Data Summary
The authors confirm all supporting data, code and protocols have been provided within the article or through supplementary data files.
Introduction
The Staphylococcus genus is comprised of a diverse group of more than 50 species and subspecies of Gram-positive cocci. Based on the capacity for coagulase enzyme production, the genus is categorized into two main phenotypic groups, coagulase-positive staphylococci and coagulase-negative staphylococci (CoNS) [1, 2]. Initially, this concept was regarded primarily as a classification based on diagnostic procedures (e.g. serum clot formation or PCR using the coagulase gene coa); however, it subsequently evolved into a clinical diagnostic approach to differentiate between the pathogenic species Staphylococcus aureus and a group of staphylococci formerly classified as non-pathogenic [3].
CoNS represent a diverse group of opportunistic pathogens, among which six species are most frequently associated with human disease, namely Staphylococcus epidermidis, Staphylococcus haemolyticus, Staphylococcus saprophyticus, Staphylococcus capitis, Staphylococcus hominis and Staphylococcus lugdunensis [4, 5]. The pathogenic potential of CoNS is primarily related to certain virulence factors, especially biofilms, toxins and antimicrobial resistance [1, 5]. To determine whether CoNS is a true cause of infection rather than a contaminant, clinical and microbiological diagnostic criteria are applied.
These include obtaining two or more positive isolates of the same species from separate longitudinal blood cultures (or distinct sterile sites), clinical symptom improvement following targeted therapy, the presence of an indwelling medical device, a time to positivity of less than 16 h and demonstrated resistance profiles [6]. Although the recovery of two or more isolates of the same CoNS species from separate longitudinal blood cultures or sterile sites is the operational threshold to apply before considering an isolate clinically significant, the presence of a different pathogen in the same specimen does not, by itself, exclude a contributory role for CoNS in true polymicrobial infections; however, a single CoNS isolate obtained in the presence of a primary pathogen is usually regarded as a likely contaminant, and the clinical context should guide interpretation [7].
To date, the most frequent CoNS species associated with human infections and the causative agent of nosocomial bacteraemia is S. epidermidis, due to its ability to form biofilms on the surface of implantable medical devices [8, 9]. For this reason, these bacteria are associated with the use of catheters, prostheses and various polymer and metal implants, which are widely used in the context of modern medicine [10, 11]. Additionally, the colonization of different parts of the skin and mucous membranes of patients by CoNS is a key source of endogenous infections, mainly in newborns, immunocompromised patients and critically ill patients [3, 12, 13].
Notably, CoNS are established as a primary cause of neonatal sepsis, significantly impacting morbidity and mortality, with prevalence ranging from 1.9 to 4.8% in the general neonatal population and reaching 9.4% in very low birth weight infants [14]. Some of the pathologies related to CoNS in this population are retinopathy of prematurity, necrotizing enterocolitis, white matter injury and bronchopulmonary dysplasia, which are mainly caused by S. epidermidis [14].
Antimicrobial resistance constitutes a significant public health burden concerning the management of bacterial infections, primarily associated with healthcare-associated infections (HAIs) [15, 16]. Methicillin was the first semisynthetic penicillin derivative antibiotic effective against S. aureus [17]. However, within 3 years of introduction, the first methicillin-resistant strains were isolated [18, 19].
Resistance to methicillin is due to the presence of mobile genetic elements, namely the staphylococcal cassette chromosome mec (SCCmec) that contains the mecA gene, responsible to produce an altered target penicillin-binding protein (PBP2a) [18, 20]. Treatment of CoNS is becoming increasingly complex due to the emergence of methicillin resistance (MR). In S. epidermidis MR has been detected in 75–90% of all hospital isolates carrying the staphylococcal SCCmec [12, 21].
MR can be detected by disc diffusion test with β-lactam molecules such as oxacillin or cefoxitin [7]. Particularly, cefoxitin has demonstrated itself to be a more sensitive and specific marker of mecA-mediated MR and it was considered the agent of choice for diffusion tests by the Clinical and Laboratory Standards Institute (CLSI), which superseded the former National Committee for Clinical Laboratory Standards [22].
A total of 664 CoNS isolates were obtained from patients with HAIs, predominantly neonates, over an 8-year period. All samples were obtained from the National Institute of Perinatology (INPer) ‘Dr. Isidro Espinoza de los Reyes’, that serves women with high-risk pregnancy.
The primary objective involved species identification and the correlation of cefoxitin disc diffusion assay with PCR-based detection to characterize methicillin-resistant S. epidermidis isolates within this population.
Methods
Isolates
A total of 664 Staphylococcus spp. isolates from patients with HAIs were collected over an 8-year period (March 2006 to November 2014) at INPer ‘Dr. Isidro Espinoza de los Reyes’.
Phenotypic identification and coagulase-negative staphylococci
Species identification of each of the 664 strains was carried out in a VITEK equipment (bioMérieux SA, 376 Chemin de l'Ome, France), following the manufacturer’s instructions. Upon arrival at the Bacterial Pathogenicity Laboratory of the Faculty of Medicine, Universidad Nacional Autónoma de México, each of the strains was cultured in Petri dishes, containing brain heart infusion agar (OXOID Ltd, Basingstoke, Hampshire, England) and mannitol salt agar (MA) (OXOID Ltd, Basingstoke, Hampshire, England). After incubation for 18–24 h at +37 °C in an Isotemp microbiological incubator 750D (Thermo Fisher Scientific, Waltham, Massachusetts, USA), colonies were examined and the morphology was verified after Gram staining (HYCEL de México S.
A. de C. V. Av.
Mexico City). The presumptively identified S. epidermidis and other CoNS colonies were examined for coagulase activity by tube plasma agglutination test (REMEL Coagulase Plasma, Santa Fe Drive, Lenexa, KS, USA). After characterization, the isolated strains were stored in an Ultima II ultra-low freezer, ULT-390–9 A31 (REVCO Technologies Asheville, NC, USA) at −80 °C in a 20% skim milk suspension, until use.
DNA extraction and purification
Genomic DNA was isolated using the Lysostaphin-Lysozyme method of Hookey [23], only modifying the lysis step. Individual colonies were used to inoculate tryptic soy broth and grown overnight, after which genomic DNA was extracted. Briefly, bacterial cells were resuspended in 250 µl of lysis solution (lysozyme 50 mg ml−1; lysostaphin 1 mg ml−1 in 25 mM Tris/HCl, 10 mM EDTA; Sigma-Aldrich, St.
Louis, MO, USA) and incubated at 37 °C for 90 min. The sample was then warmed to 95 °C for 10 min and centrifuged at 8,000 g. The lysate was extracted with equal volumes of phenol/chloroform, and the DNA was precipitated from the aqueous phase with 2-propanol, resuspended in 100 µl of sterile distilled water and stored at 4 °C.
The integrity of the DNA preparations was determined by gel electrophoresis (5 µl aliquots) and spectrophotometry at A260 nm.
Duplex PCR for coa and mecA genes
Genomic DNA was extracted from all isolates, and a duplex PCR procedure was performed for the simultaneous detection of the mecA and coa genes, following the protocol described by Hookey et al. [23]. Identification of S. aureus was determined using the coa gene as a specific marker, whereas MR was genetically confirmed through the presence of the mecA gene. The primers utilized, as previously validated in our laboratory [24], included: for coa, F-5′-AACTTGAAAT-AAAACCACAAGG-3′ and R-5′-TACCTGTACCAGCATCTCTA-3′ (yielding a 177 bp amplicon); and for mecA, F-5′-ATGGCAAAGATATTCAACTAAC-3′ and R-5′-GAGTGCTACTCTAGCAAAGA-3′ (yielding a 458 bp amplicon).
PCR was performed on a Bio-Rad C1000 thermal cycler (Bio-Rad Laboratories, Inc. Hercules, California, USA), and the amplification conditions were as follows: initial denaturation at 94 °C for 4 min and 30 cycles at 94 °C for 30 s, 56.2 °C for 30 s and 72 °C for 30 s; finally, an extension at 72 °C for 10 min. PCR products were visualized on 1% agarose gel, using a TM-10E UV transilluminator (UVP Upland, CA 91786, USA).
Disk diffusion test for cefoxitin
In a 13×100 mm test tube with 3 ml of sterile 0.85% saline, a bacterial suspension adjusted to the 0.5 McFarland standard was prepared using a DENSIMAT equipment (bioMérieux SA, Marcy-l'Étoile, France). A sterile rayon swab was then moistened with the suspension, and the inoculum was distributed homogeneously on a plate of Mueller–Hinton agar (OXOID Ltd, Basingstoke, Hampshire, England). A 30 µg cefoxitin disc (OXOID Ltd, Basingstoke, Hampshire, England) was placed, and the plate was incubated in an Isotemp 750D microbiological incubator (ThermoFisher Scientific, Waltham, Massachusetts, USA) at 35 °C for 24 h.
After that time, the diameter of the inhibition halo was measured using a vernier device. A diameter of ≤24 mm was interpreted as MR [6].
Oxacillin screen agar
Mueller–Hinton agar (OXOID Ltd, Basingstoke, Hampshire, England) plates containing 4% NaCl (Sigma-Aldrich, St. Louis, MO, USA) and 6 µg ml−1 of oxacillin (Sigma-Aldrich, St. Louis, MO, USA) were prepared.
Plates were inoculated with 10 µl of 0.5 McFarland suspension of the isolate by streaking in one quadrant and incubated at 35 °C for 24 h. Plates were carefully observed under transmitted light; any growth after 24 h was considered oxacillin-resistant.
Disk diffusion D-zone test
For the detection of inducible resistance to clindamycin, erythromycin (15 µg) and clindamycin (2 µg) discs (OXOID Ltd, Basingstoke, Hampshire, England) were placed on Mueller–Hinton agar (OXOID Ltd, Basingstoke, Hampshire, England) 15 mm apart. After a 24 h incubation at 35 °C, a positive test (iMLS_B phenotype) was defined as the flattening of the clindamycin inhibition zone in the area adjacent to the erythromycin disc, forming a ‘D’ shape. In contrast, the constitutive phenotype (cMLS_B) was identified in isolates showing resistance to both erythromycin and clindamycin with no inhibition zones around the discs.
Finally, the MS phenotype was defined as resistance to erythromycin with a circular, unaffected zone of inhibition around the clindamycin disc (negative D-test).
Statistical analysis
Data obtained from phenotypic susceptibility tests [oxacillin agar dilution (AD) and cefoxitin disc diffusion] and genotypic detection (mecA) were collected in a Microsoft Excel database and processed using the statistical software GraphPad Prism v5.0. Absolute frequencies and percentages were used to characterize mecA gene prevalence and resistance profiles in neonatal and adult populations. To determine the relationship between the presence of the gene and phenotypic resistance, Pearson’s Chi-square (X2) test was applied; when expected cell counts were <5, Fisher’s Exact Test was used to ensure statistical rigour.
A value of P<0.05 was considered statistically significant. Sensitivity (S) and Specificity (SP) were calculated, using PCR results as the reference (‘gold’ standard). Confidence intervals were established at 95% (95% CI).
The degree of agreement between genotype and phenotype was quantified using Cohen’s Kappa coefficient (κ) with 95% CI. κ values were interpreted under the categorical framework originally proposed by Landis and Koch and restated in McHugh, in which κ<0.00 indicates agreement worse than chance; 0.00–0.20, slight; 0.21–0.40, fair; 0.41–0.60 moderate; 0.61–0.80, substantial; and 0.81–1.00, almost perfect [25, 26].
Results
Population characterization and clinical origin
Over an 8-year period (March 2006 to November 2014), 664 strains of Staphylococcus were obtained from the INPer ‘Isidro Espinoza de los Reyes’, which is a tertiary level hospital where high-risk pregnant women, patients with gynaecological problems, premature newborns and gestational pregnancy loss are managed.
From the samples, 552 isolates (83.1%) were obtained from newborns, while 112 (16.9%) were from adult women. Regarding the level of care for newborns, most isolates (524; 94.9%) were obtained from those in the Intensive Care Unit (ICU), with the remaining 28 (5.1%) originating from the general inpatient wards. Of the 112 adult samples, 9 (8.0%) were from patients in the ICU, 87 (77.7%) from general inpatients and 16 (14.3%) from outpatients.
The site of infection varied by patient age (Table 1). For neonates, most isolates were obtained from catheters (237; 42.9%) and blood (198; 35.8%). In contrast, adult isolates were primarily recovered from surgical wounds (23; 23.2%), blood (14; 13.4%) and urine (14; 13.4%) (Table 1).
The 664 samples were analysed using the VITEK 2 system according to the manufacturer’s instructions. All isolates belonged to the Staphylococcus genus, with S. epidermidis being the most prevalent species (579 isolates; 87.2%). Of these, 93 (16.1%) were from adults and 486 (83.9%) were from newborns.
The second most representative species was S. hominis with 57 isolates (8.6%), comprising 9 (15.8%) from adults and 48 (84.2%) from newborns. The remaining isolates consisted of other CoNS species. Further phenotypic validation of these isolates was conducted using phenotypic tests, including growth on mannitol salt agar and catalase and coagulase testing.
Additionally, genomic DNA was extracted from the isolates for PCR analysis of the coa gene. Molecular analysis confirmed that 100% of the isolates were identified as CoNS, as evidenced by a negative result for the coa gene.
Genotype–phenotype correlation for the detection of mecA
To identify the genetic determinants of MR, 579 S. epidermidis isolates and 57 S. hominis isolates were analysed for the presence of the mecA gene by PCR. In S. epidermidis, the mecA gene prevalence was 80% (n=389) in the neonatal population and 64.5% (n=60) in adults (Table 1). Validation of phenotypic susceptibility tests against the gold standard (PCR) demonstrated performance stratification depending on the test and population studied.
Oxacillin [determined by MIC via AD at the CLSI breakpoint (≥0.5 mg l−1)] was the most sensitive marker in neonates (87.1%, 95% CI: 83.5–90.1%), surpassing cefoxitin (disc diffusion, ≤24 mm), which reached 72.2% (95% CI: 67.5–76.6%) in the same neonatal cohort (Table 2). However, both markers exhibited limited specificity (56.7% for oxacillin and 55.7% for cefoxitin), indicating that both phenotypic tests generate a substantial proportion of false-positive results within the analysed isolate cohort. Chi-square (X2) analysis confirmed that oxacillin is associated with the presence of the mecA gene in neonates (X2=88.13; P<0.001) and in adults (X2=28.53; P<0.001) (Table 3).
Conversely, cefoxitin was also significantly associated with mecA in this same adult subgroup; the markedly lower X2 (5.97, P=0.0145) than in the neonatal cohort (X2=27.21, P<0.001) reflects the reduced statistical power associated with the smaller sample size (n=93 versus n=486). Cohen’s kappa index (κ) further contextualized these observations: oxacillin showed moderate agreement with the PCR reference in both S. epidermidis subgroups (κ=0.44 in neonates and κ=0.54 in adults), whereas cefoxitin showed only fair agreement with the PCR reference in both S. epidermidis subgroups (κ=0.22 in neonates; κ=0.21 in adults), supporting the position that cefoxitin DD is unsuitable as a standalone screening marker for mecA in this setting (Table 3).
For S. hominis, 57 isolates were analysed, consisting of 48 (84.2%) from neonates and 9 (15.8%) from adults. The mecA gene was detected in 72.9% (n=35) in the neonatal group and 33.3% (n=3) in the adult population (Table 2). Phenotypic marker performance showed inconsistent diagnostic behaviour, differing markedly from the patterns observed in S. epidermidis.
Among neonates (n=48), cefoxitin demonstrated higher sensitivity than oxacillin (80 versus 71.4%), although specificity for both markers remained limited (61.5% for cefoxitin and 84.6% for oxacillin), reflecting the small mecA negative denominator (n=13) in this subgroup. These specificity values, together with the κ findings reported, point to high rates of phenotypic-method disagreement in S. hominis neonates and limit the operational value of either method as a standalone screening modality in this species (Table 3). In the adult subgroup (n=9), the small sample size prevented precise estimation: cefoxitin exhibited 100% sensitivity but 66.7% specificity, suggesting a tendency toward overestimation of MR from disc diffusion results in this dataset.
Due to the distribution of the data, Fisher’s exact test was applied, revealing only a weak statistical association. In neonates, both phenotypic markers reached strong statistical association with the mecA gene by Fisher’s exact test: oxacillin (P=0.0077) and cefoxitin (P=0.0098), both values falling below the P<0.05 threshold defined a priori. In the adult population, the limited number of positive cases meant that Fisher’s exact test failed to yield statistical significance for either phenotypic test (cefoxitin P=0.119; oxacillin P=0.6353).
Concordance assessments further underscored the unreliability of conventional phenotypic methods for S. hominis. In neonates, κ values classified as fair agreement for cefoxitin (κ=0.39) and moderate agreement for oxacillin (κ=0.48), with the higher κ for oxacillin reflecting its superior specificity (84.6%) despite its lower sensitivity (71.4%). In the adult subgroup (n=9; 3 mecA positive), the small sample size prevented reliable concordance estimation; oxacillin returned κ=0.14 (slight agreement) and cefoxitin returned κ=0.60 (moderate agreement).
These estimates, particularly those derived from the nine-isolate adult subgroup, preclude extrapolation to species-level evidence and reflect the limitations of phenotypic reliability at the current sample size.
Multi-resistance profile and macrolide and lincosamide resistance phenotypes
For the characterization of the macrolide–lincosamide resistance landscape within the CoNS collection, two complementary dimensions were investigated: the rates of erythromycin and clindamycin resistance stratified by mecA status, and the distribution of inducible, constitutive and MS phenotypes among isolates demonstrating resistance to either or both markers. The first dimension is summarized graphically in Fig. 1 and reported in full, with the row-level counts and percentages for each strata, in Table S1; the second is detailed in Table 4.
In S. epidermidis, examination of the mecA-stratified resistance landscape revealed that erythromycin resistance was elevated across all strata, peaking at 88.4% in mecA+ neonates and declining stepwise through mecA− neonates (68.0%), mecA+ adults (63.1%) and mecA− adults (57.5%), exceeding 75% only in the mecA+ neonatal subgroup. Clindamycin resistance tracked more closely with mecA carriage: in the neonatal cohort, clindamycin resistance reached 79.4% in mecA+ isolates but only 52.6% in mecA− isolates, a 26.8-percentage-point gap consistent with the persistence of erm-mediated resistance in multidrug-selected Neonatal Intensive Care Unit (NICU) clones. In adults, both mecA strata returned lower and more convergent clindamycin resistance values (48.3% in mecA+; 42.4% in mecA−), consistent with reduced selective pressure outside the NICU setting (Fig.
1) .
In S. hominis, a departure from the S. epidermidis pattern is observed in the same Fig. 1 and Table S1 comparison. Clindamycin resistance remained elevated in mecA+ neonates (65.7%) but dropped sharply in mecA− neonates (38.4%), an inversion of the gradient observed in S. epidermidis, where clindamycin resistance was higher in mecA++ isolates and consistent with the broader observation that S. hominis carries clindamycin-resistance determinants independently of the SCCmec cassette.
Among the small adult subgroup (n=9), clindamycin resistance was elevated in both mecA strata (66.7% in each); since the adult subgroup comprises only three mecA++ and six mecA− isolates, these percentages should be interpreted as point estimates rather than species-level evidence.
A complementary phenotype distribution was observed. In neonatal S. epidermidis, the constitutive phenotype (cMLS_B) dominated, affecting 73.7% (n=358) of isolates; the inducible phenotype (iMLS_B) was rare at 1.0% (n=5) and the macrolide-susceptible MS phenotype accounted for 9.7% (n=47). Among adults, cMLS_B remained the most frequent pattern but at the lower rate of 43.0% (n=40); the MS phenotype was relatively more frequent in adults (17.2%, n=16) than in neonates (9.7%, n=47), mirroring the lower overall multidrug-resistance burden in non-ICU settings.
In S. hominis, the phenotype distribution was distinctive in two respects. The inducible phenotype (iMLS_B) was substantially more frequent in S. hominis than in S. epidermidis across both cohorts (20.8%, n=10 in neonates; 22.2%, n=2 in adults). When iMLS_B and cMLS_B were combined, 70.8% (34 of 48) of neonatal S. hominis isolates carried a clindamycin-resistance mechanism, 20.8% iMLS_B (n=10) plus 50% cMLS_B (n=24), that could compromise empirical clindamycin therapy.
This rate is comparable to 74.7% seen in neonatal S. epidermidis (n=363 of 486 isolates; 1.0% iMLS_B, n=5, plus 73.7% cMLS_B, n=358), but with a markedly higher contribution from the inducible phenotype in S. hominis (20.8%, n=10) than in S. epidermidis (1.0%, n=5).
In conjunction, Fig. 1, Tables 4 and S1 delineate a two-axis resistance landscape wherein (i) phenotypic resistance rates and the mecA genotype do not coincide uniformly across species and cohorts; (ii) inducible clindamycin resistance, remaining mechanistically obscured by conventional antibiograms, is concentrated in S. hominis and concentrated further in the NICU setting; and (iii) the high cumulative clindamycin-therapy-invalidation rate in neonatal S. hominis (70.8%) constitutes the operational justification for systematic D-test screening of erythromycin-resistant CoNS isolates in this population.
Discussion
While oxacillin AD, cefoxitin disc diffusion and mecA PCR are long-established techniques, well documented in reference-centre settings, phenotypic susceptibility testing remains the operational standard in Mexican governmental hospitals, as the requisite technical infrastructure (e.g. equipment and expertise) for routine molecular confirmation is not uniformly available at the secondary-care level [27]. Under these conditions, the application of validated breakpoints to a Mexican public-sector CoNS population has not previously been combined with species-stratified analysis across S. epidermidis and S. hominis together with formal phenotype–genotype concordance analysis on both a neonatal and an adult cohort. The closest Mexican precedent combined cefoxitin disc testing with mecA PCR in a tertiary-care setting under a hospital-surveillance design [28] and, more recently, on a smaller mixed-species collection [29]; neither analysed Cohen’s κ concordance, neither separated neonates from adults and neither stratified by CoNS species.
The present study contributes, therefore, a region-specific, species-stratified validation of an established method-pair in a fixed-resource public-sector setting.
CoNS, historically considered contaminants of clinical specimens, have emerged in recent decades as the main opportunistic pathogens in the modern hospital environment [30]. This paradigm shift is primarily attributed to the increased utilization of invasive medical devices and the biofilm-forming capabilities of these species, which confer protection against host immune responses and antimicrobial agents [3]. Currently, CoNS represent the most frequent cause of bloodstream infections and late-onset sepsis in vulnerable populations, especially in preterm infants and immunocompromised patients [31].
Driven by this global epidemiological shift towards opportunistic hospital-acquired infections, NICUs function as unique ecological niches [32]. The prevalence of the mecA gene in neonates was determined to be 80% in S. epidermidis and 72.9% in S. hominis (35/48 isolates), reflecting the dissemination of multidrug-resistant endemic clones adapted to the selective pressure of the hospital environment [30, 32]. The clinical relevance of these isolates is supported by their anatomical origin: among neonates, 46.1% of S. epidermidis were recovered from catheters and 62.5% of S. hominis from blood, with a worrying 16.7% in CSF.
These findings confirm the role of these isolates as true pathogens in severe invasive infections indicative of invasive pathogenic potential rather than commensal skin colonization. In contrast, the adult population exhibited greater diversity in the origin of the samples, with a significant presence in surgical wounds (28%) and urine (16.1%). This heterogeneity suggests a distinct epidemiology, where isolates are exposed to intermittent antibiotic pressures [33, 34].
This variability in sample origin is consistent with the observation that statistical agreement was robust in neonates but null in specific adult patient subgroups, supporting the interpretation that the clinical environment directly influences the stability of resistance-genotype expression at the phenotypic level. This conceptual framing aligns with the characterization of NICUs as continuous microbial reservoirs in which environmental and infant-gut microbiomes overlap [35], and it is reinforced by recent evidence of a high multidrug-resistance burden in S. hominis collections, which positions S. hominis as a clinically significant opportunistic reservoir in its own right [36]
Cefoxitin is a substantially more efficient mecR1 inducer than oxacillin, which theoretically favours detection of MR by forcing expression of the mecA gene. Within the collection, however, this theoretical advantage did not translate into uniformly higher phenotypic agreement with the PCR. In S. epidermidis oxacillin by AD with 4% NaCl and 24 h of incubation returned higher κ than cefoxitin across both neonatal (κ=0.44 versus κ=0.22) and adult (κ=0.54 versus κ=0.21); all values fell within the ‘moderate’ (κ=0.41–0.60) or ‘fair’ (κ=0.21–0.40) categories of the Landis and Koch scale [25, 26].
Although McHugh explicitly notes that ‘moderate’ agreement is enough for clinical concordance, it is not a regulatory-grade basis for substituting one phenotypic method for another. The higher κ observed for oxacillin is therefore best interpreted as a gradient of concordance rather than as a generalizable clinical recommendation. This pattern aligns with data from prior reports suggesting that standardized 4% NaCl/24 h incubation conditions optimize the detection of PBP2a-mediated resistance [6].
For S. hominis, the comparison was constrained by an unequally distributed cohort (n=48 neonates and n=9 adults). In neonates, the two phenotypic markers showed complementary error profiles: cefoxitin had high sensitivity (80%) but limited specificity (61.50%), whereas oxacillin showed limited sensitivity (71.4%) but appreciably higher specificity (84.6%). Under the Landis and Koch scale, the κ values for neonatal S. hominis classified agreement as fair for cefoxitin (κ=0.39) and moderate for oxacillin (κ=0.48), with the higher κ for oxacillin reflecting its superior specificity (84.6%) despite its lower sensitivity (71.4%).
In the small adult subgroup (n=9), oxacillin yielded κ=0.14, corresponding to slight agreement under the scale, whereas cefoxitin yielded κ=0.60, which on the Landis and Koch scale is classifiable as moderate agreement. However, given the inclusion of only nine isolates, this point estimate is not interpreted as a robust species-level claim, consistent with McHugh’s caution that diagnostic inferences should not rest on a single κ value derived from limited sample sizes.
The χ² or Fisher’s exact tests reported determined whether each phenotypic marker was individually associated with the mecA gene, not whether the two markers differed in sensitivity; a formal paired sensitivity comparison between the two phenotypic methods would require McNemar’s test on discordant pairs, which was not performed. Compared with regional CoNS surveys in which phenotypic tests were evaluated without κ concordance and without NICU–adult stratification [28, 37], the present dataset provides a species-resolved, age-stratified concordance analysis on a fixed-resource public-sector collection. The discrepancy between these species-specific findings and the more favourable cefoxitin performance reported in some Latin American surveys may plausibly reflect the inclusion in our collection of CoNS isolates with very low or heterogeneous levels of PBP2a expression, in which the standard cefoxitin 30 µg disc and the 24 mm zone-diameter breakpoint may be insufficient to push the measured zone below the resistance threshold.
The observation that phenotypic detection findings for S. epidermidis show higher κ agreement for oxacillin than for cefoxitin sits alongside the broader observation from Fariña et al. in a regional cohort that biofilm production is more frequent in methicillin-resistant isolates, with both findings underscoring the clinical importance of identifying mecA carriage as accurately as possible in any CoNS isolate [38].
When erythromycin resistance rates are stratified by mecA status, species and cohort, the resistance landscape remains elevated across all categories, with the highest macrolide resistance in neonatal S. epidermidis (84.4%; Table 4). The cMLS_B phenotype alone reached 73.7% in neonatal S. epidermidis, exceeding the 46% reported by Gatermann et al. in 2007 and aligning with the upward trend documented by Szczuka et al. in clinical S. hominis isolates; these figures confirm a sustained global increase in macrolide resistance over the past two decades [39, 40].
Regarding clindamycin, a divergent resistance landscape between species is observed. In S. hominis, the 20.8% prevalence of the inducible phenotype (iMLS_B) underscores a silent clinical risk: an antibiogram that reports clindamycin as susceptible on an erythromycin-resistant S. hominis isolate may produce therapeutic failure. In contrast, S. epidermidis showed an overwhelming predominance of the constitutive mechanism (cMLS_B) at 73.7% in neonates, compared to only 43.0% in adults.
The ~30-percentage-point gradient is consistent with the hypothesis of stronger selective pressure for multidrug-resistant clones in NICU settings relative to general wards. Interestingly, the MS phenotype was more prevalent in the adult population (17.2%) than in neonates (9.7%). The 17.2% versus 9.7% difference is consistent in direction with the hypothesis that in non-ICU environments, resistance may be mediated more frequently by msr-type efflux pumps, although this inference requires further molecular confirmation (Fig.
1).
The therapeutic complexity is compounded by the observation that cMLS_B and iMLS_B phenotypes were detected in some mecAnegative isolates. The apparent ‘independence’ documented is a descriptive case-level observation rather than a tested population-level association. Mechanistically, the co-occurrence of mecA carriage and erythromycin/clindamycin resistance in NICU isolates without physical co-carriage on a single cassette is expected: erythromycin and clindamycin resistance in staphylococci is encoded by ermA, ermB and ermC, which are typically carried on transposons such as Tn551/Tn917 or on small plasmids rather than within the SCCmec element itself [41].
The NICU co-resistance profile is therefore consistent with co-selection under sustained multi-class antibiotic pressure rather than with any direct genetic-linkage mechanism, and the observation reinforces the role of the NICU as a high-intensity reservoir for combined resistance.
Detection in S. hominis neonates, of 20.8% inducible clindamycin resistance (iMLS_B), coupled with 50.0% constitutive resistance (cMLS_B), implying that 70.8% of isolates carry a mechanism that would invalidate clindamycin therapy regardless of any susceptibility reported by conventional antibiogram (Table 4; Fig. 1). Combined with the small adult subgroup (n=9) showing 22.2% (n=2) iMLS_B and 11.1% (n=1) MS, these values should be interpreted with substantial caution given the very small sample size.
The data position S. hominis as a particularly unpredictable resistance reservoir in the NICU setting and support the practical recommendation that clindamycin should not be prescribed empirically in this species without prior D-test screening on erythromycin-resistant isolates or, where available, PCR confirmation of ermA/B/C carriage.
Despite the methodological soundness and large size of the neonatal sample, limitations inherent to the study design must be noted. First, the numerical disparity between the neonatal and adult populations analysed (N=534 versus N=102, comprising 486 versus 93 S. epidermidis and 48 versus 9 S. hominis) limited the statistical power in the latter subgroup, as reflected in the elevated χ² P-values and the weak κ indices observed in some adult subgroups with limited sample sizes (e.g. in S. hominis isolates from adults, the χ²/Fisher test for cefoxitin returned P=0.119 with κ=0.60, and for oxacillin P=0.6353 with κ=0.14). Second, the absence of molecular typing by pulsed-field gel electrophoresis or whole-genome sequencing (WGS) prevented confirmation of the clonal structure suggested by the multi-resistance profiles; the co-resistance profiles observed cannot be interpreted as evidence of physical co-carriage of resistance genes on a single cassette, and WGS data would be required to resolve this structural question.
This consideration reinforces the corollary that, because ermA, ermB and ermC genes are typically carried on transposons such as Tn551/Tn917 or on small plasmids rather than within the SCCmec element itself, phenotypic co-resistance in NICU isolates is most plausibly explained by co-selection under sustained antibiotic pressure rather than by genetic linkage within a single cassette. Focusing exclusively on the mecA gene also left other less frequent determinants, such as mecC or mutations in native PBPs, out of analysis. Finally, the single-centre nature of the study implies that the results reflect selective pressure and local antimicrobial use policies; therefore, extrapolation of these findings to other hospitals should be carried out with caution.
Operationally, the management of CoNS infections in the NICU requires a multifaceted approach. Although oxacillin, when assessed by the reference method, proved to be an acceptable phenotypic predictor in the study cohort, the invasive-disease profile of S. hominis in neonates, with 62.5% of isolates recovered from blood and 16.7% from CSF, together with the 20.8% iMLS_B prevalence in neonatal S. hominis isolates cited above, makes systematic D-test screening of erythromycin-resistant isolates and, where available, mecA PCR confirmation necessary in this group [40]; the integration of sample-origin information (catheter, blood, CSF), D-test screening on every erythromycin-resistant CoNS isolate and molecular confirmation where available can support safe antimicrobial therapy in the neonatal population [42].
Data presented herein demonstrate that oxacillin AD maintains clinically usable diagnostic validity as a frontline phenotypic test in the NICU setting; at the same time, the inducibility of mecA expression by the mecR1/mecI regulatory system, combined with the documented diversity of CoNS species, suggests that no single phenotypic method can be relied upon across the full neonatal CoNS population, making species-stratified surveillance a suitable response to the continuing evolution of multiresistance phenotypes [43].
Supplementary material