Summary of main results

Long‐term, daily fluvastatin intake is effective at lowering blood LDL cholesterol concentrations and does so in a predictable dose‐related manner. The 'Summary of findings' table documents that fluvastatin lowers LDL cholesterol by 15% at 10 mg/day and by 33% at 80 mg/day (summary of findings Table for the main comparison). These moderate reductions reflect a reduction in synthesis of cholesterol by the liver and indicate that liver HMG CoA reductase is being inhibited by up to one third over this dose range. This has significant implications beyond circulating LDL cholesterol, as LDL cholesterol is only one of many important biochemical products that are produced by the HMG CoA reductase pathway. Those other products, including co‐enzyme Q10, heme A, vitamin D, steroid hormones and many other compounds, are also likely to be reduced by about one third with the 80 mg dose of fluvastatin. It is important to recognise that the long‐term consequences of reduction of these products is presently unknown.

In the data and analysis section it can be seen that there are more trials and data with the before and after design than from placebo‐controlled trials. For the doses where there is a large number of trials and participants, it can be seen that estimates of the effect of fluvastatin on the lipid parameters are similar with the two different trial designs. This, plus the demonstration that the placebo effect was not different from zero, justified using generic inverse variance to pool and display the combined estimates in Table 1. In addition, all trial data were entered into GraphPad Prism 4 to calculate the regression lines shown in Figure 4; Figure 3 and Figure 5. The overall efficacy results from GraphPad Prism 4 provide the best estimate of the treatment effect, because it is based on a regression line calculated from all the data for all the doses. The estimates of the average treatment effect from the regression lines are similar to the mean value for all the data for each dose (see summary of findings Table for the main comparison).

In this review, it was established using regression analysis that there was a correlation between the baseline value and fluvastatin effect on LDL cholesterol when the effect was expressed as absolute change from baseline (P < 0.0001). There was no correlation between the baseline value and the fluvastatin effect when the effect was expressed as per cent reduction from baseline (P = 0.21). This finding provides strong support for the fact that systematic reviews reporting the effect of statins on absolute changes in lipid parameters are problematic and potentially misleading.

What is the effect of fluvastatin on the end of treatment variability?

End‐of‐treatment variabilities of fluvastatin and placebo were compared to determine the effect of fluvastatin on variability of blood lipids when expressed as a co‐efficient of variation. Compared with placebo, fluvastatin at all doses increased the co‐efficient of variation of blood LDL cholesterol. Fluvastatin did not statistically significantly affect the variability of total cholesterol, HDL‐cholesterol and triglyceride measurements. In order to increase the power to answer this question we identified 66 placebo‐controlled trials from the atorvastatin (Adams 2015), rosuvastatin (Adams 2014) and fluvastatin reviews. In this comparison, the end‐of‐treatment variability expressed as the coefficient of variation for the statin was significantly increased as compared to placebo: total cholesterol (19.5 versus 15.9; P = 0.0005 N = 150) and LDL cholesterol (29.0 versus 23.3; P = 0.0004 N = 171). There was no increase in the end‐of‐treatment variability for the statin compared with placebo for HDL cholesterol (25.28 versus 25.32; P = 0.977 N = 142) and triglycerides (52.8 versus 51.1; P = 0.776 N = 123). The most plausible explanation for the increase in end of treatment variability for total cholesterol and LDL cholesterol with statins is that it reflects some individual variability in response to the statin that would not be present in the people receiving placebo.

Does fluvastatin increase withdrawals due to adverse effects?

Of 36 placebo‐controlled trials, 16 (44%) reported withdrawals due to adverse effects (WDAEs). This analysis represented only 3023 participants, 1759 of whom received fluvastatin and 1264 of whom received placebo. The pooled estimate for all doses provided a risk ratio (RR) of 1.52 (95% CI 0.94 to 2.45), demonstrating uncertainty, but the possibility of an increase in adverse effects even in these short‐term trials. As 20 (56%) of 36 placebo‐controlled trials did not report WDAEs, risk of selective reporting bias for this outcome is high, and the null effect may be a result of that bias. Furthermore, this analysis was limited to trials of three to 12 weeks’ duration and thus does not reflect adverse effects of fluvastatin that occur after intake of longer duration. Risk of participant selection bias is also high in these trials, as many of the participants studied could have been selected because they were known to tolerate statins at baseline.

Overall completeness and applicability of evidence

This review included 145 trials with 18,846 participants. As such it provided us with robust evidence of the dose‐related lipid‐lowering effects of fluvastatin. It was unknown when we did the review whether the time of fluvastatin administration is important with respect to lipid lowering. Only one trial (Scharnagl 2006) compared morning and evening administration and did not show a difference. A sensitivity analysis comparing twice‐daily versus single‐dose regimen data were available for the doses 20 mg/day, 40 mg/day and 80 mg/day. The percentage reductions in twice‐daily versus single‐dose regimens showed no difference. We therefore felt justified in combining data from both dosing regimens. Recently a Cochrane review has attempted to answer this question and concluded that statin lipid‐lowering effect is the same for morning and evening administration (Izquiero‐Palomares 2016).

Practitioners can use this evidence to calculate the expected effect of doses of fluvastatin commonly utilised in society. It is unlikely that further research will change these estimates appreciably. However, there was a fair amount of heterogeneity in many of the estimates and it is possible that this was due to differences in the populations being studied (e.g. gender or genetic differences) (Thompson 2005). To explore this, where it was possible, we compared the effect of fluvastatin in males and females plus in patients with familial and non‐familial hypercholesterolaemia. A subgroup analysis comparing male versus female participant data was available for the doses 5 mg/day, 20 mg/day and 40 mg/day and no difference was proven. However, we judged the amount of data available were insufficient to answer whether the lipid‐lowering effect of fluvastatin differed in males and females. If anything, it would be anticipated that the effect would be greater in females because on average they weigh less than males. It is important for authors to report data separately by sex and if this had been done in all these trials, we likely would have been able to answer this important question. The results of this subgroup analysis for both atorvastatin and rosuvastatin suggested a larger effect in females than males: atorvastatin 10 mg/day (Adams 2015) male versus female ‐39.2 (95% CI ‐41.6 to ‐36.9) versus ‐41.8 (95% CI ‐43.4 to ‐40.2) P = 0.08 and rosuvastatin 10 mg/day (Adams 2014) male versus female ‐45.1 (95% CI ‐47.9 to ‐42.2) versus ‐49.4 (95% CI ‐51.7 to ‐47.2) P = 0.02.

Familial versus non‐familial hypercholesterolaemia participant data were available for the fluvastatin doses 5 mg/day, 20 mg/day, 30 mg/day and 40 mg/day. These data were analysed separately for LDL cholesterol‐lowering efficacy using the generic inverse variance fixed‐effect model in RevMan 5. The percentage reduction in familial patients was less than non‐familial for all doses except 40 mg/day (see results). These findings of a lesser affect in familial hypercholesterinaemic participants is consistent with what was found for atorvastatin (Adams 2015): atorvastatin 10 mg/day ‐34.7 (95% CI ‐36.6 to ‐32.8) versus ‐36.3 (95% CI ‐36.7 to ‐35.8) P = 0.12 and 20 mg/day ‐38.0 (95% CI ‐39.8 to ‐36.2) versus ‐43.6 (95% CI ‐44.4 to ‐42.8) P < 0.00001.

The profound and relatively consistent effect of fluvastatin on lipid parameters shown in this review is probably appreciated by clinicians who treat patients with these drugs. The ability to know whether a patient is taking a statin or not is also most likely evident to investigators involved in statin placebo‐controlled randomised controlled trials (RCTs). Knowledge of the lipid parameters almost certainly leads to loss of blinding in statin RCTs. The present review calls attention to that problem and efforts to prevent this loss of blinding are needed in future statin RCTs (Higgins 2011).

We have used data from the Cholesterol Treatment Trialists’ (CTT) publications to determine the effects of fluvastatin, atorvastatin and rosuvastatin on LDL cholesterol lowering and reduction of myocardial infarction. In two RCTs a mean fluvastatin dose of 72 mg/day reduced LDL cholesterol by 31.9%, and reduced myocardial infarction, relative risk, 0.68 (95% CI 0.55 to 0.85) as compared to placebo. In five RCTs a mean atorvastatin dose of 26 mg/day reduced LDL cholesterol by 44.0% and reduced myocardial infarction, relative risk, 0.67 (95% CI 0.58 to 0.77) as compared to placebo. In four RCTs a mean rosuvastatin dose of 16 mg/day reduced LDL cholesterol by 48.8% and reduced myocardial infarction, relative risk, 0.82 (95% CI 0.73 to 0.93) as compared to placebo. Thus despite reducing LDL cholesterol by a much lesser amount with fluvastatin than atorvastatin and rosuvastatin, fluvastatin reduced myocardial infarction similarly to atorvastatin and to a greater degree than rosuvastatin. Fluvastatin 72 mg is equivalent to about 6 mg of atorvastatin and about 1.6 mg of rosuvastatin in LDL cholesterol lowering. These findings call into question the commonly held belief that the effect of statins to reduce myocardial infarction is solely due to lipid lowering. It certainly suggests that statins could be acting by some other mechanism to reduce myocardial infarction and calls for more head‐to‐head RCTs comparing different statins.

Quality of the evidence

The summary of all ’Risk of bias’ parameters for the lipid effects suggests a high risk of bias (Figure 6). However, the lipid parameter outcomes are probably relatively resistant to bias. If anything, a high risk of bias would lead to an overestimate of the lipid‐lowering effects rather than an underestimate. However, because of the objectivity of the lipid measurements we think that the estimates of effects are reasonably accurate. This view is strengthened by the fact that the two different trial designs, placebo‐controlled RCTs and before and after design produced similar results. Furthermore, we could not show evidence of funding bias. Comparing Novartis‐funded trials where an overestimate of the effect might be expected and non‐Novartis‐funded trials where a bias towards underestimating the effect of fluvastatin may be expected did not show any difference in the effect of fluvastatin on lipid parameters. Furthermore, review of funnel plots did not suggest evidence of publication bias.

Low risk of bias is not true for the harm outcome, withdrawals due to adverse effects (WDAE). This was reported in 16 (44.4%) of the 36 placebo‐controlled trials. There is therefore a high risk of selective reporting bias for this outcome and this combined with the high risk of other biases means that we cannot be confident that not finding a significant increase in WDAEs is correct (summary of findings Table 4).

Potential biases in the review process

Combining the placebo‐controlled trials with the before and after trials is a limitation of the review. We have explained why the increased risk of bias associated with the before and after design is less in this instance because the lipid parameters were measured in a remote laboratory. Another limitation of this review is that many trials did not report standard deviations for the lipid‐lowering effects. Where possible these values were determined by the method of (Furukawa 2006), from t‐statistics corresponding to the exact P values reported or from the 95% CI of the mean difference between treatment groups. In trials where the standard deviation was not reported and could not be calculated, the standard deviations were imputed as the average of this parameter from trials that reported it. Such imputation might weight some studies more or less; however, this has been shown in other reviews not to have much effect on the estimate of the effect size (Heran 2008; Musini 2014). Another limitation is that few studies were available to demonstrate the lipid‐lowering effect of fluvastatin at very low and very high doses. We did not downgrade the quality of evidence due to heterogeneity of LDL cholesterol because the confidence intervals for the pooled result estimates were narrow.

Agreements and disagreements with other studies or reviews

The best estimate of the mean per cent reduction in blood LDL‐cholesterol for any dose of fluvastatin can be calculated from the log dose‐response equation. Using this equation y = ‐19.73 log(x) + 4.869, a fluvastatin dose of 40 mg/day reduces LDL cholesterol by an average of 26.7%. This is close to the range of 22.0% to 26.0% reduction in LDL cholesterol from the six comparative trials from the Drug Effectiveness Review Project (DERP) (Smith 2009) and a range of 24.8% to 29.4% reduction in LDL cholesterol in 23 placebo‐controlled trials from (Law 2003).

Comparison of the effect with other statins

The greatest value in doing this type of review is the ability to compare fluvastatin to other statins. At present we can compare fluvastatin to atorvastatin and rosuvastatin, which have been reviewed using the same protocol. The most important finding in this review is that the slope of the dose response effect for fluvastatin on LDL, total cholesterol and triglycerides is not different from the slopes of the dose response curve for atorvastatin (Adams 2015) and rosuvastatin (Adams 2014). This provides some confirmation that the three statins are all causing lipid lowering by a similar mechanism. However, it also demonstrates that fluvastatin is much less potent than the other two drugs: fluvastatin is 12‐fold less potent than atorvastatin in lowering LDL cholesterol and 46‐fold less potent than rosuvastatin. This means that fluvastatin 80 mg/day reduces LDL cholesterol on average by 32.7 %; the dose of atorvastatin and rosuvastatin to achieve the same reduction in LDL cholesterol is 7 mg/day and 2 mg/day, respectively.