Distinguishing Kingella kingae from Pyogenic Acute Septic Arthritis in Young Portuguese Children

(1) Background: We aim to identify clinical and laboratorial parameters to distinguish Kingella kingae from pyogenic septic arthritis (SA). (2) Methods: A longitudinal, observational, single-centre study of children < 5 years old with microbiological positive SA admitted to a paediatric hospital from 2013–2020 was performed. Clinical and laboratorial data at admission and at 48 h, as well as on treatment and evolution, were obtained. (3) Results: We found a total of 75 children, 44 with K. kingae and 31 with pyogenic infections (mostly MSSA, S. pneumoniae and S. pyogenes). K. kingae affected younger children with low or absent fever, low inflammatory markers and a favourable prognosis. In the univariate analyses, fever, septic look, CRP and ESR at admission and CRP at 48 h were significantly lower in K. kingae SA. In the multivariate analyses, age > 6 months ≤ 2 years, apyrexy and CRP ≤ 100 mg/L were significative, with an overall predictive positive value of 86.5%, and 88.4% for K. kingae. For this model, ROC curves were capable of differentiating (AUC 0.861, 95% CI 0.767–0.955) K. kingae SA from typical pathogens. (4) Conclusions: Age > 6 months ≤ 2 years, apyrexy and PCR ≤ 100 mg/L were the main predictive factors to distinguish K. kingae from pyogenic SA < 5 years. These data need to be validated in a larger study.


Introduction
Kingella kingae, a Gram-negative aerobic coccobacillus, was first described in the 1960s by Elizabeth King [1]. Some strains can cause invasive disease, and in recent years K. kingae has emerged as an important cause of septic arthritis (SA) in children younger than 4 years, ranging from 30% to 93.8% of cases [2][3][4][5][6], mostly due to improved culture techniques and molecular detection methods [7].
K. kingae infections usually affect children between 6 to 48 months, are usually milder, have a different microbiological and clinical profile, and require a less aggressive management [1,[7][8][9]. In contrast, SA due to pyogenic infections, such as Staphylococcus aureus or Streptococcus spp., or less commonly due to Gram-negative enteric microorganisms, as occurs more often in older children and adults, is usually associated with high fever and inflammatory parameters, longer days of intravenous (IV) treatment and length of stay (LOS), and has a worse prognosis [2,9,10]. Joint drainage and irrigation are still the standard of care in major joint infections [10,11]. However, the antibiotic choice is a matter of discussion. In children younger than 4 years, a first-or second-generation cephalosporin is recommended to cover for K. kingae and also S. aureus, as oxacillin does not provide adequate coverage against K. kingae. In older children, oxacillin is a good option in countries with a low prevalence of methicillin-resistant S. aureus (MRSA), such as Portugal [10][11][12][13][14]. It is thus important to early distinguish K. kingae from other SA infections, to guide management and initial antibiotic therapy [2].
Several studies compare the clinical and laboratorial characteristics between K. kingae and pyogenic infections, and although differences were noted, few have suggested initial distinguishing features [8,15]. Ceroni et al. [15], in a retrospective study, proposed that fever, CRP above 55 mg/L, leucocytosis above 14,000 cells/mm 3 and neutrophil band shift ≥ 150 cells/mm 3 were able to distinguish K. kingae from other SA infections. However, this model is still a matter of controversy, lacking validation by other authors [16].
The aim of this study was to compare the signs and symptoms of SA in children under 5 years old caused by K. kingae and typical pathogens and to establish predictive parameters to allow the differential diagnosis.

Materials and Methods
We reviewed all children aged less than 5 years old with microbiological positive SA admitted to a Lisbon paediatric hospital. Clinical, microbiological and imaging data, treatment, complications and sequelae were collected from January 2003 to December 2020. Children less than 3 months of age or that developed infection after surgery, orthopaedic hardware implementation or open trauma were excluded.
Acute SA was defined based on clinical complaints (local pain, swelling, decreased range of movement and imaging findings suggestive of infection) when the duration of symptoms was ≤ 14 days. Only children with a positive culture or molecular bacterial identification were included in this study [10]. Complications and sequelae were considered as previously defined [6].
Microorganism identification and antimicrobial susceptibility were determined by the local microbiology laboratory. Real-time PCR for K. kingae was implemented in 2014, using a primer that targeted the rtxA gene [17]. Children with arthritis due to K. kingae and typical pathogens were compared by bivariate analysis. A multivariate logistic regression was performed to determine which variables at admission best predicted K. kingae and pyogenic infections. K. kingae diagnostic accuracy was analysed by the area under the ROC curve (AUC) and a cut-off AUC > 0.8 was considered discriminative. All analyses were performed in SPSS Statistics ® version 27 (IBM Corp, New York, NY, USA). The study was subject to approval by our Hospital Ethics Committee (EC70-2011, 23 April 2011).

Kingella kangas
A total of 44 cases of K. kingae SA were identified (Table 1), all in children above 6 months and under 4 years, and only 3 (6.8%) were older than 36 months. There were three to eight cases per year, with 59.1% between November and February. Most children (61.4%) reported a preceding respiratory tract infection, suggestive of hand, foot and mouth disease in seven (15.9%), and only 6.8% described previous trauma. Two children were from the same kindergarten, but no additional investigation was undertaken. None had chronic disease.

Comparing Data between K. kingae and Pyogenic SA
Comparing data between K. kingae and pyogenic SA on bivariate analyses, fever at admission, fever duration, fever at 48 h and septic look, CRP and ESR were significantly lower in K. kingae infections (Table 1). In addition, the LOS and duration of treatment were also inferior in the K. kingae group.
We observed that only 37.2% of children with K. kingae SA had fever, CRP was higher than 55 mg/L in 34.1% of cases and WBC was elevated in 40%. In comparison, 83.9% of cases with SA due to typical pathogens had fever, CRP level was above 55 mg/L in 80.6% and WBC was elevated in 54.8% of cases. In our cohort, the predictive value for K. kingae accounting for WBC count < 14,000 cells/mm 3 was 63.2% and for CRP < 55 mg/L 82.9%. Applying the Ceroni Score to our cohort, the sensibility (less than two criteria) for K. kingae was low (65%), with a good positive predictive value of 86.7%, and for typical pathogens (considering ≥ 2 criteria, as band test was not performed) a better sensibility (87.1%) but a low PPV (65.9%). On multivariate analyses, our best model to distinguish K. kingae from pyogenic infections at admission was age above 6 months and ≤2 years, apyrexia and CRP < 100 mg/L, with an overall PPV of 86.5%, 88.4% for K. kingae and 83.9% for pyogenic infections ( Table 2). For this model, AUC assessed by ROC curves was capable of differentiating (AUC 0.861, 95% CI 0.767-0.955) K. kingae arthritis from typical pathogens (Figure 1).

Comparing Data between K. kingae and Pyogenic SA
Comparing data between K. kingae and pyogenic SA on bivariate analyses, fever at admission, fever duration, fever at 48 h and septic look, CRP and ESR were significantly lower in K. kingae infections (Table 1). In addition, the LOS and duration of treatment were also inferior in the K. kingae group.
We observed that only 37.2% of children with K. kingae SA had fever, CRP was higher than 55 mg/L in 34.1% of cases and WBC was elevated in 40%. In comparison, 83.9% of cases with SA due to typical pathogens had fever, CRP level was above 55 mg/L in 80.6% and WBC was elevated in 54.8% of cases. In our cohort, the predictive value for K. kingae accounting for WBC count < 14,000 cells/mm 3 was 63.2% and for CRP < 55 mg/L 82.9%. Applying the Ceroni Score to our cohort, the sensibility (less than two criteria) for K. kingae was low (65%), with a good positive predictive value of 86.7%, and for typical pathogens (considering ≥ 2 criteria, as band test was not performed) a better sensibility (87.1%) but a low PPV (65.9%). On multivariate analyses, our best model to distinguish K. kingae from pyogenic infections at admission was age above 6 months and ≤ 2 years, apyrexia and CRP < 100 mg/L, with an overall PPV of 86.5%, 88.4% for K. kingae and 83.9% for pyogenic infections ( Table 2). For this model, AUC assessed by ROC curves was capable of differentiating (AUC 0.861, 95% CI 0.767-0.955) K. kingae arthritis from typical pathogens (Figure 1).
Although in our series diagnosis was only presumed by oropharyngeal K. kingae positivity in eight (18.2%) cases, the predictive positive value of the OPS in young children with osteoarticular infections is very high (91%), making this diagnosis highly probable [19,20]. Furthermore, genotyping of oropharyngeal K. kingae without prior culture indicates that these strains matched the most frequent invasive strains [21].
In our study, several common K. kingae characteristics can be identified, such as early childhood affection (average 15.3 months, with only 6.8% children older than 36 months), previous upper respiratory symptoms, fall predominance, clinical paucity, low or absent fever, low inflammatory markers and a favourable prognosis [1,3]. All were milder and monoarticular, affecting more commonly the knee, hip and ankle. This low virulence has been evidenced by shorter length of stay, fewer adverse events and a better outcome than pyogenic infections [2,7,15,16], as documented in our series. The number of surgical interventions was similar in both groups, but K. kingae infections had shorter duration of treatment and LOS and no admissions to the ICU, confirming the milder nature of these infections. However, 15.9% needed a second intervention, due to persistent effusion, and complications were reported in 11.4%. This is different than that reported by Basmaci et al. [8], with only 1/64 (1,6%) complicated infections that needed a second surgical drainage.
Basmaci and Ceroni have previously characterised the minor increase in biological markers in K. kingae arthritis [15,16]. In our study, a lower CRP and ESR were also suggestive of K. kingae, [15,22] being less than 100 mg/L in about 88.6% of cases in our cohort, although it could not eliminate classic pathogens, as already suggested [16]. The duration of fever is longer in classic, pyogenic pathogens, with about a third still febrile for more than 48 h after treatment [8,16,23], as observed in our study.
Leukocytes are usually lower than 16,000 cells/mm 3 [22,24], which is comparable to our patients (12,700 cells/mm 3 ), but slightly higher than that reported by Ceroni (10,538 cells/mm 3 ) [15]. We did not find any significative difference in leukocyte count between K. kingae and classic pathogens, similar to Basmati et al. [16] but different from Ceroni [15].
Ceroni et al. [15] proposed a model to discriminate K. kingae from pyogenic osteoarticular infections in children less than 4 years. According to this model, the best predictors of K. kingae osteoarticular infections consists of the following four variables: temperature at admission <38 • C, CRP < 55 mg/L, WBC count < 14,000 cells/mm 3 and neutrophil band shift < 159 forms/mm 3 . In our cohort, applying these parameters (with the exception of band shift that we could not quantify), the predictive value for WBC count < 14,000 cells/mm 3 (63.2%) and CRP < 55 (82.8%) was lower than that reported by Ceroni (77.1% and 90.3%, respectively) [15], but similar to Basmaci [16]. Applying the Ceroni Score to our cohort, the sensibility (less than two criteria) for K. kingae was low (65.9%), with a higher positive predictive value (PPV) of 86.7%, and for typical pathogens (considering ≥ two criteria, as band test was not performed) a sensibility of 87.1%, but a low PPV (65.9%). Indeed, these recommended algorithms to differentiate K. kingae infections from other pyogenic bacteria have different discrimination power, due to diverse geographic patterns [16] and different applications (Ceroni applied to all osteoarticular infections, Basmati included older children), and cannot be applied to our cohort. Our study provides evidence that age ≥ 6 months but ≤2 years, apyrexy and CRP ≤ 100 mg/L was a better model to distinguish K. kingae SA from classic pathogens in children <5 years, with an overall PPV of 86.7%, 88.6% for K. kingae and 83.9% for pyogenic infections.
Our study has limitations, such as being a small, single-centre study needing validation in a larger prospective multi-centre study. We plan to increase our sample in the future in order to strengthen the analyses and if possible, include other centres in the study. Furthermore, other parameters such as respiratory rate, heart rate or blood pressure at admission or disseminated disease, being used in severity scores for acute osteomyelitis in children [24][25][26] were not tested and could have improved our analyses.
K. kingae SA is less severe than pyogenic SA and needs less aggressive management and a different antibiotic empiric treatment approach. In children with SA under 5 years of age, we propose that the presence of three positive criteria, age ≥ 6 months but ≤2 years, apyrexy and CRP ≤ 100 mg/L, could be reliably used to diagnose K. kingae SA at admission.