a two-arm design (statin versus placebo). Upon consultation with the author group, the type of statin chosen for inclusion in the main effects model for these two studies were those most commonly evaluated in the other studies.

We reviewed included studies using a set of quality criteria developed by the Cochrane Centre. Based on this, we concluded that the included studies were of sound methodological quality. Where participant drop out occurred, all studies reported a greater than approximately 80% retention rate. All studies used intention to treat analysis. Details of how investigators and participants were blinded to treatment condition and the process of randomization were unclear in most studies. The key elements of trial quality for each study are presented in Table 2.

The 7 RCTs included in the analysis represented 2,105 participants (1,133 treatment group versus 972 placebo group). A test for overall effect demonstrated no significant differences in psychological outcomes between participants receiving statins (simvastatin, lovastatin, provastatin) or a placebo (SMD = -0.08, 95% CI -0.29 to 0.12; P = 0.42) (Figure 1). Given the high heterogeneity observed across studies (I2 = 71%), sensitivity analyses were conducted to separately analyze depression (n = 5) and mood (n = 2) outcomes. This analysis yielded low heterogeneity (I2 = 0%) and revealed that statins were associated with significant improvements in mood scores (SMD = -0.43, 95% CI -0.61 to -0.24) (Figure 2), as measured by the Profile of Mood States.

When we compared long-term (3 to 4 years), intermediate (3 to 6 months) and short-term (4 weeks) effects of statin use, no statistically significant differences were observed between groups. When results were explored separately for lovastatin, simvastatin and provastatin, simvastatin was the only agent associated with psychological improvements (SMD = -0.11, 95% CI -0.41 to 0.19), although these effects were not significant.

Further sensitivity analyses were conducted to exclude studies comprising participants with a CVD history [26] or 'healthy' volunteers [24]; this had no impact on results. In a separate analysis, we also excluded studies from which we applied approximations [19, 23, 24]. The effect of statins on psychological outcomes remained non-significant (SMD = -0.02, 95% CI -0.25 to 0.20).

To explore the effects of statins on anxiety, we included available data from studies that measured anxiety, as distinct from depression (n = 3: one presented data in graphical form [26], therefore were not included in our analysis). This subanalysis compared anxiety data of 419 treatment participants versus 222 control participants, revealing no significant between-group differences (SMD = -0.01, 95% CI -0.17 to 0.16).