Abstract
BACKGROUND:
Despite appropriate antibiotic therapy, the risk of mortality in neonatal sepsis still remains high. We conducted a systematic review to comprehensively evaluate different adjuvant therapies in neonatal sepsis in a network meta-analysis.
METHODS:
We included randomized controlled trials (RCTs) and quasi-RCTs that evaluated adjuvant therapies in neonatal sepsis. Neonates of all gestational and postnatal ages, who were diagnosed with sepsis based on blood culture or sepsis screen were included. We searched MEDLINE, CENTRAL, EMBASE and CINAHL until 12th April 2021 and reference lists. Data extraction and risk of bias assessment were performed in duplicate. A network meta-analysis with bayesian random-effects model was used for data synthesis. Certainty of evidence (CoE) was assessed using GRADE.
RESULTS:
We included 45 studies involving 6,566 neonates. Moderate CoE showed IVIG [Relative Risk (RR); 95% Credible Interval (CrI): 1.00; (0.67–1.53)] as an adjunctive therapy probably does not reduce all-cause mortality before discharge, compared to standard care. Melatonin [0.12 (0–0.08)] and granulocyte transfusion [0.39 (0.19–0.76)] may reduce mortality before discharge, but CoE is very low. The evidence is also very uncertain regarding other adjunctive therapies to reduce mortality before discharge. Pentoxifylline may decrease the duration of hospital stay [Mean difference; 95% CrI: –7.48 days (–14.50–0.37)], but CoE is very low.
CONCLUSION:
Given the biological plausibility for possible efficacy of these adjuvant therapies and that the CoE from the available trials is very low to low except for IVIG, we need large adequately powered RCTs to evaluate these therapies in sepsis in neonates.
Introduction
Despite remarkable advancements in neonatal care, the incidence of sepsis and associated mortality still remains high. A recent systematic review estimated the incidence of early- and late-onset sepsis as 2,469 and 946 per 100,000 live births respectively [1]. The Global Burden of Disease study reported an annual incidence of 1.3 million cases of neonatal sepsis and 200,000 sepsis-related neonatal deaths [2, 3]. The burden is much higher in low- and middle-income countries (LMICs) when compared to high income countries. A recent large population-based study from LMICs showed that sepsis is the cause of approximately 35% of neonatal deaths in South Asia and Sub-Saharan Africa [4].
Newborn infants are more susceptible to sepsis and sepsis-related mortality due to their inadequate host defence mechanisms and immature organs systems that are overwhelmed when faced with systemic inflammation and hemodynamic compromise [5, 6]. Standard of care for neonatal sepsis includes timely treatment with appropriate antibiotics and cardiorespiratory supportive measures. The high mortality rates despite institution of standard care emphasises the need to explore adjuvant therapeutic interventions that could potentially aid in earlier resolution of septicemia and hence reduce mortality. Multiple immunological and non-immunological interventions such as intravenous immunoglobulin (IVIG), granulocyte or granulocyte-monocyte colony stimulating factor (GCSF or GMCSF), granulocyte transfusion, exchange transfusion, pentoxifylline and melatonin have been studied as adjuvant therapies to reduce mortality in neonatal sepsis [7–12]. We conducted a systematic review to assess the relative efficacy of various adjuvant therapies in addition to standard care in neonatal sepsis to reduce mortality and improve clinical outcomes. Given the choice of multiple competing interventions, we aimed to synthesise the evidence in a network meta-analysis (NMA).
Methods
The protocol was registered with PROSPERO [13]. We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline - NMA extension [14]
Study design, population, interventions and outcomes
We included randomized clinical trials (RCTs) and quasi-RCTs that had evaluated adjuvant therapies in neonatal sepsis in addition to standard care. Newborn infants of all gestational and postnatal ages, who were diagnosed with sepsis based on either blood culture or sepsis screen and clinical features were eligible for inclusion.
Twelve interventions in addition to standard care were evaluated: IVIG, IgM enriched IVIG (IgM-IVIG), IVIG enriched with immunoglobulins against Group B Streptococci (GBS-IVIG), GCSF, GMCSF, exchange transfusion, granulocytes transfusion, fresh frozen plasma (FFP), pentoxifylline, pentoxifylline-IgM IVIG, melatonin and zinc. Standard care was defined as appropriate antibiotic therapy with cardiorespiratory and other systemic supportive measures as required.
The primary outcome was all-cause mortality before hospital discharge. Secondary outcomes included bronchopulmonary dysplasia (BPD) defined as oxygen requirement at 28 days of life or 36 weeks postmenstrual age, necrotising enterocolitis (NEC) any stage, duration of hospital stay, periventricular leukomalacia (PVL), all-cause mortality at 18–24 months and neurodevelopmental disability at 18–24 months.
Literature search and risk of bias assessment
We searched four major databases Medline, CENTRAL, Embase and CINAHL from their inception till April 12, 2021. Reference lists of included trials and published systematic reviews were also searched to identify additional studies. Studies published in non-English language were also included. Four authors (TA, TB, SHS and NBS), in pairs of two, screened the title and abstract of all studies using an online software tool (Rayyan-QCRI, Doha, Qatar) [15] and independently assessed the full-text articles for inclusion. The search strategy is provided in Supplement eTable 1.
Two authors (TB and VVR) independently assessed the risk of bias of all included trials using the Cochrane risk of bias tool, version 2.0 [16]. Discrepancies were resolved by consulting a third author (MK).
Data extraction and data synthesis
Four authors (TA, TB, NBS and SHS), in pairs of two, independently extracted data from the included trials using a structured proforma. A NMA with bayesian random-effects model was used for data synthesis [17]. Non-informative priors and generalized linear models with 4 chains, burn-in of 50,000 iterations, followed by 100,000 iterations and 10,000 adaptations were used [18]. Geometry of networks for all outcomes was evaluated using network plots. Model convergence was assessed with the Gelman-Rubin statistic, trace plots and density plots [19]. Fit of the model was analysed with leverage plots, total residual deviance, and deviance information criterion. Node splitting was used to detect inconsistency [20]. I2 statistic and Cochran Q test were used to evaluate heterogeneity in pairwise meta-analyses of direct evidence. Publication bias was assessed using a funnel plot when a meta-analysis included 10 or more trials.
Statistical analysis
Statistical analysis was performed with R (R Foundation for Statistical Computing, Vienna, Austria) [21]. Network estimates were expressed as risk ratio (RR) or mean difference (MD), with 95% credible interval (CrI), and were illustrated with league plots and forest plots. Ranking of interventions for all outcomes was done with surface under the cumulative ranking curve (SUCRA) plots [22]. Two authors [VVR, TA] ascertained the certainty of evidence (CoE) for all of the estimates according to the GRADE Working Group recommendations for a network meta-analysis [23].
We also conducted the following sensitivity analyses for the primary outcome all-cause mortality before hospital discharge: (1) separately analysing the studies that included only preterm, only term and those that included both term and preterm neonates, and (2) separately analysing the studies that included neonates with blood culture proven sepsis alone, probable sepsis alone, and those that included both culture positive and probable sepsis.
Results
A total of 2,953 titles and abstracts were screened, of which 45 studies that had enrolled 6,566 neonates were included (Supplement e24–68]. Amongst the included studies were 2 three-armed RCTs [26, 44] and 1 four-armed RCT [28]. Twenty-seven (60%) studies included only preterm, one (2.2%) study included only term and 17 (37.8%) studies included both term and preterm neonates. Ten (22%) studies included only blood culture proven sepsis, one (2.2%) study included only probable sepsis and 34 (75.6%) included both blood culture proven and probable sepsis. The characteristics of the included trials are given in Table 1.
Characteristics of included studies
Characteristics of included studies
Abbreviations: ANC, absolute neutrophil count; EOS, early onset sepsis; FFP, fresh frozen plasma; GBS, group B streptococci; GCSF, granulocyte colony stimulating factor; GMCSF, granulocyte monocyte colony stimulating factor; IQR, interquartile range; IV, intravenous; IVIG, intravenous immunoglobulin; LOS, late onset sepsis; SC, subcutaneous; SD, standard deviation.
Amongst the 45 included trials, 30 (66.7%) had high risk, 8 (17.8%) had some concerns and 7 (15.6%) had a low risk of bias. The risk of bias assessment of individual trials is given in Table 2. Majority of the studies had some concerns for the domain ‘randomisation process’ as the methods of allocation concealment and / or random sequence generation were not described. Also, most studies had some concerns for the domain ‘selection of reported results’ due to non-availability of a published protocol.
Risk of bias of included trials assessed using Cochrane risk of bias tool version 2.0
Risk of bias of included trials assessed using Cochrane risk of bias tool version 2.0
Forty-four studies including 6,502 neonates had reported on the primary outcome ‘all-cause mortality before hospital discharge’ [24–57, 59–68]. Figure 1 shows the network plot, and SUCRA plot with standard care as the common comparator. Supplement eFigure 2 shows the league plot that depicts the network estimates for various comparisons. Node-splitting analysis showed inconsistency in the network for the comparisons granulocyte transfusion vs. standard care and granulocyte transfusion vs. IVIG (Supplement eFigure 3). Forest plots for the direct evidence are provided in Supplement eFigure 4. CoE assessment for the primary outcome is listed in Table 3. The characteristics of the networks for all of the outcomes are given in Supplement eTable 2.
All-cause mortality before hospital discharge. A. Network plot; B. SUCRA plot Abbreviations: FFP, fresh frozen plasma; GBS, group B streptococci; GCSF, granulocyte colony stimulating factor; GMCSF, granulocyte monocyte colony stimulating factor; IVIG, intravenous immunoglobulin.
Certainty of evidence using GRADE recommendations
1Downgraded by two levels for very serious risk of bias. 2Downgraded by one level for serious imprecision due to small sample size not meeting the OIS. 3Downgraded by two levels for very serious imprecision due to data from one study with small sample size and confidence interval including ‘no difference’. 4Downgraded by two levels for very serious imprecision due to single digit event rate and confidence interval including ‘no difference’. 5Downgraded by one level for serious risk of bias. 6Downgraded by one level for serious imprecision due to confidence interval including ‘no difference’. 7RR not estimable due to zero event rate in both groups. 7Downgraded by two levels for very serious heterogeneity. 8Downgraded by one level for serious inconsistency due to heterogeneity. 9Downgraded by one level for serious heterogeneity.
1Downgraded by two levels for very serious imprecision due to data from one study with small sample size and confidence interval including ‘no difference’. 2Downgraded by one level for serious risk of bias. 3Downgraded by two levels for very serious risk of bias. 4Downgraded by two levels for very serious imprecision due to single digit event rate and confidence interval including ‘no difference’. 5Downgraded by one level for serious imprecision due to confidence interval including ‘no difference’.
1Downgraded by one level for serious risk of bias. 2Downgraded by one level for serious imprecision due to confidence interval including ‘no difference’. 3Downgraded by two levels for very serious imprecision due to data from one study with small sample size. Values in bold are statistically significant. GRADE Ranking the Quality of Evidence. High quality - Very confident that the true effect lies close to that of the estimate of the effect. Moderate quality - Moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality - Confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low quality - Very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
Abbreviations: FFP, fresh frozen plasma; GBS, group B streptococci; GCSF, granulocyte colony stimulating factor; GMCSF, granulocyte monocyte colony stimulating factor; IVIG, intravenous immunoglobulin.
Moderate CoE showed that neither IVIG [RR 1.00 (95% CrI 0.67, 1.53)] nor GBS IVIG [RR 0.00 (95% CrI 0.00, 1.05)] as an adjunctive therapy reduce all-cause mortality before discharge when compared to standard care alone. When compared to standard care alone, melatonin [0.12 (0, 0.08)] and granulocyte transfusion [0.39 (0.19, 0.76)] reduced the risk of all-cause mortality before discharge. Melatonin was ranked as the most beneficial intervention (SUCRA value 0.99) followed by granulocyte transfusion (SUCRA value 0.86). However, the CoE was very low for both the interventions. A comparison of interventions among themselves showed granulocyte transfusion was more beneficial than IVIG [0.39 (0.17, 0.81)], melatonin was more beneficial than pentoxifylline-IgM IVIG [0.09 (0, 0.87)] and most of the interventions were better than GBS IVIG. The CoE for all these comparisons was very low (eTable 3).
Sensitivity analysis evaluating the trials that included only preterm neonates showed that granulocyte transfusion was more beneficial than standard care in reducing the all-cause mortality [0.41 (0.16, 0.94)]. Analysis of trials that included both term and preterm neonates did not show beneficial effect for any adjuvant therapy (Supplement eFigures 5–11).
Sensitivity analysis including only neonates with blood culture proven sepsis showed that none of the adjuvant therapies were beneficial in decreasing the risk of the primary outcome, when compared to standard care alone. Analysis of trials that included both blood culture proven and probable sepsis showed melatonin and granulocyte transfusion to be more effective in reducing all-cause mortality before discharge when compared to standard care alone (Supplement eFigures 12–17).
NMA evaluating seven interventions including standard care showed that none of the adjunctive therapies were effective in reducing the risk of NEC during or after the sepsis episode (Supplement eFigures 18–20). Similarly, NMA of six interventions including standard care showed no adjunctive therapy was effective in reducing the risk of BPD (Supplement eFigures 21–23). The CoE was moderate for IVIG vs. standard care and low to very low for all other comparisons for both the outcomes (eTable 3).
NMA evaluating 10 interventions including standard care showed that pentoxifylline reduced the duration of hospital stay when compared to standard care alone [MD –7.48 (–14.50, –0.37)] and pentoxifylline with IgM IVIG [–15.51 (–30.46, –0.53)] (Supplement eFigures 24–26).
Limited data was available from the included RCTs on other secondary outcomes and hence NMA could not be performed. Pairwise meta-analysis did not show any beneficial effect of adjuvant therapies on outcomes such as PVL, mortality at 18–24 months, neurodevelopmental disability at 18–24 months (Supplement eFigures 27–30).
Discussion
This systematic review and NMA included 45 trials including 6,566 neonates to evaluate 12 adjuvant therapies in neonatal sepsis. Moderate CoE showed that IVIG as an adjuvant therapy in neonatal sepsis did not reduce all-cause mortality before discharge. The CoE was low to very low for other adjunctive therapies.
Considering the fact that host defences in neonates are immature [5, 6], many adjuvant therapies to boost the immune mechanisms have been tried in neonatal sepsis (Fig. 2). Actively acquired immunity is negligible in neonates due to lack of prior exposure to antigens as well as negligible endogenous production of immunoglobulins. In preterm neonates, passively acquired immunity is also deficient especially in those who are born prior to the third trimester when the maximal transplacental transfer of immunoglobulins occurs. These immunoglobulins are essential to bind to cell surface receptors, provide opsonic activity, activate complement and promote antibody-dependent cytotoxicity, all of which are hence deficient in neonates. In lieu of these, administration of exogenous immunoglobulins as an adjunctive therapy to sepsis in neonates has been evaluated in many trials. The largest multi-centric RCT (INIS trial) evaluating IVIG as an adjuvant therapy in neonatal sepsis did not find a significant reduction in mortality before hospital discharge [52]. Similarly, the recent Cochrane systematic review did not show a reduction in mortality before discharge with either IVIG or IgM enriched IVIG as adjuvant therapy in neonatal sepsis [10]. Whilst the CoE was not evaluated in the previous review [10], the CoE was moderate for the network estimates in our NMA.
Infographics depicting the therapeutic role of different adjuvant therapies in neonatal sepsis. Abbreviations: FFP, fresh frozen plasma; GBS, group B streptococci; GCSF, granulocyte colony stimulating factor; GMCSF, granulocyte monocyte colony stimulating factor; IVIG, intravenous immunoglobulin; MODS, multiorgan dysfunction syndrome; PMN, polymorphonuclear; PRR, pattern recognition receptor; SIRS, systemic inflammatory response syndrome; TLR, toll-like receptor.
Sepsis in neonates is frequently associated with neutropenia. Further, neutrophils and macrophages in neonates are functionally immature, and have impaired chemotactic and phagocytic ability [5, 6]. Hence, adjuvant therapies such as granulocyte transfusion and GCSF/GMCSF have been evaluated as adjuncts in the treatment of neonatal sepsis. The recent Cochrane systematic reviews did not find a reduction in all-cause mortality before discharge with either of these adjuvant therapies [7, 8]. In this NMA, though granulocyte transfusion was associated with a reduction in all-cause mortality, the evidence was very uncertain due to very low CoE. Similarly, the effect of GCSF/GMCSF on the outcome all-cause mortality before discharge is very uncertain as per the results of our NMA.
Oxidative stress due to reactive oxygen and nitrogen species induced by pro-inflammatory mediators is an important component of septicemia and its complications [69]. Melatonin is an effective anti-oxidant (Fig. 2). It also has anti-inflammatory properties by preventing the translocation of nuclear factor-kappa B of activated B cells, which mitigates the upregulation of pro-inflammatory cytokines [70]. A recent systematic review showed that melatonin significantly reduced C-reactive protein levels and improved the clinical status in neonates with sepsis [71]. Quite similarly, our NMA showed melatonin adjuvant therapy to significantly reduce all-cause mortality before discharge. However, the data was from a single small RCT and the CoE was very low. Future trials which are adequately powered may evaluate the potential role of melatonin as an adjuvant to standard care in neonates with sepsis.
Pentoxifylline is a phosphodiesterase inhibitor, which has anti-inflammatory effects, improves tissue microcirculation and mitigates the procoagulant state in sepsis (Fig. 2) [72]. The recent Cochrane systematic review on pentoxifylline as an adjuvant therapy in sepsis showed significant reduction in all-cause mortality before discharge. This was based on data from six small RCTs and the CoE was low [9]. However, in our NMA, pentoxifylline was found to have a significant effect only on the duration of hospital stay. The effect on mortality and other outcomes were not significant, with the CoE being very low to low.
Zinc is an immunomodulator and has anti-inflammatory effects mediated by inhibition of calprotectin complexes [73]. Zinc deficiency is shown to be associated with impaired immunity in neonates and children [74]. A meta-analysis of three small RCTs showed zinc as adjuvant therapy significantly reduced all-cause mortality before discharge [11]. Double volume exchange transfusion is a common practice in neonates with severe sepsis. It probably works by removing the bacterial toxins and pro-inflammatory cytokines from the blood and replacing fresh and immunologically replete blood in the neonate. A recent meta-analysis of both RCTs and observational studies showed a reduction in mortality with exchange transfusion [75]. However, we did not find a significant reduction in mortality before discharge with either zinc or exchange transfusion, with CoE being very low to low.
An NMA that compared four immunotherapies namely IVIG, IgM-IVIG, GCSF and GMCSF also showed none of these was effective in reducing all-cause mortality or duration of hospital stay in neonates with sepsis [12]. Though we do have the rationale and biological plausibility to consider that these adjuvant interventions in sepsis could be effective [76], the available evidence is limited and does not support routine use of any of these adjuvant therapies. We need large well designed RCTs to evaluate these interventions. Further, the molecular mechanisms and pathophysiology of neonatal sepsis and its complications need to be studied in detail. This would not only guide us on the optimal choice and timing of adjuvant therapy, but also would aid in identifying newer therapeutic interventions.
To the best of our knowledge, this is the first NMA to comprehensively evaluate all the adjuvant therapies in neonatal sepsis. There are several limitations. Many of the included trials were conducted two decades ago, when the diagnosis and management of sepsis in neonates were underdeveloped compared to the present times. Most included trials had small sample size and high risk of bias, which were the major reasons to downgrade the CoE. The number of trials evaluating only culture positive sepsis was less. We did not assess the adverse effects of adjuvant therapies and other outcomes such as intraventricular hemorrhage and retinopathy of prematurity. We had only limited data on long-term mortality and neurodevelopmental outcomes. Finally, the severity of sepsis and risk of mortality being starkly different in preterm neonates when compared to term neonates, there is a possibility of intransitivity in our NMA. The definition of probable sepsis and the dosage regimens of the adjuvants were also variable across the studies. Though we had addressed these issues by multiple sensitivity analyses, we consider this as one of the major limitations of this NMA. We could not perform a sensitivity analysis based on sepsis severity score score and culture profile of the organisms (gram positive versus gram negative sepsis) due to availability of limited data in the included studies.
To conclude, given the biological plausibility for possible efficacy of these adjuvant therapies and that the CoE from the available trials is very low to low, we need large adequately powered RCTs with proper study designs to evaluate these therapies in sepsis in neonates.
Conflict of interest
None
Contributors’ statement
Dr Thangaraj Abiramalatha and Dr Viraraghavan Vadakkencherry Ramaswamy conceptualized the systematic review. Abdul Kareem Pullattayil S devised the literature search strategy for the databases and extracted the full texts of included articles. Dr Tapas Bandyopadhyay, Dr Thangaraj Abiramalatha, Dr Sanjana Hansoge Somanath and Dr Nasreen Banu Shaik were responsible for literature search and data extraction. Dr Viraraghavan Vadakkencherry Ramaswamy and Dr Thangaraj Abiramalatha did the statistical analysis of data. Dr Venkat Reddy Kallem designed the infographics. Dr Thangaraj Abiramalatha and Dr Monika Kaushal did the GRADE assessment. Dr Thangaraj Abiramalatha provided the first draft of the manuscript. Dr Viraraghavan Vadakkencherry Ramaswamy and Dr Monika Kaushal provided further intellectual inputs and revised the initial draft. All authors approved the final version for submission and agree to be accountable for all aspects of the work.
