Summary of main results

We analysed nine randomised placebo‐controlled studies in children with heterozygous familial hypercholesterolaemia (FH). The studies showed a clinically significant reduction in both serum total cholesterol and low‐density lipoprotein (LDL) cholesterol among children treated with a statin compared with those treated with a placebo. In addition, statin therapy slightly increased serum high‐density lipoprotein (HDL) cholesterol and slightly decreased serum triglyceride concentration; however, when compared with the substantial change in serum LDL cholesterol, these changes are likely to be of minor importance. The magnitude of LDL cholesterol lowering varied from study to study, most likely due to different statins and doses and possibly due to different definitions about true monogenic heterozygous FH. We did not do any formal subgroup analyses because the choice of statin treatment was heterogeneous between studies.

Endothelial dysfunction represents one of the earliest stages of atherogenesis, and has a clear predictive value for future cardiovascular disease. A number of studies have shown that endothelial function measured as flow‐mediated dilation is impaired in children with FH (de Jongh 2002b; Vlahos 2014). The effect of simvastatin on flow‐mediated dilatation of the brachial artery in children with FH was reported in one study (de Jongh 2002a). It was found that simvastatin therapy restored endothelial function in the studied participants (50 children with FH; 9 to 18 years). Clearly more studies are needed to confirm this result in children with FH.

In addition to early changes in the function of the arterial endothelium in children with FH, which result from the high LDL cholesterol concentration in the blood, accumulation of the LDL cholesterol in the subendothelial space of the carotid arterial wall leads to increased intima‐media thickness (IMT) of the carotid arteries (Tonstad 1996). Carotid IMT represents the combined intima and media thickness of the arterial wall, and numerous studies have shown that this surrogate marker of atherosclerotic vessel wall change is a reliable indicator of clinical outcomes later in life (Koeijvoets 2005). Accordingly, studies examining the sensitivity of this surrogate marker to risk intervention are important. We found only one study that used carotid IMT as the primary efficacy outcome in children with FH treated with statins (Wiegman 2004). The authors found that two years of pravastatin therapy induced a small but significant regression in mean change in IMT between statin‐treated and placebo groups in children with FH. This clearly encouraging result calls for further studies with pravastatin or other statins.

In the largest study, all children were from families where a molecular diagnosis had been made in one parent and where the recruited child had LDL cholesterol twice greater than 4.0 mmol/L. The authors judged this to mean the child had a greater than 99.6% chance of having inherited the family mutation (Wiegman 2004).This sample is therefore highly likely to consist of all FH individuals. In the most recent study (Braaskamp 2015a), all children had a mutation in LDLR or APOB genes or had a parent where the mutation had been identified, and thus all of these children have molecularly defined monogenic FH. In all the other studies, children were recruited as having LDL cholesterol above a cut‐off point which varied between the studies (Starr 2008). The other criterion was having a first degree relative, either with elevated LDL cholesterol, or with a family history of premature coronary artery disease. It is therefore likely the vast majority of the children in the studies included in this review have monogenic FH, but it cannot be ruled out that a small percentage (not more than 10%) may not have.

Although in the majority of studies published in recent years molecular testing was performed and only mutation‐positive participants recruited, the diagnosis of FH was not confirmed by direct molecular testing of the children in some of the earlier studies included in this review. Although idiopathic elevated LDL cholesterol levels occur less frequently in children than in adults, this raises the question of whether only a proportion of the children in the earlier studies have true monogenic FH. This may affect the conclusions made above, since the lipid‐lowering response to a statin may be different in mutation carriers compared to those with a polygenic cause of their phenotype. In adults with a clinical diagnosis of “Definite” FH a causative mutation can be found in between 70% to 80% of individuals, while only around 30% of people with a clinical diagnosis of “Possible” FH carry a causative mutation (Graham 2005; Humphries 2006b; Futema 2013). It is now known that in people with a clinical diagnosis of FH but with no detectable mutation in any of the three common FH genes there is a polygenic (not a monogenic) cause of their phenotype (Talmud 2013), and they have been incorrectly been given the diagnosis of FH. This polygenic cause has also been demonstrated to explain the elevated LDL‐C levels in children with a diagnosis of FH where no mutation can be found (Futema 2015).

In the absence of molecular confirmation, it is possible to estimate the probable dilution of monogenic FH children with children with a polygenic aetiology from family studies. There is a considerable overlap in LDL‐cholesterol levels in the mutation‐carrying and non‐mutation carrying siblings of a parent with FH, such that using the intersection between the two peaks of LDL‐cholesterol levels observed results in a false positive diagnostic rate of 6% to 8% (Kwiterovich 1974; Leonard 1976). Thus selecting children as “FH” based only on having a parent with FH and elevated LDL‐C levels may have resulted in the inclusion of 6% to 8% non‐mutation carriers, in earlier studies where no DNA testing was carried out. Based on this we can conclude that the earlier published estimates of the effect of statin treatment in children with a clinical diagnosis of FH where no molecular testing had been performed are unlikely to have been significantly influenced by the incorrect inclusion of non‐monogenic individuals.

The importance of distinguishing between monogenic and polygenic elevation of LDL cholesterol is whether children with a monogenic cause might have a much smaller than average LDL cholesterol‐lowering response than children whose hypercholesterolaemia is due to polygenic causes. Although we are not aware of any data addressing this directly in children, there is evidence that adults with a clinical diagnosis of FH without a detected mutation have a better response to statins than those in which a mutation has been found (Sun 1998; Heath 1999). Another issue to consider is that children with different LDLR mutations, or in those where FH is caused by mutations in the APOB or PCSK9 genes, may respond differently to statins. There is no direct evidence for this in children, but adults carrying the APOB mutation have been reported to respond better to statins than those carrying an LDLR mutation (Myant 1993) or a PCSK9 mutation (Humphries 2006b). Furthermore, the class of LDLR mutation can affect the untreated LDL cholesterol levels (Humphries 2006b; Futema 2013) and affect the response to statins (Couture 1998; Vohl 2002; Miltiadous 2005). These variations according to mutations are pertinent because founder effects are seen in many countries, e.g. South Africa (Kotze 1993), Finland (Vuorio 2001) and Holland, (Umans 2002). Thus the variability in the prevalence of different mutations and molecular causes of FH across countries may contribute to a small extent to between‐study differences in response but although there may be a small overestimate of the response in statin‐sensitive mutation‐carrying FH children, it is unlikely to be more than 5% (van der Graaf 2011).

Recent guidelines vary in their recommendations as to when statin treatment should be started between 8 years to 14 years (McCrindle 2007; SIGN 2007; Daniels 2008; NICE 2008; Descamps 2011; Goldberg 2011; Sullivan 2012). None of the guidelines recommended statins before the age of eight years in cases of heterozygous FH (Vuorio 2013). In regard to dosing, this varied considerably between the studies. In the earliest study, children with FH were treated with pravastatin doses from 5 mg/day to 20 mg/day (Knipscheer 1996). In the later studies there was a tendency to use larger doses. Wiegman used pravastatin doses of 20 mg/day or 40 mg/day (Wiegman 2004), and de Jongh titrated simvastatin doses of up to 40 mg/day (equivalent dose of pravastatin, 80 mg/day) (de Jongh 2002a). McCrindle titrated atorvastatin doses from 10 mg/day up to 20 mg/day (equivalent dose of pravastatin, up to 80 mg/day) if the LDL cholesterol level remained over 3.4 mmol/L (McCrindle 2003). In this study the mean serum LDL cholesterol concentration among FH children treated with statins was 3.39 mmol/L at the end of follow‐up. In the most recent study (Braaskamp 2015a), the 5 mg starting dose of rosuvastatin was titrated at 3‐monthly intervals to a maximum tolerated dose of 10 mg (six‐ to nine‐year olds) or 20 mg (10‐ to 17‐year olds) to achieve an LDL‐C goal of (2.85 mmol/L (110 mg/dL). In all cases the minimal effect dose is advised.

It has been estimated that elevations in aminotransferase levels over three times the upper limit of normal occur in less than 1% of adults on any statin (Cohen 2006). However, there was no increase in aminotransferase levels when compared with a placebo group in a recent meta‐analysis (de Denus 2004), and adults with elevated aminotransferase levels during statin treatment do not appear to have a higher risk of liver dysfunction (Chalasani 2004). The risk of acute severe liver dysfunction in the general population with no statin medication is about one to two cases per million (Law 2006). Consequently, severe liver dysfunction is extremely rare and routine monitoring is recommended, but it will be effective only when it is active and includes not only laboratory test but also clinical follow‐up (Golomb 2013). The studies of children with FH used liver transaminases as the method for detecting possible liver dysfunction. The putative risk of statin‐induced severe acute liver dysfunction at this stage of life should be outweighed by the reduced cardiovascular risk achieved by statin treatment (NICE 2008). The most recent guidelines by the National Lipid Association's Statin Safety Assessment Task Force give some useful considerations (McKenney 2006). They underline the importance of monitoring any possible symptoms like abdominal pain related to liver dysfunction and advise to consider using a fractionated bilirubin for detection of liver dysfunction. This kind of monitoring was carried out systematically in only half of the studies we analyzed, and the monitoring protocol varied between the studies (Stein 1999; de Jongh 2002a; Wiegman 2004; Clauss 2005). None of the studies included fractionated bilirubin in their laboratory analysis. It can be concluded that even though liver dysfunction was not present in the included studies, the risk exists. Therefore, any new studies of children with FH should be planned so that possible hepatotoxicity symptoms are routinely monitored using standardized methods and, additionally, new laboratory standards should be used in detecting possible liver dysfunction.

The incidence of rhabdomyolysis has been estimated to be about 3.4 per 100,000 person‐years in adults (Law 2006). Although this figure is very low, the lesson learned with cerivastatin should be keenly kept in mind (Pasternak 2002). The rate of fatal rhabdomyolysis with this drug was unexpectedly and exceptionally high; 16 to 80 times greater than with other statins (Staffa 2002) and even after excluding individuals treated simultaneously with gemfibrozil, the rate of fatal rhabdomyolysis was still 10 to 50 times higher than of other statins (Staffa 2002). All statins used so far in children with FH (atorvastatin, lovastatin, pravastatin, and simvastatin) appear to have a low risk of rhabdomyolysis as compared to adults, which is estimated to be about 0.08% to 0.09% of persons treated with these statins (Pasternak 2002). Since the mechanism of myopathy is not well understood, it is of the utmost importance to monitor adverse reactions and adjust the therapy accordingly (Pasternak 2002).

The terminology of clinical adverse events in the analyzed studies varied. The comparison between the studies would have been more reliable if the definition of adverse events had been standardized. It is important to note that some drugs that interact with statins (macrolide antibiotics and azole fungals) may have been consumed, thus altering the adverse event risk. In practice any interaction risk can be mitigated with patient education. Statin therapy combined with alcohol abuse potentially increases the risk of liver dysfunction. Alcohol consumption was not monitored in any of the included studies; however, alcohol abuse is uncommon in children but when adolescence is reached discussions about alcohol consumption should be had with the young person. It is unclear as to whether statins increase the risk of cataracts with about equal numbers of studies supporting the theory versus those against (Harris 1995; Cenedella 1996; Pedersen 1996; Chodick 2010; Hippisley‐Cox 2010; Fong 2012; Leuschen 2013).

One of the potential long‐term side effects of statin treatment in children with FH is the increased risk of developing type 2 diabetes (T2D) that has been noted in statin treatment of non‐FH individuals. A meta‐analysis of published randomised controlled trials in over 91,000 high risk individuals from the general population (Sattar 2010) reported statin therapy was associated with a 9% increase in the likelihood of new T2D during follow‐up. Interestingly, a second meta‐analysis showed pravastatin (40 mg/day) was associated with the lowest (7%), atorvastatin (80 mg/day) with an intermediate (15%) and rosuvastatin (20 mg/day) and simvastatin (40 mg/day) with the highest (25% and 21% respectively) risk of new onset T2D (Navarese 2013). The exact molecular mechanism of this statin‐associated T2D risk is unknown (please refer to the references in Vuorio 2016), and it is unclear whether this is an on‐target or off‐target effect of the drug; that is, whether the dysglycaemic effect is a direct consequence of inhibition of HMG‐CoA, the intended target of statins. Using the approach of Mendelian Randomisation, variants in the gene encoding HMG‐CoA reductase (HMGCR, chr 5q13.3) associated with lower LDL‐C were used as proxies for statin treatment. Both statin treatment and the genetic variants were associated with higher T2D risk and higher bodyweight, and the genetic variants with higher plasma glucose and insulin, and waist and hip circumferences (Swerdlow 2015). This directional concordance strongly suggested that the higher T2D risk caused by statin therapy is at least in part a direct consequence of HMG‐CoA reductase inhibition. Reassuringly, many studies have reported that the prevalence of T2D is low in adults with FH, and in a study of over 63,000 people from Holland (Besseling 2015), even in treated adults with FH the prevalence of T2D was significantly lower than in their unaffected relatives (1.75% versus 2.93%). Follow‐up studies in adults (Skoumas 2014) and in children (Kusters 2014) are also reassuring, with 10‐year follow‐up in 194 statin‐treated children (mean age at baseline 13 years) seeing one new case of T2D, with a similar incidence in their 83 non‐FH siblings (Kusters 2014). It is clear that overall the benefits of statin treatment for preventing cardiovascular disease in people with FH far outweighs the modest potential risk of T2D. It is possible the dietary and lifestyle advice given to all people with FH encourages them to maintain an ideal body weight and thus to ameliorate any of the statin‐associated risk of developing T2D. Based on published evidence it would appear that treatment with pravastatin is associated with the lowest risk, although long‐term follow‐up studies of treated FH children are needed to confirm this.

Relatively little is known about the potential statin‐related neurologic side‐effects such as sleep disturbances (Bays 2006), effects on cognitive function and peripheral neuropathy (Backes 2003; Chong 2004). In most cases, the onset of symptoms was reported within six months of commencing medication, and most of the peripheral neuropathies were confirmed by nerve conduction studies. The reports were related to all statins, and discontinuation of statin treatment improved the symptoms. In a recent review based on adult studies, peripheral neuropathy was concluded to be an idiosyncratic effect of statin use (Brass 2006). In the included studies Wiegman showed that there was no difference on academic performance between the statin treated and placebo group (Wiegman 2004).

Particularly important are the concerns related to any potential impact on sexual and physical maturation (McCrindle 2007). Long‐term effects on maturation will need to be studied in longer and larger controlled follow‐up studies. Physical maturation was followed in some studies by measuring height but this is unreliable in short‐term studies. There is concern about pregnancy during statin treatment (McCrindle 2007) as statins can affect foetal development. Females of child‐bearing age should receive counselling and contraceptive advice (Arambepola 2007; McCrindle 2007; NICE 2008; Nordestgaard 2013).

In summary, this review found statins to be an effective treatment for FH in children. It did not find any difference between the statin and control group in the proportion of participants who experienced a clinically significant increase in liver transaminase values (over three‐fold increase in alanine transferase or aspartate aminotransferase) or creatine kinase values (over 10‐fold increase). We did not find any significant difference between those treated with a statin and those treated with placebo with respect to their sexual maturation measured by the Tanner staging. Overall, the data suggest the risk of adverse events in children treated with statins is similar to that observed in statin‐treated adults over the short term and the adverse event rate was the same between statin and placebo group. In the absence of long‐term treatment and follow‐up of children, it is not possible to rule out any long‐term adverse effects. Our findings are similar to those reported in two systematic reviews (Arambepola 2007; Avis 2007).

Overall completeness and applicability of evidence

The quality of the data concerning change in serum LDL cholesterol and adverse events was, according to GRADE, classified as moderate. These data are applicable in the treatment of FH children. In conclusion, statins lowered LDL cholesterol effectively (advantage). Significant adverse events were not present during the statin treatment (potential harm).

Quality of the evidence

Blinding (performance bias and detection bias) was not present in any studies. Bias from random sequence generation (selection bias) was not present in two studies and the data were not clearly stated in seven studies. Bias from allocation concealment (selection bias) and selective reporting (reporting bias) can not totally be excluded, but it is very unlikely. In conclusion it can be stated that all the studies appeared to be well run and we do not think any above mentioned factors will influence the results in a negative way. Quality of evidence varied from moderate (change in serum LDL cholesterol and adverse events) to low (change in carotid intima‐media thickness, change in measures if growth and maturation, liver dysfunction, myopathy and change in endothelial function).

Potential biases in the review process

A comprehensive literature search was carefully carried out and we consider that most controlled studies were identified. Study protocols varied between the included studies and data were presented slightly differently between the studies. In conclusion we found no potential bias in the review process.

Agreements and disagreements with other studies or reviews

Our findings are similar to those reported in two systematic reviews (Arambepola 2007; Avis 2007). In the Arabepola review, results of a parallel‐group randomised placebo‐controlled trial concerning heterozygous FH children with LDL and HDL cholesterol and triglycerides as outcomes were pooled using standard meta‐analytical methods (Arambepola 2007). In the Avis review, they performed a meta‐analysis of randomized, double‐blind, placebo‐controlled trials evaluating statin therapy in children aged 8 to 18 years with heterozygous FH and six studies (n = 798 children) with 12 to 104 weeks of treatment were included (Avis 2010).