Probiotic and synbiotic therapy in critical illness: a systematic review and meta-analysis

Background Critical illness is characterized by a loss of commensal flora and an overgrowth of potentially pathogenic bacteria, leading to a high susceptibility to nosocomial infections. Probiotics are living non-pathogenic microorganisms, which may protect the gut barrier, attenuate pathogen overgrowth, decrease bacterial translocation and prevent infection. The purpose of this updated systematic review is to evaluate the overall efficacy of probiotics and synbiotic mixtures on clinical outcomes in critical illness. Methods Computerized databases from 1980 to 2016 were searched. Randomized controlled trials (RCT) evaluating clinical outcomes associated with probiotic therapy as a single strategy or in combination with prebiotic fiber (synbiotics). Overall number of new infections was the primary outcome; secondary outcomes included mortality, ICU and hospital length of stay (LOS), and diarrhea. Subgroup analyses were performed to elucidate the role of other key factors such as probiotic type and patient mortality risk on the effect of probiotics on outcomes. Results Thirty trials that enrolled 2972 patients were identified for analysis. Probiotics were associated with a significant reduction in infections (risk ratio 0.80, 95 % confidence interval (CI) 0.68, 0.95, P = 0.009; heterogeneity I2 = 36 %, P = 0.09). Further, a significant reduction in the incidence of ventilator-associated pneumonia (VAP) was found (risk ratio 0.74, 95 % CI 0.61, 0. 90, P = 0.002; I2 = 19 %). No effect on mortality, LOS or diarrhea was observed. Subgroup analysis indicated that the greatest improvement in the outcome of infections was in critically ill patients receiving probiotics alone versus synbiotic mixtures, although limited synbiotic trial data currently exists. Conclusion Probiotics show promise in reducing infections, including VAP in critical illness. Currently, clinical heterogeneity and potential publication bias reduce strong clinical recommendations and indicate further high quality clinical trials are needed to conclusively prove these benefits.


Background
Critical illness is characterized by a loss of commensal flora and an overgrowth of potentially pathogenic bacteria, leading to a high susceptibility to acquired nosocomial infections [1,2]. Further, sepsis following infection is still a leading cause of death worldwide [3]. The U.S. Centers for Disease Control indicates death rates from critical illness/sepsis have increased at a rate greater than any other common cause of mortality in the last year for which data were available [4]. Thus, therapies to reduce the risk and incidence of infection and sepsis in critical illness are urgently needed.
According to the World Health Organization and the Food and Agriculture Organization, probiotics are living non-pathogenic microorganisms, which have demonstrated well-documented beneficial health effects administered in optimum amounts in the prevention and treatment of several disease states [5]. So far, several mechanisms by which probiotics may exert beneficial effects have been described, including modification of the gut flora by inducing host cell antimicrobial peptides, release of antimicrobial factors, suppression of the immune cell proliferation, stimulation of mucus and IgA production, anti-oxidative activity, inhibition of epithelial cell nuclear factor kappa B activation, and other potentially vital gut epithelial barrier protective effects [6][7][8].
As the gut is hypothesized to play a central role in the progression of critical illness, sepsis and multiple organ dysfunction syndrome [9], maintenance of the gut barrier and a healthy gut microbiome, potentially via reintroduction of commensal bacteria (probiotic therapy), may be essential to optimizing outcomes in critically ill patients.
According to current literature, the efficacy of probiotics in the prevention of infectious complications has been extensively evaluated in many animal studies and clinical trials in heterogenous intensive care unit (ICU) patient populations. These studies suggest that probiotics may reduce the incidence of infection, particularly ventilator-associated pneumonia (VAP) [10], which is a common serious complication in intubated, mechanically ventilated patients [11]. Nonetheless, the effect of probiotics on the prevention of VAP still remains controversial and inconclusive [12][13][14][15][16][17]. In fact, its effect depends on the patient population and the probiotic strain studied. Despite the outcome benefits of probiotics therapy, recent guidelines have been unable to make a definitive recommendation for the routine use of probiotics in ICU patients. To date, these guidelines have suggested the use of probiotic therapy in select medical and surgical patient populations in whom trials have documented safety and clinical benefits [18,19].
Over the last few years, several systematic reviews and meta-analyses have evaluated the effects of probiotics in critically ill patients [12][13][14][15][16][17]. In 2012, after aggregating 11 trials that reported on infections [14], we demonstrated that probiotics may reduce infections, including the incidence of VAP, although the effect on VAP was not statistically significant given the available data. Moreover, probiotics were associated with a trend toward reduced ICU mortality, but did not influence hospital mortality. Since our last systematic review and meta-analyses, seven new trials of probiotic therapy have been published [20][21][22][23][24][25][26]. Further, to date, no recent meta-analysis has examined the effect of probiotic versus synbiotic (probiotic and prebiotic fiber) therapy. Finally, a Canadian survey [27] on the use of probiotics as a prophylactic strategy for VAP showed that most Canadian ICU pharmacists have used probiotics at least once, although routine use is considered controversial and considerable practice variability exists. Thus, any increased understanding that the newly published trials can yield will be vital to clarifying clinical probiotic use in the ICU and areas in need of future research focus.
Therefore, as probiotic use in the ICU remains widespread and controversial, current guidelines are not conclusive, and with a significant number of new trials of probiotic use published recently we conducted a comprehensive systematic review and meta-analysis of probiotic and synbiotic use in critically ill patients. Our aim was to elucidate the overall efficacy of probiotics, as a single strategy or in combination with fiber therapy (synbiotics) on relevant clinical outcomes, particularly infection and VAP, in adult critically ill patients.

Search strategy and study identification
A literature search was conducted in MEDLINE, Embase, CINAHL, the Cochrane Central Register of Controlled Trials and the Cochrane Database of Systematic Reviews to identify all relevant randomized controlled trials (RCTs) published between 1980 and April 2016. The literature search used broad search terms containing "randomized," "clinical trial," "nutrition support," "enteral nutrition", "probiotics," and "synbiotics". No language restrictions were applied. Personal files and reference lists of relevant review articles were also reviewed.

Eligibility criteria
We included trials with the following characteristics: 1. Type of study: randomized controlled parallel group trials 2. Population: adult (≥18 years of age) critically ill patients. If the study population was unclear, we considered a mortality rate higher than 5 % in the control group to be consistent with critical illness 3. Intervention: Probiotics alone or associated with prebiotics (synbiotics) compared to a placebo 4. Outcomes: pre-specified clinical outcomes in ICU patients such as infectious complications, VAP, mortality, ICU and hospital length of stay (LOS), and diarrhea We excluded trials that reported only nutrition, biochemical, metabolic, or immunologic outcomes. Data published in abstract form were included only if additional information about the study design was obtained from the authors. The methodological quality of the included trials was assessed in duplicate by two reviewers independently using a data abstraction form with a scoring system from 0 to 14 according to the following criteria: 1. The extent to which randomization was concealed 2. Blinding 3. Analysis based on the intention-to-treat (ITT) principle 4. Comparability of groups at baseline 5. Extent of follow up 6. Description of treatment protocol 7. Co-interventions 8. Definition of clinical outcomes Consensus between both reviewers on the individual scores of each of the categories was obtained. We attempted to contact the authors of included studies and requested additional information not contained in published articles. We designated studies as level I if all of the following criteria were fulfilled: concealed randomization, blinded outcome adjudication and an ITT analysis, all which are the strongest methodological tools to reduce bias. A study was considered as level II if any one of the above-described characteristics were unfulfilled.

Data synthesis
All analyses, except the test for asymmetry, were conducted using RevMan 5.3 (Cochrane IMS, Oxford, UK) with a random effects model. We combined data from all trials to estimate the overall weighted mean difference (WMD) with 95 % confidence intervals for LOS data the pooled risk ratio (RR) with 95 % confidence intervals (CIs) for the incidence of infections and mortality, and diarrhea. WMDs were estimated by the inverse variance approach and pooled RRs were calculated using the Mantel-Haenszel estimator. The random effects model of DerSimonian and Laird was used to estimate variances for the Mantel-Haenszel and inverse variance estimators [28]. RRs were undefined and excluded for studies with no event in either arm. Heterogeneity was tested by a weighted Mantel-Haenszel χ 2 test and quantified by the I 2 statistic as implemented in RevMan. Differences between subgroups were analyzed using the test of subgroup differences described by Deeks et al., and the results expressed using the P values. We considered P <0.05 to be statistically significant and P <0.10 as an indicator of trends. Funnel plots were used to assess the possibility of publication bias and the Egger regression test was used to measure funnel plot asymmetry [29].

Clinical outcomes
Overall infections were the primary outcome for this meta-analysis. Secondary outcomes were VAP, mortality, ICU and hospital LOS, and finally diarrhea. We used definitions of infections as defined by the authors in their original articles. From all trials, we combined hospital mortality where reported. Mortality specified at either 28 days or 90 days was not considered as ICU or hospital mortality, respectively. Nonetheless, if the mortality time frame was not specified as either ICU or hospital, it was presumed to be the later.

Subgroup analysis
We utilized predefined subgroup analyses to assess a number of possible influences on the effect of probiotic supplementation on clinical outcomes, and thus to explore the possible causes of heterogeneity. On the basis that the higher the daily dose the greater the effect, we first examined trials that administered a high dose of probiotics defined as >5 × 10 9 colony-forming units (CFU)/day vs. lower dose probiotics defined as <5 × 10 9 CFU/day. Second, we compared the results of RCTs that administered Lactobacillus plantarum as probiotic therapy vs. no L. Plantarum, and compared trials using Lactobacillus rhamnosus strain GG (LGG) vs. those administering other non-LGG strains.
Moreover, based on a larger treatment effect in those more seriously ill patients with higher risk of death, we compared studies including patients with higher mortality vs. lower mortality. Mortality was considered to be high or low based on whether it was greater or less than the median control group mortality of all the trials. Trials of higher quality, defined as those with a methodological score equal to or higher than the median quality score, may demonstrate a lower treatment effect.

Study identification and selection
A total of 79 relevant citations were identified from the search of computerized bibliographic databases and a review of reference lists from related articles. Of these, we excluded 49 due to the following reasons: 21 trials did not include ICU patients (mostly surgical patients); 12 articles were systematic reviews and meta-analyses; 4 trials were published as an abstract and we were unable to obtain the data from the authors to complete our data abstraction process; 5 articles were duplicates of included trials; 3 studies did not evaluate clinical outcomes; 2 trials tested multiple interventions; 1 study was not a RCT, and finally 1 study administered probiotics as oral swabs.
Finally, 30 RCTs [10, 20-26, 30-51] met our inclusion criteria and were included, covering a total of 2972 patients (see Tables 1 and 2). The reviewers reached 100 % agreement on the inclusion of the trials. The mean methodological score of all trials was 9, whereas the median value was 9.5 on a maximum of 14 (range 5-13). Randomization was concealed in 9/30 trials (30 %), ITT analysis was performed in 18/30 trials (60 %), and double blinding was done in 20/30 of the studies (67 %). There were five level-I studies and 25 level-II studies. The details of the methodological quality of the individual trials are shown in Table 1.

Primary outcome: infections Overall effect on new infections
Aggregating the results of the 14 trials reporting overall infections, probiotics were associated with a significant        Fig. 1).

Overall effect on hospital length of stay
Aggregating the data from the nine RCTs that reported hospital LOS, there were no significant differences between the groups (WMD -0.58, 95 % CI -3.66, 2.50, P = 0.71; I 2 = 74 %, P < 0.00001).

Antibiotic days
When we aggregated the data of four trials reporting on antibiotic days, we found that probiotics were significantly associated with a reduction in the duration of antibiotic therapy (WMD -1.12, 95 % CI -1.72, -0.51, P = 0.0003; I 2 = 32 %, P = 0.22; Fig. 5).

Subgroup analysis Probiotics daily dose
There were similar rates of infectious complications in RCTs using high-dose probiotic therapy (n = 8 trials)

Higher vs. lower mortality
The median hospital mortality rate of all the trials (or ICU mortality when hospital mortality was not reported) in the control group was 19 %. After aggregating nine studies with a higher mortality rate, probiotics significantly reduced the incidence of infections (RR 0.74; 95 % CI 0.57, 0.96; P = 0.02; I 2 = 58 %, P = 0.01) (Fig. 7). However, probiotics did not have an effect on infections in the five studies with lower mortality (RR 0.85; 95 % CI 0.66, 1.11; P = 0.24; I 2 = 23 %, P = 0.27). The test for subgroup differences was not significant (P = 0.43) (Fig. 7).

Publication bias
There was indication that potential publication bias influenced the observed aggregated results. In fact, funnel plots were created for each study outcome and the tests

Discussion
To date, our systematic review and meta-analysis is the largest and most updated evaluation of the overall effects of probiotics in the critically ill. It is also the first to include an analysis of symbiotic (probiotic/fiber combinations). Based on the analysis of 30 trials enrolling 2972 patients we demonstrated that probiotics are associated with a significant reduction in ICU-acquired infections, including VAP, which is the most common infectious complication in the critically ill. This significant effect on VAP is a new finding from our previous systematic reviews. Further, the beneficial effect of probiotics on reduction of infections is stronger with the publication of the new trials and the data no longer show a statistically significant effect of heterogeneity on this endpoint. Despite the probiotic effect of reducing infectious complications, this therapy did not influence ICU or hospital mortality, although none of the trials were powered to detect an effect on mortality. Overall, there was a tendency towards a reduction in ICU LOS and probiotic therapy did not influence other clinical endpoints such as hospital LOS, and diarrhea. Statistical and clinical heterogeneity was observed for some endpoints, although this was significant for the key endpoints of infectious complications and VAP. In addition, publication bias for overall infections and hospital LOS means that larger, well-powered, and more definitive clinical trials are urgently needed aimed to avoid these biases. Moreover, subgroup analysis showed that those trials with lower methodological quality exhibit the best treatment effect, which is another issue indicating that larger, welldesigned studies are needed. Again, with the exception of four trials, most of the included studies (n = 14) that reported mortality had small sample sizes, and hence were underpowered and inadequate to detect any clinically important treatment effects of probiotic therapy on mortality. Moreover, the inferences we can make from our current findings are further weakened, as randomization was concealed in 30 % of trials, whereas double-blinding was performed in 67 % of trials. Over recent years, several systematic reviews and meta-analyses have been conducted, although our metaanalysis is the largest and most current to date, as it contains the seven new suitable trials published since the most recent comprehensive meta-analysis publication on this topic, which focused on overall infections and other outcomes (not primarily VAP) in 2012 [14]. Further, these previous systematic reviews did not include analysis of synbiotic therapy. Overall, we have examined several    relevant clinical outcomes in a heterogenous ICU patient population, and therefore our results could be applied to a broad group of critically ill patients with sepsis, trauma, severe pancreatitis, or who have undergone surgery. Specific to pancreatitis, concerns have been raised about the safety of probiotic therapy following the 2008 trial, Probiotic prophylaxis in patients with predicted severe acute pancreatitis (PROPATRIA) [43], which showed that Ecologic 641® given with fiber post-pyloric was associated with higher mortality and bowel ischemia. This postpyloric method of administration was associated with an increase in small bowel necrosis, which was subsequently associated with death in a number of patients receiving the prebiotic fiber/probiotic mixture. It is possible that the post-pyloric administration of this fiber/multiple probiotic strain mixture in patients with pancreatitis may carry  OR odds ratio significant risk and should likely be avoided [52]. Unfortunately, there were significant ethical and statistical concerns raised about the conduct of the trial [53], limiting the utility of the data. Further, more recently, a systematic review and meta-analysis by Gou et al. [54] found that probiotics had neither beneficial nor adverse effects in patients with pancreatitis.
Despite the limitations of the PROPRIATA trial, it has contributed to concerns around the safety of probiotic administration in critical illness and limited the design of larger-scale clinical trials and/or more routine clinical administration of live probiotics. To address this, the American Health Care Research and Quality (AHRQ) agency reviewed and reported on the safety of probiotic therapy in over 600 published clinical trials and case reports [55]. It should be reassuring to future investigators that the overall conclusion of the extensive AHRQ report indicates that probiotic therapy in both adults and pediatric populations was not been found to be associated with any increased risk of infectious or other adverse events in either healthy or ill patients. Importantly, their report revealed a trend towards less adverse events in probiotic-treated critically ill patients, although isolated adverse effects of probiotic administration have been reported [56]. In any case, careful and appropriate safety monitoring in all future probiotic clinical trials should be conducted.
Recent data indicate that infection during critical illness continues to be a major challenge worldwide. A multi-national ICU study of 14,414 patients in 1265 ICUs from 75 countries, revealed that 51 % of ICU patients were considered infected on the day of survey and 71 % were receiving antibiotics [57]. Of the infections in this study, 64 % were of respiratory origin and the ICU mortality rate in infected patients was more than twice that of non-infected patients (25 % vs. 11 %, P < 0.001), as was the hospital mortality rate (33 % infected vs. 15 % non-infected, P < 0.001) [57].
Currently, VAP is the second most common nosocomial infection in the USA and the most prevalent ICUacquired infection. Notwithstanding, its incidence is highly variable depending on diagnostic criteria used to identify this infectious complication. In fact, in 2015 Ego et al. [58] reported that the incidence of VAP ranged from 4 % to 42 % when using the six published sets of criteria and from 0 % to 44 % when using the 89 combinations of criteria for hypoxemia, inflammatory response, bronchitis, chest radiography, and microbiologic findings. In our systematic review we found that the incidence of VAP ranged from 9 % [46] to 80 % [39]. Additionally, the apparent effect of probiotics on VAP is largely driven by the studies of Kotzampassi et al. [39] study and the Zeng et al. [26]; both trials explain 45.5 % of the signal and thus, provide an unstable estimate. Moreover, current knowledge shows that VAP is associated with high cost and poor clinical outcomes [59]. In 2002, Rello et al. [60] demonstrated that VAP leads to an additional US$40,000 in hospital charges per patient, and recently it has been suggested that the use of prophylactic probiotics may be cost-effective for prevention of VAP from a hospital perspective [61].
Probiotic therapy may prevent VAP and other infections by restoring non-pathogenic flora, which competes with nosocomial pathogens inhibiting their overgrowth, modulating local and systemic immune response, and improving gut barrier function. However, in spite of these protective effects the role of probiotics as a nonpharmacological strategy in preventing VAP has previously been inconclusive. In 2010, Siempos et al. [12] aggregated five probiotic trials demonstrating a reduction in the incidence of VAP, whereas in 2012 Petrof et al. [14] and subsequently Barraud et al. [13] and Wang et al. [15] did not demonstrate any significant effect of probiotic therapy on VAP. More recently, a Cochrane review of probiotic therapy specifically for VAP [17], found with low quality of evidence that probiotic therapy is associated with a reduction in the incidence of VAP. Our current systematic review demonstrates a significant treatment effect of probiotics in reducing VAP and did not demonstrate statistical heterogeneity, strengthening the signal that this may be an effective therapy for VAP. Recently, a Canadian survey [27] on the use of probiotics as a prophylactic strategy for VAP showed that most Canadian ICU pharmacists have used probiotics at least once, although they do not routinely recommend probiotics for the prevention of VAP.
Currently, a large number of clinical trials have demonstrated that probiotics may reduce the incidence of antibiotic-associated diarrhea and Clostridium difficile infections, and systematic reviews have confirmed a significant signal of benefit on reduction of diarrhea and C. difficile-related colitis in all patients (not confined to ICU patients) [62,63]. Our results, when focused on ICU patients do not currently demonstrate a treatment benefit of probiotics in preventing and treating diarrhea in the critically ill, including antibioticassociated diarrhea.
An interesting finding of our meta-analysis was a reduction in antibiotic use in those patients who received probiotics. Nonetheless, only four trials [10,21,26,48] comprising 13 % of included studies reported duration of antibiotic therapy as an outcome. In addition, the study of Zeng et al. contributed to 90 % of the signal, which is a very unstable estimate that weakens our finding. Therefore, probiotics may shorten the duration of antibiotic therapy, although the limited clinical trial data available for this endpoint limits the strength of these findings and further investigation of this effect is needed.
We currently have a greater understanding about the potential benefits of probiotics therapy in critical illness, although much more data are needed. Subgroup analysis found that certain strains such as L. plantarum alone or in combination was associated with a significant reduction in overall infections, although the test for subgroup differences was not significant (P = 0.21). Certain specific biological properties have been described for L. plantarum, including an ability to prevent adhesion of pathogens to the intestinal epithelium secondary to the production of adhesins, enolase, and phosphoglycerate kinase on the bacterial surface [64,65]. These mechanisms may be crucial to reduction of bacterial translocation and modulation of local inflammatory response, and therefore the effect of this strain on systemic infectious complications. Interestingly, probiotics alone had a greater effect than synbiotics on infections, although the difference between these subgroups was not significant (P = 0.98) and more data on the specific effects of different prebiotic fibers are needed. Finally, future trials also need to focus on evaluating the changes in the microbiome following critical illness and the effect of probiotic or synbiotics on restoring a healthy microbiome in treated patients [66]. Recent advances in microbiome sequencing technology (16 s rRNA) in the last few years have resulted in an unprecedented growth in the amount of sequence data that can be collected at a previously unattainable low cost [66]. Thus, if we speculate that a specific probiotic or synbiotic therapy can be used to treat dysbiosis (a pathological change in the patient's bacterial flora) and restore a healthy microbiome, we need to evaluate this with the new accessible microbiome analysis techniques currently available. This may help us target probiotic or probiotic mixtures in the future and increase the personalization of care.
The strength of this current systematic review includes the use of several methods to reduce bias (comprehensive literature search, duplicate data abstraction, specific criteria for searching and analysis), and the analysis of relevant clinical outcomes in the critically ill. However, several important limitations in drawing strong treatment inferences are present. These include the significant potential for publication bias for the infection and hospital LOS outcomes, and the small numbers of trials included in subgroup analyses. In addition, the variety of probiotic strains, wide range of daily doses, and length of administration of probiotic therapy among the different trials weaken any possible clinical conclusions and recommendations. We were also unable to perform subgroup analysis for all clinical outcomes due to the limited number of studies evaluating each endpoint.
Based on our current data, there is not currently sufficient evidence to make a final strong recommendation for probiotics to be utilized in the prevention of infections, including VAP, in the critically ill. However, our current guideline recommendations suggest that probiotics should be considered to improve outcome in critically ill patients [19]. Future trials continue to need to address questions about timing, daily dose, and duration of therapy, which still remain unanswered.

Conclusion
In the largest systematic review and meta-analysis of probiotics to date, we demonstrated that in 30 trials enrolling 2972 patients, probiotics significantly reduced the incidence of infectious complications, including new episodes of VAP in critically ill patients. This finding is limited by clinical heterogeneity and potential publication bias for the overall infection outcome. This precludes a more meaningful statistical conclusion of the efficacy of probiotic therapy on overall infections and potentially the prevention of VAP in critical illness. Moreover, according to our findings probiotics has been demonstrated to be more effective in those trials with higher mortality in the control group. Probiotic therapy with L. plantarum currently demonstrates the most significant effect on the reduction of infections. Overall, the variety of strains, wide range of daily doses, and length of administration of probiotics weakens the strength of our conclusion. Certainly, additional large-scale, adequately powered, well-designed clinical trials, aimed at confirming our observations, are needed and warranted.

Key messages
Critical illness is characterized by a loss of commensal flora and an overgrowth of potentially pathogenic bacteria, leading to a high susceptibility of nosocomial infections Probiotics are living non-pathogenic microorganisms, which may protect the gut barrier, attenuate pathogen overgrowth, decrease bacterial translocation and prevent infection in ICU patients Probiotic use in the ICU remains widespread and controversial, current guidelines are not conclusive, and a significant number of new trials of probiotics have been published recently, which requires a current and comprehensive systematic analysis of probiotic and synbiotic therapy in critically ill patients Probiotics were associated with a significant reduction in infections and a significant reduction in the incidence of ventilator-associated pneumonia (VAP) was found in critically ill patients receiving probiotics alone versus synbiotic mixtures, demonstrating the greatest improvement in infectious outcome, limited synbiotic trial data are currently available Currently, clinical heterogeneity and potential publication bias reduce strong clinical recommendations and indicate further high-quality clinical trials are needed to conclusively prove these benefits Probiotics shows promise for the reduction of infections, including VAP in critical illness, and should be considered in critically ill patients Abbreviations CFU, colony-forming unit; CI, confidence interval; C.Random, concealed randomization; EN, enteral nutrition; ICU, intensive care unit; Ig A, immunoglobulin A; ITT, intention to treat; LGG, Lactobacillus rhamnosus strain GG; LOS, length of stay; MV, mechanical ventilation; NA, non-attributable; NR, non-reported; OR, odds ratio; RCT, randomized controlled trial; RNA, ribonucleic acid; RR, risk ratio; VAP, ventilator-associated pneumonia; WMD, weighted mean difference

Funding
No funding for the development, writing or submission of this manuscript was received.
Authors' contributions WM contributed to development of the concept of the manuscript, study grading, study selection, evaluation and interpretation of data, and also performed primary authoring and editing of all drafts of the manuscript. ML contributed to study grading, selection, evaluation and interpretation of data, performed much of the primary statistical analysis, meta-analysis and data analysis, and also contributed to the writing of the manuscript. PL contributed to development of study grading, study selection, evaluation and interpretation of data, and also contributed substantially to the writing of the manuscript. PW contributed to development of the concept of the manuscript, evaluation and interpretation of data, and also performed authoring and editing of all drafts of the manuscript. All authors read and approved the final manuscript.