Introduction

The neonatal mortality rate is a crucial health indicator. Infections cause almost one-fourth (23%) of neonatal deaths worldwide, with 15% of these deaths resulting from neonatal sepsis [1]. Sepsis is a systemic inflammatory reaction caused by microorganisms invading the bloodstream, leading to extreme symptoms such as fever and shock [2]. Neonatal sepsis is classified into early-onset sepsis (EOS) and late-onset sepsis (LOS). EOS is defined as sepsis within 72 h of birth, and LOS defines as sepsis occurring at or after 72 h of life [3]. Early detection of neonatal sepsis is challenging, so antibiotics are given empirically when sepsis is suspected to prevent severe consequences.

The unnecessary use of broad-spectrum antibiotics in empirical therapy leads to an increase in multidrug-resistant microorganisms in neonatal intensive care units (NICU) and puts a high burden on developing countries. The world health organization (WHO) defines antibiotic resistance as a major public health issue that requires immediate attention [4].

Gram-negative (GN) bacteria-induced neonatal sepsis is a crucial cause of morbidity and mortality in neonates [5]. Neonatal GN sepsis is becoming more prevalent globally, with a concerning rise in multidrug-resistant infections [3, 6]. It has been estimated that 214,000 deaths from neonatal sepsis are attributed to resistant pathogens annually [7]. Sepsis is the fourth leading cause of neonatal mortality in Iran, with an estimated 16% prevalence in hospitalized neonates [8,9,10]. The high use of empirical and prophylactic antibiotics goes against the recommended therapies [11]. Healthcare professionals face a challenge due to antibiotic resistance. We conducted a systematic review and meta-analysis of published data on gram-negative neonatal sepsis from various regions of Iran due to the increasing evidence of multidrug resistance in neonatal sepsis caused by GN bacteria. The aim was to determine the prevalence of gram-negative bacteria and their antibiotic resistance pattern in neonatal sepsis.

Materials and methods

The systematic review followed the Preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines for systematic reviews and meta-analyses [12]. The review methods were not established prior to the conduct of the review.

Eligibility criteria

Cross-sectional studies reporting bacterial blood culture and antibiotic resistance/sensitivity testing for neonates with sepsis were included if published in English or Persian language, performed in Iranian hospitals, and used a recognized standard for interpreting antibiotic susceptibility testing (European Committee on Antimicrobial Susceptibility Testing (EUCAST), Clinical and Laboratory Standards Institute (CLSI), British Society for Antimicrobial Chemotherapy). According to the WHO definition, a neonate or newborn infant is a child who is under 28 days old. Any samples over 28 days in age were excluded from the studies. Studies that only reported antibiotic sensitivity were excluded from the analysis. Studies that only reported gram-positive bacteria were excluded. Review studies, letters, case reports, and conference papers were excluded.

Information sources and search strategy

Four international electronic databases (PubMed/MEDLINE, EMBASE, Scopus, and Web of Science) and five Iranian databases (Magiran, Iranmedex, Irandoc, Scimed, and SID) underwent a broad electronic search. Additionally, we manually searched the first 100 records on Google Scholar. The databases were searched from the beginning up until July 28, 2023. Additionally, the references of included studies were searched for other potentially essential studies. Experts in neonatology and library science were consulted to select the search keywords. The used keywords in this study were as follows: ‘sepsis’, ‘septicemia’, ‘bacteremia’, ‘blood infection’, ‘infant’, ‘newborn’, ‘neonate’, ‘antibiotic resistance’, ‘antimicrobial resistance’, ‘Prevalence’, and their Persian equivalent. Our search was restricted to English and Persian publications. Detailed search strategies for PubMed database available in Supplementary file 1.

Study selection

All records have been imported to EndNote X8 and duplicates were eliminated. The records were screened by two reviewers, who independently considered inclusion and exclusion criteria based on title and abstract (MST, KM). The full-text of the selected articles was reviewed independently by two different reviewers (PRH, NM). Any disagreement was resolved through discussion among at least three reviewers (KM, MST, NM) until they reached a consensus.

Data extraction and data items

We used a researcher made data extraction checklist. The data extraction sheet underwent a pilot test on 10 randomly selected articles, followed by revisions and approval by consensus among researchers. The data items collected for every study consisted of author names, publication year, province, duration, hospital type, sample size (categorized by gender), positive culture (categorized by gender), early or late-onset sepsis, pathogen type, and antibiotic resistance. Data extraction was done by two reviewers independently. In case of disagreement, a third author was involved.

Quality assessment

The quality of the included studies was assessed using the analytical cross-sectional study checklist from the Joanna Briggs Institute (JBI) [13]. The checklist has eight questions that are signed with the answer “Yes”, “No”, and “Unclear”. Articles that scored above 7 were considered high-quality, while those between 4 and 6 were medium-quality, and those below 4 were low-quality. Two reviewers (MN and TSS) conducted the quality assessment and resolved discrepancies through consensus.

Synthesis of results

The Mantel–Haenszel method was used in performing a meta-analysis with comprehensive meta-analysis (CMA) (Version 2) software. Statistical heterogeneity was evaluated through the calculation of I2 statistics. We utilized a fixed or random-effect model with a 95% confidence interval (CI) depending on the level of heterogeneity. In the following of Cochrane criteria if the heterogeneity was ≥ 50 we used the random-effect model. To investigate sources of heterogeneity, sensitivity, and subgroup analyses were conducted, as well as meta-regression models. For each variable, the event rate was determined alongside a 95% CI. Egger's test and funnel plots were used to evaluate the presence of publication bias.

Results

Study selection

Figure 1 displays the flow diagram according to PRISMA guidelines, illustrating the search process and study selection. A total of 717 titles were retrieved from the databases. After removing duplicates, 191 papers were screened by title and abstract for possible inclusion in the study. After applying the eligibility criteria, 48 full-text articles remained for assessment. Based on the exclusion criteria, 17 articles were excluded after the assessment (Age of patients in seven studies was above 28 days, five studies reported just gram-positive bacteria, in two studies only antibiotic sensitivity was reported, two review studies and one study was conference paper). The review included 31 articles [14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44] that met the eligibility criteria.

Fig. 1
figure 1

The follow diagram of the literature selection process

Full size image

Characteristics of the selected studies

The studies that were included were published between 1998 and 2021, with the majority conducted in Iran's Center (n = 10) [35,36,37,38,39,40,41,42,43,44] based on geographical location, followed by the Northwest (n = 5) [30,31,32,33,34], West (n = 5) [25,26,27,28,29], North (n = 4) [14,15,16,17], East (n = 4) [21,22,23,24], and South (n = 3) [18,19,20]. The duration of the studies varied from six months to 10 years. Of all the studies, 22 were conducted in NICUs of maternity hospitals and nine in NICUs of children's hospitals. Studies assessed 104,566 neonates, among whom 6348 patients had positive blood cultures (6.07% of all cultures). Of all isolates, 69.8% were GN bacteria. Out of 24 studies that report blood cultures based on gender, 2474 males and 1919 females were found to have positive blood cultures. According to Table 1, sepsis was divided into EOS (n = 1030) and LOS (n = 486) into 12 studies.

Table 1 Characteristics of included studies
Full size table

Assessment quality of articles

Table 2 displays the results of the methodological evaluation of the included studies. The methodological quality of the studies included had a final score range of 5 to 8.

Table 2 Methodological evaluation of included studies
Full size table

There were six high-quality studies and 25 medium-quality studies. All studies were included eventually. All studies highlighted Q8 as the most important quality aspect, which confirmed the use of the right statistical analysis. Also, question number 6 which implied “Were strategies to deal with confounding factors stated?” was addressed in seven studies [14, 22, 28, 31, 37, 38, 41].

Total prevalence of gram-negative bacteria and sensitivity analysis

There was a high rate of heterogeneity in the prevalence of GN bacteria (I2 = 96.026, P < 0.001). According to 31 studies with a sample size of 104,566, GN bacteria in neonates with sepsis was estimated to be 53.6% [95% CI: 45.9– 61.1: P = 0.362] (Fig. 2). The studies conducted by Bahmani [29] and Rajabi [38] reported the lowest and highest prevalence of GN bacteria as 9.5% and 95.8%, respectively (Fig. 2).

Fig. 2
figure 2

Prevalence of Gram Negative bacteria in neonates with sepsis in Iran

Full size image

Sensitivity analysis for the prevalence of GN bacteria in Fig. 3 shows that after removing one study at a time, the result is still robust.

Fig. 3
figure 3

Sensitivity analysis for the prevalence of Gram Negative bacteria in neonates with sepsis in Iran

Full size image

Subgroup analysis of the prevalence of gram-negative bacteria cause neonatal sepsis based on geographical region

Among GN bacteria that caused neonatal sepsis, Klebsiella pneumoniae (K.pneumonia) (23.2% [95% CI: 17.5–30.0, P < 0.001]) followed by Escherichia coli (E.coli) (13.5% [95% CI: 9.4–18.9, P < 0.001]) were more prevalent. However, this pattern varied between different regions. As shown in Table 3, in the Center, Northwest, and West of Iran, K.pneumonia had the highest prevalence rate among GN bacteria causing neonatal sepsis (24.6% [95%CI: 16.1–35.6, P < 0.001], 17.4% [95%CI: 10.2–28.0, P < 0.001], and 19.6% [95%CI: 7.5–42.2, P = 0.012], respectively). Also, in the East, North, and South of Iran, E.coli (32.0% [95%CI: 18.0–50.1, P < 0.001], 34.4% [95%CI: 21.1–90.5, P = 0.009], and 28.8% [95% CI: 4.6–77.2, P = 0.403], respectively) had the highest prevalence rate.

Table 3 Subgroup analysis for the prevalence of Gram-negative bacteria in neonates with sepsis in Iran
Full size table

Subgroup analysis of the prevalence of Gram-negative bacteria cause neonatal sepsis based on hospital

Hospitals exhibited varying patterns of GN bacteria prevalence. The data in Table 3 shows that E.coli (23.3%, [95% CI: 20.6 -28.8, P < 0.001]) and K.pneumonia (20.3%, [95% CI: 15.4–33.7, P < 0.001]) were the most common bacteria found in maternity hospitals. While in the children’s hospitals, K.pneumonia (20.5%, [95% CI: 11.8–33.2, P < 0.001]) followed by Enterobacter (11.6%, [95% CI: 5.9–21.7, P < 0.001]) were more prevalent.

Prevalence of antibiotic resistance in gram-negative bacteria

There was a high level of heterogeneity in antibiotic resistance prevalence among GN bacteria (I2 = 96.18, P < 0.001). Cefixime had the highest resistance rate among third-generation cephalosporins (62.0%, [95% CI: 45.8–75.9, P = 0.146]) as shown in Fig. 4. Ampicillin and Amikacin had the highest resistance rates among penicillin and aminoglycosides, respectively (58.6%, [95% CI: 47.3- 69.0, P = 0.137] and 51.4%, [95% CI: 42.7–60.0, P = 0.616]).

Fig. 4
figure 4

Prevalence of antibiotic resistant in Gram-negative bacteria among neonates with sepsis in Iran

Full size image

Subgroup analysis of the prevalence of antimicrobial resistance based on geographic region

Figure 5 displays the pattern of antibiotic resistance rate in different regions of Iran. Ampicillin was found to have the highest rate of antibiotic resistance among neonates with sepsis in the Center of Iran (72.8%, [95% CI: 58.1–83.7, P = 0.003]). High resistance to Gentamicin (86.7%, [95% CI: 59.8- 96.6, P = 0.013]) was observed in the Eastern region of Iran. Ceftriaxone showed the highest resistance rate in the North, Northwest, and West regions (75.8%, [95% CI: 44.8–92.4, P = 0.098], 57.9% [95% CI: 29.9–81.6, P = 0.593] and 57.7%, [95% CI: 27.8–82.9, P = 0.629], respectively). The South of Iran had the highest resistance to Amikacin at 63.0% [95% CI: 40.4–81.0, P = 0.117]. Imipenem showed the lowest resistance in the Center of Iran (11.9%, [95% CI: 3.9–31.0, P = 0.001]). Both East and West regions exhibited low resistance to Cephalothin (9.7%, [95% CI: 1.6- 41.2, P = 0.017] and 34.7%, [95% CI: 11.9–67.7, P = 0.366]). Gentamicin showed the lowest resistance rate in the North of Iran (27.6%, [95% CI: 10.9–54.4, P = 0.097]). Cotrimoxazole had the lowest resistance in the South (45.1% [95% CI: 20.7–72.1, P = 0.751]). Northwest had the lowest resistance rate for Ciprofloxacin (28.9%, [95% CI: 15.2–48.1, P = 0.032]).

Fig. 5
figure 5

Prevalence of antimicrobial resistance on gram negative bacteria based on geographic region

Full size image

Subgroup analysis of the prevalence of antimicrobial resistance based on the type of bacteria

Cefixime was less effective against K.pneumonia, the most resistant GN bacteria causing neonatal sepsis (80.7%, [95% CI: 56.2–93.2, P = 0.018]). E.coli was more resistant to Ampicillin (61.7%, [95% CI: 44.3–76.5, P = 0.188]), Enterobacter was resistant to Cephalothin (74.2%, [95% CI: 36.6–91.4, P = 0.052]) and Acinetobacter was resistant to Cefotaxime (90.0%, [95% CI: [95% CI: 64.7- 97.8, P = 0.007]). Pseudomonas aeruginosa (P.aeruginosa) was more resistant to Ceftizoxime (94.7%, [95% CI: 79.5–98.8, P < 0.001]). Table 4 displays the antibiotic resistance pattern of two common GN bacteria. Supplementary file 2, Table S1 demonstrates the resistance pattern of other bacteria.

Table 4 Subgroup analysis for the antibiotic resistance pattern among two more prevalent gram-negative bacteria
Full size table

Meta-regression

In Iran, there has been a statistically significant decreasing trend in the prevalence of GN bacteria in neonates with sepsis in recent years, as shown by a meta-regression model that considers the published year of studies (P < 0.001) (Fig. 6). The meta-regression model revealed that Ampicillin resistance has been on the rise in recent years in the Center of Iran (P < 0.001), while Gentamicin resistance has significantly decreased in the Northwest. The other antibiotics did not exhibit a significant trend (P < 0.001).

Fig. 6
figure 6

Meta-regression model for the prevalence of gram negative bacteria in neonates with sepsis according to the published year of studies

Full size image

Publication bias

Based on the funnel plot in Fig. 7 and the results of Egger's test, Publication bias was not observed among the included studies (p = 0.295).

Fig. 7
figure 7

Funnel plot for investigating of publication bias in the included studies

Full size image

Discussion

Our study analyzed the occurrence of GN bacteria and their antibiotic resistance in septic neonates from Iran. Based on the meta-analysis, the occurrence of GN bacteria was found to be 53.6%. Based on the year of studies, the meta-regression model for GN bacteria exhibited a significant decreasing trend. Different studies have reported neonatal sepsis caused by GN agents ranging from 18 to 78% [45,46,47]. In two systematic reviews conducted in Iran in 2020, Akbarian-Rad et al.[8] reported that Enterobacter (23.04%) and K.pneumonia (17.54%) were common neonatal sepsis GN pathogens after combining 22 articles with a sample size of 14,683. In a review of 17 studies (sample size: 89,472), Akya et al. [9] found that K.pneumonia (24.2%) and P.aeruginosa (16.6%) were the main causative pathogens of neonatal sepsis. The results of our meta-analysis of 31 studies with a total of 104,566 Iranian neonates with sepsis showed that K.pneumonia (23.2%) was the most prevalent GN bacteria, followed by E.coli (13.5%). The advantages of this study over previously published meta-analyses include a larger sample size, the use of cross-sectional studies, and the exclusion of studies with samples over 28 days old. These factors, which were not accounted for in previous meta-analyses, can impact the final evaluation and accuracy of prevalence. Our findings are supported by a 2014 systematic review in resource-limited countries, which demonstrated that in Africa, South-East Asia, and the Middle East, K.pneumonia is often the cause of neonatal sepsis more than other pathogens [48]. Moreover, a systematic review carried out in 2021 in developing countries [49] discovered that K.pneumonia (26.36%) and E.coli (15.30%) were the dominant pathogens responsible for neonatal sepsis. Geographical variation in GN bacteria prevalence was observed among Iranian neonates with sepsis through region-based subgroup analysis. The highest prevalence rate of E.coli was found in the East and North of Iran, at 32.0% and 34.4%, respectively. A systematic review and meta-analysis carried out in Iran in 2019 found that the prevalence rates of urinary tract infection (UTI) and asymptomatic bacteriuria (ASB) in pregnant women were 9.8% and 8.7%, respectively [50]. A higher prevalence of UTI and ASB was observed in the North and East of Iran than in other regions. In addition, E.coli was reported as the predominant microorganism involved in UTI (61.6% [95%CI: 51.6–70.7]) and ASB (63.22% [95%CI: 51.2–73.8]). One reason for the alignment of the results of the current study with that study may be the fact that newborns can get gram-negative bacteria from the vaginal fecal flora of the mother and the environment. Differences in socioeconomic factors, quality healthcare, and racial diversity may explain the variation in neonatal GN bacteria prevalence across geographic regions. The prevalence of GN agents in neonatal with sepsis in Iran, based on the type of hospital, shows that E.coli (23.3%) has the highest prevalence in maternity hospitals and K.pneumonia (20.5%) is more prevalent in children’s hospitals. The rate of prevalence of K.pneumonia in children’s hospitals from 26 to 48% has been reported by various authors [51, 52]. Another study reported K.pneumonia as the most frequently isolated pathogen (32.5%) among extramural admissions [53]. K.pneumonia handles a significant proportion of hospital-acquired infections, such as septicemias [51, 53].

WHO recommends Ampicillin-Gentamicin as the first-line treatment for neonatal sepsis in low- and middle-income countries [54]. Ampicillin and aminoglycoside (Amikacin/Gentamicin) are the primary empirical antibiotics for neonatal sepsis in Iranian NICUs [21]. According to our meta-analysis, nearly 54.0% of GN pathogens that were isolated showed resistance to the WHO-recommended first-line antibiotics. Excessive and irrational use of antibiotics in hospitals may be the cause of high resistance in Iran [11]. The findings of this study align with those of other studies when it comes to levels of resistance to first-line antibiotics [55, 56]. In Africa, Asia, and South America, other reports indicate that 50–80% of neonates have a high resistance rate to commonly used antibiotics, like aminoglycosides, cephalosporins, and ampicillin [57,58,59,60,61]. Depending on the region, the resistance pattern in Iran varied. The increased resistance of GN bacteria to Ampicillin in Iran's Center and its upward trend over the past decade highlights the urgency to re-evaluate the current treatment protocols and implement antibiotic stewardship. The resistance to Gentamicin has lowered in Northwest Iran, perhaps because Amikacin is now the preferred first-line treatment. Local prevention policies and clinical management decisions can be influenced by geographical variations. Ampicillin resistance was observed in both E.coli and K.pneumonia in the current study. Germany, China, and Africa also reported similar findings [48, 62, 63]. A United States report found that 67% of E.coli isolates were resistant to Ampicillin and 17% were resistant to aminoglycosides. Additionally, nearly 10% of the isolates were resistant to both Ampicillin and Gentamicin [64]. Another similar report in 2015–2017 in the United States shows 7.8% of neonatal sepsis caused by E.coli in NICU was resistant to both Ampicillin and Gentamicin [65]. According to previous studies, resistance in E.coli and K.pneumoniae is commonly acquired through plasmidmediated extended-spectrum beta-lactamase (ESBL) production, which has been recognized as a significant threat to public health for the past two decades [66, 67]. ESBL-producing multidrug-resistant bacteria cause infections that are resistant to a variety of beta-lactams, such as third-generation cephalosporins [68]. The effectiveness of third-generation cephalosporins as a second-line treatment is still being debated [63]. Our study found a high level of resistance (57.3%) to third-generation cephalosporins. The reviewed articles in this study were laboratory-based, exploring the resistance of GN bacteria to various types of antibiotic discs. According to the results, Cefixime was found to have the highest resistance in K.pneumoniae. In Iran, Cefixime isn't used as a treatment for neonatal sepsis and Cefotaxime is the second-line treatment for sepsis among third-generation cephalosporins. Acinetobacter showed the highest level of resistance to Cefotaxime. Other studies have reported the high resistance of Acinetobacter to Cefotaxime [69, 70]. Antimicrobial resistance patterns in GN bacteria in Iran make it difficult to choose the right antibiotic for initial empirical therapy. In the NICU, selecting the right empirical antibiotics and treatment duration for suspected sepsis has a lot of variation. Recent studies indicate that implementing NICU-specific antimicrobial stewardship programs (ASP) can significantly reduce the use of inappropriate antibiotics [71, 72]. The use of ASP along with suitable antimicrobial treatments can reduce the negative impact caused by antibiotic resistance in newborns.

Excessive use of broad-spectrum antibiotics in NICUs has led to a serious problem of infections caused by multidrug-resistant GN bacteria in some developing countries. Developed countries face this problem with less severity. The occurrence of multidrug-resistant GN bacteria in the present study is akin to that of China and India [63, 73]. Multidrug resistance was found in more than 50% of GN bloodstream isolates in the present study. This level of resistance highlights the significance of GN multidrug resistance in Iranian neonates. Improving infection control strategies should be prioritized. The essential method for preventing GN multidrug resistance colonization and infection is to restrict horizontal transmission. Infection control measures, such as proper hand hygiene, suitable gloving, disinfection, decontamination, and sterilization practices, should be taken seriously. It is important to prevent unit overcrowding and understaffing. NICU-specific ASPs play a crucial role in reducing resistance. Neonatal ESBL bacterial sepsis incidence can be reduced by limiting cephalosporin. Nevertheless, an important challenge is to minimize the use of third-generation cephalosporins and carbapenems. Additional clinical research is urgently required to address these challenges.

In this meta-analysis, most studies did not differentiate between EOS or LOS cases in sepsis. Unfortunately, grouping by sepsis type for analysis was not feasible. The neonates were not classified based on gender, so a detailed analysis could not be conducted. Another limitation of this study was the uneven distribution of samples across the study regions.

The study's findings are crucial for WHO's antibiotic recommendations for neonatal sepsis. Many neonates may not receive sufficient coverage from common first-line and second-line antibiotics. Therefore, these findings can aid in the creation of NICU-specific antibiotic use guidelines.

Conclusion

The study emphasizes that K.pneumoniae and E.coli are the most frequent gram-negative pathogens that cause neonatal sepsis in Iran. The distribution of sepsis-causative pathogens differs among hospitals and regions, as shown in this systematic review. GN bacteria showed the greatest resistance to third-generation cephalosporin and aminoglycosides. Neonatologists in Iranian hospitals should carefully discuss this alarming result and consider changing the treatment regimen if needed.