La consommation des fruits à coque pour la prévention primaire des maladies cardiovasculaires
Résumé scientifique
Contexte
Les fruits à coque contiennent un certain nombre d'attributs nutritionnels qui peuvent s’avérer être cardioprotecteurs. Un certain nombre d'études épidémiologiques ont montré que la consommation de fruits à coque pouvait avoir un effet bénéfique chez les personnes présentant des facteurs de risque de maladies cardiovasculaires (MCV). Toutefois, les résultats des essais contrôlés randomisés (ECR) sont moins cohérents.
Objectifs
Déterminer l'efficacité de la consommation des fruits à coque dans la prévention primaire des MCV.
Stratégie de recherche documentaire
Nous avons effectué des recherches dans les bases de données électroniques suivantes : le Registre Cochrane des essais contrôlés (CENTRAL), MEDLINE, EMBASE, Web of Science Core Collection, CINAHL, Database of Abstracts of Reviews of Effects (DARE), Health Technology Assessment Database (HTA) et Health Economics Evaluations Database (HEED) jusqu'au 30 juillet 2015. Nous avons aussi fait des recherches dans les registres d'essais et les références de revues pertinentes. Aucune restriction de langue n’a été imposée.
Critères de sélection
Nous avons inclus des ECR sur les conseils diététiques ou la provision de fruits à coques pendant au moins trois mois dans le but d’augmenter leur consommation chez des adultes en bonne santé ou des adultes à risque modéré ou élevé de MCV. Le groupe témoin n'a fait l'objet d'aucune intervention ou d'une intervention minimale. Nous nous sommes intéressés aux résultats portant sur les événements cliniques et les facteurs de risque de MCV.
Recueil et analyse des données
Deux auteurs ont indépendamment sélectionné les essais à inclure, ont résumé les données, et ont évalué les risques de biais des essais inclus.
Résultats principaux
Nous avons inclus cinq essais (435 participants randomisés) et un essai en cours. Une étude est en attente de classification. Tous les essais ont porté sur la distribution de fruits à coque visant à augmenter leur consommation plutôt que sur les conseils diététiques. Aucun des essais inclus n'a examiné les manifestations cliniques de MCV (critères de jugement principaux), mais les essais étaient de petite envergure et de courte durée. Par contre, les cinq essais ont examiné l’effet des fruits à coque sur les facteurs de risque de MCV. Quatre essais ont fourni des données dans un format utilisable pour les méta‐analyses, mais l'hétérogénéité entre les essais n’a pas permis de conduire de méta‐analyses pour la plupart des données. Dans l'ensemble, les essais ont été jugés comme présentant un risque peu clair de biais.
Les effets de la consommation de fruits à coque sur les facteurs de risque de MCV (taux de lipides et tension artérielle) étaient variables et peu uniformes. Trois essais ont surveillé l’apparition d’évènements indésirables. Un essai a fait état d'une réaction allergique et trois essais n'ont rapporté aucun gain de poids significatif avec une consommation accrue de fruits à coque. Aucun des essais inclus n'a rapporté d’autres critères de jugement tels que la survenue du diabète de type 2 comme facteur de risque majeur de MCV, la qualité de vie liée à la santé, et les dépenses de santé.
Conclusions des auteurs
À l'heure actuelle, nous manquons de données concernant les effets de la consommation de fruits à coque sur les événements cardiovasculaires dans la prévention primaire des MCV et disposons de très peu de données concernant leurs effets sur les facteurs de risque de MCV. Aucune conclusion ne peut être tirée et des essais de meilleure qualité, de plus longue durée, et incluant plus de participants, sont nécessaires pour répondre à la question posée dans cette revue.
PICO
Résumé simplifié
Consommer des fruits à coque pour prévenir les maladies cardiovasculaires
Problématique de la revue
Cette revue Cochrane a pour objectif d’évaluer si la consommation de fruits à coque peut jouer un rôle dans la prévention les maladies cardiovasculaires.
Contexte
Les maladies cardiovasculaires sont un groupe de maladies qui affectent le cœur et les vaisseaux sanguins. Elles sont une cause majeure de décès dans le monde entier. Notre alimentation peut influencer le risque de développer une maladie cardiovasculaire. Les fruits à coques, s’ils sont consommés régulièrement et à des doses relativement élevées (50 g à 100 g), peuvent théoriquement réduire le cholestérol total et le cholestérol des lipoprotéines de basse densité (LDL) (''mauvais cholestérol'').
Caractéristiques de l'étude
Cette revue comprend des essais contrôlés randomisés d’une durée minimale de 12 semaines. La plupart des participants avaient entre 37 et 54 ans. Les données ont été collectées jusqu'au 30 juillet 2015.
Principaux résultats
Nous avons inclus cinq essais (435 participants), dont l'un comportait deux groupes de traitement. Les cinq essais ont porté sur les effets de la consommation de fruits à coque. Aucune étude n'a été conduite sur l'effet des conseils diététiques ayant pour but d’augmenter la consommation de fruits à coque. Aucune des études n'a examiné les effets sur les événements cardiovasculaires ou les décès. Aucun des résultats ne montre un effet manifeste des fruits à coque sur le taux de cholestérol total et la tension artérielle. Une étude a signalé un cas de réaction allergique. Trois études n'ont fait état d'aucun gain de poids significatif avec l'augmentation de la consommation de fruits à coque. Aucun autre événement indésirable n'a été signalé.
Qualité des données
Tous les essais inclus sont de petite taille, comptant de 60 à 100 participants, et présentent un niveau élevé de variation (hétérogénéité). Les résultats doivent donc être interprétés avec prudence. Dans l'ensemble, nous avons considéré que les essais inclus présentaient un risque de biais mal défini.
Authors' conclusions
Background
Description of the condition
Cardiovascular diseases (CVD), which include coronary heart disease (CHD) and cerebrovascular disease, are a variety of conditions that affect the heart and blood vessels (WHO 2013). They are the leading cause of death worldwide, with over 17 million deaths per year attributed to CVD (WHO 2013). In 2011, CVD accounted for nearly 160,000 deaths in the UK (BHF 2014). Around 74,000 of these deaths were caused by CHD (BHF 2014). Low‐ and middle‐income countries (LMICs) are also affected by CVD. In 2001, three quarters of global deaths from CHD took place in LMICs (WHO 2013). According to Gaziano 2010, the rapid increase in CHD burden in LMICs is attributable to the acquisition of lifestyle‐related risk factors, socio‐economic changes and an increase in life span in these countries.
Dietary factors may play a vital role in the development of CVD and its risk factors, and may contribute to the geographic variability in CVD morbidity and mortality (Yusuf 2001; Scarborough 2011). These factors are important, not only because they have been linked to CVD development, but also because they can be modified. This makes them one of the main targets for interventions aimed at the primary prevention and management of CVD.
Description of the intervention
Nuts have been in the human diet for thousands of years. Indeed, records show that people ate pistachio nuts as far back as 7000 BC (King 2008). This pattern has continued today with nuts eaten globally in a variety of ways, including as ingredients in recipes and as snacks (King 2008). However, the amount of nuts consumed around the world differs. For instance, nut consumption in countries with a Mediterranean style diet is twice that of those with an American diet (Dreher 1996; Sabaté 2006).
In botanical terms, nuts are considered as a dry fruit with one seed that has a hard shell or pericarp (Sabaté 2006). Nuts are an energy dense food containing around 44% to 76% total fat (Sabaté 2010). Their saturated fatty acid content is low, being 4% to 16%, with almost half of their total fat content consisting of unsaturated fatty acids (Ros 2006). They also contain protein (around 10% to 30%) (Blomhoff 2006) and a number of micronutrients and minerals, such as folic acid, selenium, zinc and niacin (Brufau 2006). Many nuts also contain antioxidants (Vinson 2012). Almonds, for instance, contain a number of flavonoids such as catechins, flavonols and flavonones, whilst walnuts contain a variety of polyphenols and tocopherols (Blomhoff 2006). Raw walnuts and toasted almonds in particular have shown a high antioxidant efficacy (Vinson 2012). This may explain why almonds and walnuts are the main types of nuts studied in trials (based on 25 nut consumption trials included in Sabaté 2010).
How the intervention might work
A number of epidemiological studies show that nuts have a beneficial effect on CVD risk factors (Fraser 1992; Albert 2002). For example, in the Iowa Women's Health Study (IWHS) CVD mortality was lower in people who ate nuts/peanut butter five or more times per week (hazard ratio (HR) 0.67 (95% confidence interval (CI) 0.56 to 0.81; Blomhoff 2006). Evidence also comes from systematic reviews of observational studies. Mukuddem‐Petersen 2005 conducted a systematic review of observational studies looking at the effect of nuts on the lipid profile of normal and hyperlipidemic participants. They found that eating 1.5 to 3.5 servings of nuts five times or more per week along with a heart healthy diet significantly reduced low‐density lipoprotein cholesterol (LDL‐C) and total cholesterol. In particular, three studies of almond consumption showed that consuming 50 g to 100 g of almonds per day was associated with 4% to 17% lower total cholesterol and 7% to 19% lower LDL‐C in both hypercholesterolemia and non‐cholesterolemic participants. Randomised controlled trials (RCTs) also provide some evidence that nuts may be beneficial for people with CVD risk factors (Sabaté 1993; Jenkins 2002). One RCT comparing a recommended cholesterol‐lowering diet with walnuts to one without found that incorporating 84 g of walnuts on a daily basis for four weeks decreased serum levels of total cholesterol by 12% (Sabaté 1993).
The mechanisms by which nuts reduce CVD risk are not exactly known. However, nuts contain a number of nutritional attributes that have been linked to cardioprotection (Sabaté 2010). For instance, walnuts contain high levels of n‐3 fatty acids, which are known to be cardioprotective (Sabaté 2010). It is thought that the individual nutrients contained in nuts or the composite of cardioprotective nutrients that they contain, or both, may account for the beneficial effect of nuts on CVD and its risk factors (Kris‐Etherton 2008). Furthermore, the beneficial ratio of unsaturated fatty acids to saturated fatty acids may be an important factor in the health benefits associated with frequent nut intake (Sabaté 2010).
Why it is important to do this review
Despite the potential benefits from increased nut intake, there are few systematic reviews examining the effects of nut consumption on CVD prevention and, of those available, most include observational studies which are subject to bias and confounding (Mukuddem‐Petersen 2005; Afshin 2014; Zhou 2014). One systematic review of the effects of walnut consumption on lipid levels included only short term trials (average six weeks duration) so the sustained effects could not be established (Banal 2009). To our knowledge, there are currently no systematic reviews examining RCT evidence on CVD risk factors or clinical outcomes over the longer term (three months or more). To address this, this Cochrane review will examine evidence from RCTs of three months or longer duration on nut consumption for the primary prevention of CVD in the general population and in people at high risk of CVD.
Objectives
To determine the effectiveness of nut consumption for the primary prevention of CVD.
Methods
Criteria for considering studies for this review
Types of studies
We included RCTs, reported as full‐text articles, those published as abstract only, and unpublished data.
Types of participants
We included trials of adults (aged 18 years and older) from the worldwide population with and without CVD risk factors (e.g. hypertension, hyperlipidemia, overweight/obese). As this Cochrane review is interested in the primary prevention of CVD, we excluded people who experienced a myocardial infarction (MI), stroke, revascularization procedure (coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI)), and those with angina or angiographically defined CHD.
We excluded trials with > 25% of participants with diagnosed CVD or type 2 diabetes at baseline. Whilst type 2 diabetes is a major risk factor for CVD, interventions targeting specifically this patient group are covered by Cochrane reviews from the Cochrane Metabolic and Endocrine Disorders Group.
Types of interventions
We included trials comparing the provision of nuts or advice to increase nut consumption with no intervention or minimal intervention (e.g. leaflets with no face‐to‐face intervention or reinforcement). Trials were excluded in which the control group received an intervention which was not also given to the intervention group, e.g. particular foods other than nuts. We included trials of at least 12 weeks duration. Longer term studies are most informative in terms of behavioral change and sustained changes for public health interventions, with follow‐up being seen as the time elapsed since the start of the intervention.
We did not include multifactorial intervention studies, such as those including exercise or lifestyle interventions, in this review in order to avoid confounding.
Types of outcome measures
Primary outcomes
-
Cardiovascular mortality.
-
All‐cause mortality.
-
Non‐fatal endpoints such as MI, CABG, PCI, angina or angiographically defined CHD, stroke, carotid endarterectomy and peripheral arterial disease (PAD).
Secondary outcomes
-
Changes in blood pressure (systolic and diastolic) and blood lipids (total cholesterol, LDL‐C, high‐density lipoprotein (HDL) cholesterol and triglycerides).
-
Type 2 diabetes as a major CVD risk factor.
-
Health‐related quality of life (using any validated scale).
-
Costs.
-
Adverse effects (as defined by the authors of the included trials, e.g. weight gain, anaphylaxis).
Search methods for identification of studies
Electronic searches
We systematically searched the following bibliographic databases on 30 July 2015:
-
Cochrane Central Register of Controlled Trials (CENTRAL, Issue 6 of 12, 2015) on the Cochrane Library.
-
MEDLINE (Ovid, 1946 to July week 4 2015).
-
EMBASE Classic and EMBASE (Ovid, 1947 to 2015 July 29).
-
CINAHL (EBSCO, 1937 to 17 July 2015).
-
Web of Science Core Collection (Thomson Reuters, 1970 to 29 July 2015).
-
Database of Abstracts of Reviews of Effects (DARE, Issue 2 of 4, 2015), Health Technology Assessment Database (HTA, Issue 2 of 4, 2015) and Health Economics Evaluations Database (HEED, Issue 2 of 4, 2015) on the Cochrane Library.
The search strategies are presented in Appendix 1. We applied the Cochrane sensitivity‐maximizing RCT filter (Lefebvre 2011) to MEDLINE (Ovid) and adaptations of it to the other databases, except the Cochrane Library.
We did not apply any restrictions on language of publication.
Searching other resources
We searched ClinicalTrials.gov (www.ClinicalTrials.gov) and the WHO International Clinical Trials Registry Platform (WHO ICTRP) Search Portal (apps.who.int/trialsearch/) on 12 March 2015. Search terms used were 'cardiovascular and nuts'.
We contacted trial authors for additional information. We checked reference lists of included trials reports and review articles for additional references.
Data collection and analysis
Selection of studies
Two review authors (AA, RG, NM, KR) independently screened titles and abstracts for inclusion of all the potential studies identified by the searches and coded them as 'retrieve' (eligible or potentially eligible/unclear) or 'do not retrieve'. We retrieved the full‐text study reports/publications. Two review authors (AA, RG, NM, KR) independently screened the full‐text articles and identified trials for inclusion, and identified and recorded reasons for exclusion of the ineligible studies. We resolved any disagreements through discussion. We identified and excluded duplicates and collated multiple reports of the same trial so that each trial, rather than each report, is the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram (Figure 1) and Characteristics of excluded studies table.
Study flow diagram.
Data extraction and management
We used a data collection form for trial characteristics and outcome data, which was piloted on at least one trial included in the review. Two review authors (AA, NM) extracted trial characteristics from included trials. We extracted the following trial characteristics:
-
Methods: trial design, total duration of trial, details of any 'run in' period, number of trial centres and location, trial setting, withdrawals and date of trial.
-
Participants: number, mean age, age range, gender, severity of condition, diagnostic criteria, inclusion criteria and exclusion criteria.
-
Interventions: intervention, comparison, concomitant medications and excluded medications.
-
Outcomes: primary and secondary outcomes specified and collected, and time points reported.
-
Notes: funding for trial, and notable conflicts of interest of trial authors.
Two review authors (AA, NM) independently extracted outcome data from included trials. We resolved disagreements by consensus or by involving a third review author (KR). One review author (NM) transferred data into the RevMan 2014 file. We double‐checked that data were entered correctly by comparing the data presented in the systematic review with the study reports. A second review author (KR) spot‐checked trial characteristics for accuracy against the trial report.
Assessment of risk of bias in included studies
Two review authors (AA, NM) independently assessed risk of bias for each included trial using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved any disagreements by discussion or by involving another review author (KR). We assessed the risk of bias according to the following domains:
-
Random sequence generation.
-
Allocation concealment.
-
Blinding of participants and personnel.
-
Blinding of outcome assessment.
-
Incomplete outcome data.
-
Selective outcome reporting.
-
Other bias (e.g. industry funding).
We graded each potential source of bias as either 'high', 'low' or 'unclear' and provided a quote from the study report together with a justification for our judgment in the 'Risk of bias' section of the Characteristics of included studies table. We summarised the 'Risk of bias' judgements across different trials for each of the domains listed. Where information on risk of bias related to unpublished data or correspondence with a trial author, we noted this in the 'Risk of bias' section of the Characteristics of included studies table.
When considering treatment effects, we took into account the risk of bias for the trials that contributed to that outcome.
Assessment of bias in conducting the systematic review
We conducted this Cochrane review according to the published protocol (Martin 2015) and reported any deviations from it in the Differences between protocol and review section.
Measures of treatment effect
We analysed continuous data as mean difference (MD) with 95% CIs. For continuous variables we present data for the change from baseline rather than end point data. We entered data presented as a scale with a consistent direction of effect, with the exception of HDL cholesterol where an increase in this outcome is a positive finding.
We planned to analyse dichotomous data as odds ratios or risk ratios with 95% CIs but that did not apply to any of our included analyses. Narratively describing skewed data reported as medians and interquartile ranges also did not apply to this review.
Unit of analysis issues
Studies with multiple intervention groups
Data for the control group have been used for each intervention group comparison. We reduced the weight assigned to the control group by dividing the control group N by the number of intervention groups analysed.
Cluster RCTs
We planned to analyse cluster RCTs using the unit of randomisation (cluster) as the number of observations. However, none of the included trials are cluster RCTs.
Cross‐over studies
For included cross‐over studies we only used the first period.
Dealing with missing data
We contacted trial authors or study sponsors in order to verify key trial characteristics and obtain missing numerical outcome data where possible (e.g. when a study is identified as abstract only). Where papers did not report results as change from baseline we calculated this. For the standard deviation differences we followed the methods presented in the Cochrane Handbook for Systematic Reviews of Interventions for imputing these (Section 16.1.3.2: Imputing standard deviations for changes from baseline; Higgins 2011), and assumed a correlation of 0.5 between baseline and follow‐up measures as suggested by Follman 1992.
Assessment of heterogeneity
We used the I² statistic to measure heterogeneity among the trials in each analysis. If no heterogeneity was present, a fixed‐effect meta‐analysis was performed. If we identified substantial heterogeneity (I² statistic > 50%) we reported it and explored possible causes by prespecified subgroup analysis if there were a sufficient number trials. If the heterogeneity could not be explained, we either provided a narrative overview or used a random‐effects model with appropriately cautious interpretation.
Assessment of reporting biases
If we had been able to pool more than 10 trials, we would have created and examined a funnel plot to explore possible small study biases for the primary outcomes. However, this did not apply to this Cochrane review.
Data synthesis
We performed statistical analysis using RevMan 2014. None of the included trials reported dichotomous data, which we would have added as events and the number of participants. We entered continuous data as means and standard deviations. We performed meta‐analysis if treatments and participants were similar enough. In the absence of substantial heterogeneity (> 50%) and provided that there are sufficient trials, we combined the results using a fixed‐effect model.
Subgroup analysis and investigation of heterogeneity
We planned to carry out the following subgroup analyses:
-
Type of nut.
-
Dosage.
-
Duration of the intervention.
-
Type of intervention (provision or advice).
-
Risk level of participants (presence of CVD risk factors versus no CVD risk factors).
We planned to use the formal test for subgroup interactions in RevMan 2014.
There were an insufficient number of trials included in this Cochrane review to perform these analyses.
Sensitivity analysis
We planned to carry out sensitivity analyses by only including trials at low risk of bias. However an insufficient number of trials met the inclusion criteria of the review to do this.
Results
Description of studies
Results of the search
The search of the main databases retrieved 4050 references, of which 1311 were duplicates, leaving 2739 to screen. The search of clinical trial registers retrieved 52 trial reports, of which 6 were duplicates, leaving 46 to screen. One reference was identified through reference checking. In total we screened 2786 references by title and abstract and excluded 2521. We retrieved and screened the full text of 265 references, which led to the exclusion of 252. Eleven references were eligible for inclusion and reported on five studies. We identified one ongoing study and one study is awaiting classification. A PRISMA flow chart (Figure 1) illustrates this process.
Included studies
This Cochrane review includes five trials (reported in 11 references) with a total of 435 participants (see Characteristics of included studies).
All studies were RCTs. Two studies were trials with two parallel treatment arms (Balci 2012; Abazarfard 2014). One trial had two relevant treatment arms of different amounts of hazelnuts consumed and are denoted Tey 2013 (30 g) and Tey 2013 (60 g) as they are entered separately in the analysis. One study had four treatment arms (Tey 2011) of which we were only interested in the nut intervention and control arm. One study, Sabaté 2005, was a cross‐over trial and we only used the first period, as planned and described in the Unit of analysis issues section.
Two trials were conducted in New Zealand (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g)), one in USA (Sabaté 2005), one in Iran (Abazarfard 2014) and one does not provide the information (Balci 2012). None of the included RCTs investigated advice to increase nut consumption but instead they all provided nuts to the intervention groups. Almonds were studied in one trial (Abazarfard 2014, 50 g per day), walnuts in two trials (Balci 2012, 10 g per day; Sabaté 2005, 28 g to 56 g per day) and hazelnuts in two trials (Tey 2013 (30 g), 30 g per day; Tey 2013 (60 g), 60 g per day; Tey 2011, 42 g per day). Control groups were asked to follow their usual diet (Sabaté 2005; Tey 2011; Tey 2013 (30 g); Tey 2013 (60 g)) or to follow advice for a balanced diet (Abazarfard 2014) or healthy nutrition (Balci 2012). All participants, regardless of allocation to intervention or control group, were advised to maintain their usual activity habits (Sabaté 2005; Tey 2013 (30 g); Tey 2013 (60 g); Abazarfard 2014) or had their physical activity measured at baseline and during the intervention (Tey 2011). One trial did not provide information on physical activity (Balci 2012).
In all studies, recruitment took place via public advertisements from the general population. The number of randomised participants per trial ranged from 60 to 110 (N = 60 in Balci 2012; N = 110 in Tey 2013 (30 g) and Tey 2013 (60 g) combined; N = 108 in Abazarfard 2014; N = 94 in Sabaté 2005 and N = 63 in the two arms of interest in Tey 2011). The mean age of participants ranged from 37.4 to 54.3 years.
Studies varied in the types of participants recruited. One trial recruited only female participants (Abazarfard 2014) while the other trials included male and female participants with similar ratios (Balci 2012, 45% male; Tey 2011, 47% male; Tey 2013 (30 g) and Tey 2013 (60 g), 43% male; Sabaté 2005, 44% male). Abazarfard 2014 included pre‐menopausal women aged 20 to 55 years with a BMI ≥ 25 kg/m². Balci 2012 included participants with prediabetic metabolic syndrome. Tey 2013 (30 g) and Tey 2013 (60 g) included participants with BMI ≥ 25 kg/m², aged between 18 and 65 years. Tey 2011 also included participants aged 18 to 65 years with a BMI < 30 kg/m². Information for Sabaté 2005 differed between the two papers reporting this study regarding the inclusion of participants with BMI < 35 kg/m²/BMI ≤ 35 kg/m², aged between 30 and 72 years.
The duration of the intervention varied between the included trials. Two RCTs had a follow‐up of 12 weeks (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g)), two RCTs of three months (Balci 2012; Abazarfard 2014) and one RCT for six months (Sabaté 2005).
We have provided details about one ongoing study, NCT01950806, in the Characteristics of ongoing studies table. This study investigates whether pecans are effective in reducing the risk of CVD or diabetes.
One study awaiting classification (Njike 2015) is a published conference abstract. We have contacted the study authors for further details and are awaiting a response (Characteristics of studies awaiting classification).
Excluded studies
We have listed 116 references (reporting 112 studies) of the 252 excluded references with reasons for exclusion in the Characteristics of excluded studies table, because they most closely missed the inclusion criteria. The studies were excluded because they were not randomised (N = 12), the intervention was not of interest (N = 9), the intervention for the control group was inappropriate (N = 23), studying the wrong type of participants (N = 12) or the follow‐up was less than 12 weeks (N = 56).
Risk of bias in included studies
We have presented details of the risk of bias in included trials as part of the Characteristics of included studies table, and a 'Risk of bias' graph and summary in Figure 2 and Figure 3, respectively.
'Risk of bias' graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.
'Risk of bias' summary: review authors' judgements about each 'Risk of bias' item for each included study. One study had two intervention arms (Tey 2013 (30 g) and Tey 2013 (60 g)), and whilst the two arms were treated separately in the analysis, we assessed the risk of bias for the total trial. The risk of bias for the second arm of Tey 2013 (60 g) is the same as Tey 2013 (30 g).
Overall, the risk of selection bias was either at low risk of bias or unclear. Blinding of participants was judged to be of unclear risk of bias as participants cannot be blinded to the food they eat and it was unclear whether or not personnel were blinded. Blinding of the outcome assessors and incomplete outcome reporting were judged to be at low risk of bias in two studies and unclear risk of bias in the remaining three. Reporting bias and other sources of bias were judged to be unclear for all included studies. One trial had two intervention arms (Tey 2013 (30 g) and Tey 2013 (60 g)), and whilst the two arms were treated separately in the analysis, we assessed the risk of bias for the total trial.
Allocation
The methods of random sequence generation were unclear in two studies (Sabaté 2005; Balci 2012). Three of the five studies which stated the method of random sequence generation were judged to have a low risk of bias in this domain (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g); Abazarfard 2014). The methods of allocation concealment were unclear in three studies (Sabaté 2005; Balci 2012; Abazarfard 2014) and judged to be of low risk in two studies (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g)).
Blinding
Most trials were judged to be of unclear risk of bias in the domain blinding of participants and personnel as we consider it to be impossible to blind participants to the food they consume and it was unclear whether personnel were blinded or not. One trial's clinical trial registry entry stated it as double‐blinded (Abazarfard 2014) but because participants are unable to be blinded to food they eat, it was judged to be at unclear risk of bias. One trial, Tey 2011, revealed from communication with the author that the personnel were blinded and this trial was considered to be at low risk of bias for this domain.
Outcome assessors were said to have been blinded in two trials (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g)). They have therefore been judged to be of low risk of bias for this domain.
Incomplete outcome data
One of the five included trials, Abazarfard 2014, reported losses to follow‐up with similar numbers of losses in the intervention and control arm but without an intention‐to‐treat (ITT) analysis. We judged this study to be of unclear risk of bias. Two studies could not be judged on attrition bias as no information was provided (Sabaté 2005; Balci 2012). Two trials conducted an ITT analysis but did not report on how missing data were dealt with (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g)). We therefore judged them to be of unclear risk of bias.
Selective reporting
Selective reporting was judged to be an unclear risk of bias in all five studies (Sabaté 2005; Tey 2011; Balci 2012; Tey 2013 (30 g) and Tey 2013 (60 g); Abazarfard 2014) because of insufficient information to make this judgement.
Other potential sources of bias
For all included RCTs there was insufficient information to judge the risk of bias from other potential sources.
Effects of interventions
We found no studies investigating the effect of advice to increase nut consumption.
All five included RCTs investigated the effect of nut provision. None of the five trials reported on our primary outcomes: cardiovascular mortality, all‐cause mortality and non‐fatal endpoints such as MI, CABG, PCI, angina or angiographically defined CHD, stroke, carotid endarterectomy and PAD. Four of the five included trials provided usable data for analysis (Tey 2011; Balci 2012; Tey 2013 (30g) and Tey 2013 (60g); Abazarfard 2014). One trial did not provide any usable data and the trial author did not reply to our enquiry for the missing information (Sabaté 2005).
Changes in blood pressure (systolic and diastolic)
Three trials (267 participants) measured mean change in systolic and diastolic blood pressure (mmHg) from baseline to last follow‐up (Balci 2012; Tey 2013 (30g) and Tey 2013 (60g); Abazarfard 2014), which was three months for Abazarfard 2014 and Balci 2012 and 12 weeks for Tey 2013 (30 g) and Tey 2013 (60 g). For systolic blood pressure, heterogeneity was substantial between trials (I² statistic = 55%) and we used a random‐effects model. There was no statistically significant effect of nut consumption on systolic blood pressure (MD 3.22 mmHg, 95% CI ‐2.70 to 9.14; P = 0.29; Analysis 1.1). However, one trial was a clear outlier with an extremely large reduction in systolic blood pressure in the control arm during the trial period of 14 mmHg (Abazarfard 2014). Excluding this trial from the sensitivity analysis gave a MD of ‐0.28 mmHg (95% CI ‐5.79 to 5.23; P = 0.92).
For diastolic blood pressure, because of the substantial heterogeneity (I² statistic = 87%) the pooled effect estimate was suppressed and results from individual trials are plotted (Analysis 1.2). One trial showed a statistically significant reduction (Abazarfard 2014), the two arms of the Tey 2013 trial showed a statistically significant increase in diastolic blood pressure with the intervention (Tey 2013 (30g); Tey 2013 (60g)) and the remaining trial showed no effect of the intervention (Balci 2012).
Changes in blood lipids (total cholesterol, LDL‐C, HDL cholesterol and triglycerides)
Three trials (266 participants) measured mean change in total cholesterol (mmol/L), HDL cholesterol (mmol/L), LDL cholesterol (mmol/L) and triglycerides (mmol/L) from baseline to last follow‐up (Tey 2011; Tey 2013 (30g) and Tey 2013 (60g); Abazarfard 2014), which was three months for Abazarfard 2014 and 12 weeks for Tey 2011, Tey 2013 (30g) and Tey 2013 (60g).
For total cholesterol, there was substantial heterogeneity between the trials (I² statistic = 79%) which precluded meta‐analysis. Results for individual trials are plotted and the pooled effect estimate is suppressed (Analysis 1.3). One trial showed large and statistically significant reduction in total cholesterol with nut consumption (Abazarfard 2014), one a small and borderline significant reduction (Tey 2011), and the remaining two arms of the Tey 2013 trial showed no effect of the intervention on total cholesterol levels (Tey 2013 (30g); Tey 2013 (60g)).
For LDL cholesterol, there was low to moderate heterogeneity between the trials (I² statistic = 25%) and we used a fixed‐effect model. There was no statistically significant effect of nut consumption in lowering LDL cholesterol (MD ‐0.02, 95% CI ‐0.1 to 0.06; P = 0.57; Analysis 1.4).
For HDL cholesterol, there was no heterogeneity between the trials (I² statistic = 0%), and we used a fixed‐effect model for the analysis. There was no effect of nut consumption on HDL levels in the pooled analysis (MD ‐0.00, 95% CI ‐0.04 to 0.04; P = 0.82; Analysis 1.5).
For triglycerides, there was substantial heterogeneity between trials (I² statistic = 84%) and we did not perform a meta‐analysis. We plotted individual trials and the pooled effect estimate is suppressed (Analysis 1.6). One trial showed a significant reduction in triglyceride levels with nut consumption (Abazarfard 2014), the remaining two trials showed no effect of the intervention on triglyceride levels (Tey 2011; Tey 2013 (30g); Tey 2013 (60g)).
Occurrence of type 2 diabetes as a major CVD risk factor
None of the included trials reported on type 2 diabetes as a major CVD risk factor.
Health‐related quality of life (using any validated scale)
None of the included trials reported on quality of life.
Costs
None of the included trials reported on costs.
Adverse effects (as defined by the authors of the included trials, e.g. weight gain, anaphylaxis)
One trial, Tey 2011, reported one adverse event (1/32, allergic reaction to nuts) but also found that "nuts can be incorporated into the diet without adversely affecting body weight". This was also found by two other trials. Sabaté 2005 reports a non‐significant weight gain when consuming walnuts daily (mean daily consumption 35 g) for six months. Tey 2013 (30g) and Tey 2013 (60g) noted "no significant weight change" for both nut groups over the study period.
Discussion
Summary of main results
We included five trials (435 participants randomised), one ongoing trial and one study is awaiting classification. All trials examined the provision of nuts to increase consumption rather than dietary advice. None of the included trials reported on our primary outcomes, CVD clinical events, but trials were small and short term. All five trials reported on CVD risk factors. Four of these trials provided data in a useable format for meta‐analyses, but heterogeneity precluded meta‐analysis for most of the analyses. Overall trials were judged to be at unclear risk of bias.
Heterogeneity was significant for most analyses of CVD risk factors with inconsistent effects seen for systolic blood pressure, diastolic blood pressure, total cholesterol and triglycerides. Heterogeneity was low for LDL and HDL cholesterol where there was no significant effect of nut consumption on these outcomes although the number of studies contributing to these analyses was small. Three trials monitored adverse events. One trial reported an allergic reaction to nuts and three trials no significant weight gain with increased nut consumption. None of the included trials reported our other secondary outcomes, occurrence of type 2 diabetes as a major risk factor for CVD, health‐related quality of life and costs.
Overall completeness and applicability of evidence
None of the included RCTs reported on our primary outcomes.
Five trials reported on CVD risk factors (lipid levels and blood pressure) with inconsistent findings. There were limitations in the available data as only four trials provided data in a useable format for meta‐analyses and for the remaining study we were unable to obtain additional information from the trial authors. The findings to date for these outcomes are inconclusive.
Quality of the evidence
Overall, the included RCTs were at unclear risk of bias for most of the 'Risk of bias' domains. Provison of nuts is a behavioural intervention and as such participants cannot be blinded to group allocation although trial personnel were blinded in two studies (Tey 2011; Abazarfard 2014). Outcome assessors were blind to group allocation in three trials (Tey 2011; Tey 2013 (30 g) and Tey 2013 (60 g); Abazarfard 2014).
All studies recruited small numbers of participants and small study bias is of particular concern for this review (Sterne 2000; Sterne 2001; Nüesch 2010). Due to the relatively small number of included studies we were unable to examine the effects of publication bias in funnel plots.
Potential biases in the review process
We performed a comprehensive search across major databases for interventions involving increased nut consumption for this Cochrane review. In addition, we screened the reference lists of systematic reviews and contacted study authors for information when needed. Two review authors independently performed all screening, inclusion and exclusion of articles, and extracted data from the included trials.
We only included trials with a minimum of 12 weeks duration as longer term studies are most informative in terms of behavioral change and sustained changes for public health interventions. However, this limited the number of trials eligible for inclusion and we excluded most studies because they were short term interventions.
Agreements and disagreements with other studies or reviews
In terms of our primary outcome, CVD clinical events, none of the included trials reported this outcome. Previous evidence for an association between nut consumption and clinical endpoints comes from observational studies which are subject to bias and confounding (Blomhoff 2006).
We were unable to determine the effectiveness of nut consumption on major CVD risk factors (lipid levels and blood pressure) with the trials included in the current review due to missing information, heterogeneity between trials and the limited number of trials available. There are few systematic reviews examining the effects of nut consumption on CVD prevention and most include observational studies which are subject to bias and confounding (Mukuddem‐Petersen 2005; Afshin 2014; Zhou 2014). One systematic review of the effects of walnut consumption on lipid levels included only short term trials (average six weeks) so the sustained effects could not be established (Banal 2009).
A recent systematic review, Mohammadifard 2015, examined the effects of nut consumption on blood pressure and found beneficial effects on systolic blood pressure in participants at high risk of CVD. However, it is unclear if these effects are sustained as the review included very short term trials.
The global burden of disease (GBD) study suggests that 40% of the disability‐adjusted life‐years from ischemic heart disease are attributable to diets low in nuts and seeds (Lim 2012). This systematic review shows no RCT evidence on the effect of nuts on mortality or heart disease, bringing into question the validity of the evidence on the effect of diet on heart disease. The primary source of data for nuts in the GBD study is from non‐randomised studies.
Study flow diagram.
'Risk of bias' graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.
'Risk of bias' summary: review authors' judgements about each 'Risk of bias' item for each included study. One study had two intervention arms (Tey 2013 (30 g) and Tey 2013 (60 g)), and whilst the two arms were treated separately in the analysis, we assessed the risk of bias for the total trial. The risk of bias for the second arm of Tey 2013 (60 g) is the same as Tey 2013 (30 g).
Comparison 1 Secondary outcomes, Outcome 1 Systolic blood pressure.
Comparison 1 Secondary outcomes, Outcome 2 Diastolic blood pressure.
Comparison 1 Secondary outcomes, Outcome 3 Total cholesterol.
Comparison 1 Secondary outcomes, Outcome 4 LDL cholesterol.
Comparison 1 Secondary outcomes, Outcome 5 HDL cholesterol.
Comparison 1 Secondary outcomes, Outcome 6 Triglycerides.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Systolic blood pressure Show forest plot | 4 | 267 | Mean Difference (IV, Random, 95% CI) | 3.22 [‐2.70, 9.14] |
| 2 Diastolic blood pressure Show forest plot | 4 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 3 Total cholesterol Show forest plot | 4 | Mean Difference (IV, Random, 95% CI) | Totals not selected | |
| 4 LDL cholesterol Show forest plot | 4 | 266 | Mean Difference (IV, Fixed, 95% CI) | ‐0.02 [‐0.10, 0.06] |
| 5 HDL cholesterol Show forest plot | 4 | 266 | Mean Difference (IV, Fixed, 95% CI) | ‐0.00 [‐0.04, 0.04] |
| 6 Triglycerides Show forest plot | 4 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
