COVID-19 Biomarkers Comparison: Children, Adults and Elders

Background and Objectives: this study aimed to research links between C-reactive protein (CRP), lactate dehydrogenase (LDH), creatinekinase (CK), 25-OH vitamin D (25-OHD), ferritin (FER), high-density lipoprotein cholesterol (HDL)cholesterol and clinical severity in patients from the western part of Romania, and compare their potential use as biomarkers for intensive care units (ICU) admission and death in children, adults and elders. Materials and Methods: this study is a retrospective cohort study, performed on patients positively diagnosed with COVID-19. Available CRP, LDH, CK 25-OH vitamin D, ferritin, HDL cholesterol and clinical severity were recorded. The following were assessed: median group differences, association, correlation and receiver operating characteristic. Results: 381 children, 614 adults and 381 elders were studied between 1 March 2021 and 1 March 2022. Most children and adults presented mild symptomatology (53.28%, 35.02%, respectively), while most elders presented severe symptomatology (30.04%). ICU admission was 3.67% for children, 13.19% for adults and 46.09% for elders, while mortality was 0.79% for children, 8.63% for adults and 25.1% for elders. With the exception of CK, all other biomarkers showed some significant associations with clinical severity, ICU admission and death. Conclusions: CRP, LDH, 25-OH vitamin D, ferritin and HDL are important biomarkers for COVID-19 positive patients, especially in the pediatric population, while CK was mostly within normal ranges.


Introduction
From 2003 to the present, a variety of coronaviruses have been identified in humans and animals, according to the literature. This kind of virus can easily spread from one person to another, resulting in infections of varying degrees, most commonly the common cold but also more severe infections [1,2].
At the end of 2019, a new coronavirus, initially called 2019-nCoV, and later renamed SARS-CoV-2, was recognized in China, causing the COVID-19 disease. It was proclaimed a vitamin D, ferritin and/or HDL data. Pediatric patients were defined as younger than 18 years, while elders were defined as older than 65, with adults being in-between.
The severity of the disease was classified as asymptomatic, mild, moderate, severe and critical according to the clinical characteristic and laboratory results: asymptomatic-cases with a positive RT-PCR test and without any clinical findings; mild-cases with upper respiratory tract infection symptoms, such as fever, fatigue, myalgia, cough, sore throat; moderate-pneumonia with complaints of fever and cough but without the symptoms of dyspnea and hypoxemia; severe-cases with fever and cough in the early period which develop dyspnea and central cyanosis within a week (arterial oxygen saturation of <92%); critical-cases which develop acute respiratory distress or respiratory failure rapidly, and who tend to develop shock, encephalopathy, myocardial affection, coagulation dysfunction and acute kidney injury.

Data Collection
The study has a retrospective cohort character, so the patient's written consent was not required, as the data were obtained from the computer register of the laboratories. Epidemiological information such as age, sex and diagnosis were also assessed. Identification data of the patients such as names, ID numbers or contact information were not collected for the present study.

Laboratory Testing
A fully automated clinical chemistry instrument Cobas Integra 400 plus (Roche Diagnostics, Bucharest, Romania) was used for analyzing the serum sample. Nasopharyngeal swabs of the patients were examined using the BIO-RAD CFX96 Real-Time System C1000 Touch Thermal Cycler Device.

Statistical Analysis
Sample size was calculated using the G*Power software, v 3.1.9.6 (Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany), using an a priori test to calculate the minimum sample size for a medium effect size (d = 0.25), a power of 80% and an alpha of 0.05. The result indicated that a minimum of 270 patients were needed.
Statistical analysis was performed using the MedCalc Statistical Software, version 20.218 (MedCalc Software bv, Ostend, Belgium) All continuous variables were assessed with the Kolmogorov-Smirnov test, to check for normal distribution. Continuous variables with a normal distribution were expressed as mean and standard deviation (SD), and were compared using the ANOVA with Tukey' post hoc test. Non-parametric variables were expressed as median and interquartile range (IQR), and were compared using the Kruskal-Wallis with Dunn's post hoc tests.
In order to find associations between the existence of clinical respiratory signs and the elevated status of the studied biomarkers, contingency tables were created and analyzed using the Chi 2 test. Afterwards, links between biomarkers levels were assessed using Spearman's rank correlation, where r ≤ 0.10 was considered "very weak", 0.10 < r ≤ 0.33 was considered "weak", 0.33 < r ≤ 0.66 was considered "moderate" and r > 0.66 was considered "strong".
The receiver operating characteristic (ROC) curve was used to assess the diagnostic value of any of the studied values in predicting admission to the ICU and risk of death due to COVID-19. The cutoff point was determined using the Youden index. A comparison of independent ROC curves was also done. For all tests, p < 0.05 was considered statistically significant.

Results
A total of 7789 patients were tested during the study period. Of these, 1783 were children, with only 381 found to be positive for COVID-19 upon admission to the ER. A total of 4625 were adults, with only 614 found to be positive for COVID-19 upon admission to ER. There were 1381 elders in total, and only 243 were positive for COVID-19 upon admission to ER.

Children
The total number of studied pediatric patients was 381. Regarding sex grouping, no differences were observed at the Kruskal-Wallis test, with the exception of ferritin, which is physiologically different between the two sexes. At the Chi 2 test, no associations could be observed with regard to severity, ICU admission or death. All the data can be observed in Table 3. With regard to severity, differences were observed at the Kruskal-Wallis test, with the exception of age and CK levels. At the Chi 2 test, associations could be observed with regard to severity and both ICU admission or death. All the data can be observed in Table 4. A further link between data was assessed using Spearman's rank correlation test. A total of 15 (3.94%) out of 381 patients had available data on all the studied variables, as seen in Table 5. Considering that CK and age did not show differences at the Kruskal-Wallis test, they were omitted from this analysis.  The following correlations were observed: a moderate inverse relationship between CRP and HDL, a strong direct relationship between CRP and ICU admission, a moderate inverse relationship between ferritin and HDL, a strong direct relationship between ferritin and ICU admission, a moderate direct relationship between ferritin and deceased, a strong inverse relationship between HDL cholesterol and ICU admission and a moderate direct relationship between ICU admission and deceased.
Demographic data is presented in Table 6. The sex distribution was similar (M/F = 52.12%/ 47.88%). Most patients presented mild symptomatology (n = 215, 35.02%). There were 81 (13.19%) cases with ICU admission, of which 33 (40.74%) were severe and 30 (37.04%) were critical. There were 53 (8.63%) deceased patients. Median and IQR values for the quantifiable data of the pediatric patients can be observed in Table 7. Regarding sex grouping, no differences were observed in the Kruskal-Wallis test, with the exception CK, which is physiologically different between the two sexes and 25-OH vitamin D. Males tended to have lower 25-OH vitamin D levels (median = 24.28, p = 0.0028). No associations could be observed from the Chi 2 test with regard to severity, ICU admission or death. All the data can be observed in Table 8. With regard to severity, differences were observed in the Kruskal-Wallis test, with the exception of age and CK levels. Associations could be observed from the Chi 2 test with regard to severity and both ICU admission or death. All the data can be observed in Table 9, with Dunn's post-hoc test results in Table 10.   A further link between data was assessed using Spearman's rank correlation test. A total of 220 (35.83%) out of 614 patients had available data on all the studied variables, as seen in Table 11. Considering that CK and age did not show differences at the Kruskal-Wallis test, they were omitted from this analysis.
The following correlations were observed: a moderate direct relationship between CRP and LDH, a moderate inverse relationship between CRP and 25-OH vitamin D, a moderate direct relationship between CRP and ferritin, a weak inverse relationship between CRP and HDL cholesterol and a moderate direct relationship with both ICU admission and deceased, a moderate inverse relationship between LDH and 25-OH vitamin D, a weak direct relationship between LDH and ferritin and with both ICU admission and deceased, a weak inverse relationship between 25-OH vitamin D and ferritin, a weak direct relationship between 25-OH vitamin D and HDL cholesterol, a moderate direct relationship with both ICU admission and deceased, a weak inverse relationship between ferritin and HDL cholesterol, a weak direct relationship between ferritin and ICU admission, a moderate relationship with deceased, weak inverse relationships between HDL cholesterol and both ICU admission and deceased and a strong direct relationship between ICU admission and deceased.

Elders
The total number of studied elder patients was 243. Among these patients, 229   Median and IQR values for the quantifiable data of the pediatric patients can be observed in Table 13. Regarding sex grouping, alongside CK and ferritin, which are physiologically different between the two sexes, a noticeable difference was observed for LDH, which was higher in males (median = 223, p = 0.0462). No associations could be observed from the Chi 2 test with regard to severity, ICU admission or death. All the data can be observed in Table 14.  With regard to severity, differences were observed in the Kruskal-Wallis test, with the exception of age and 25-OH vitamin D levels. Associations could be observed from the Chi 2 test with regard to severity and both ICU admission or death. All the data can be observed in Table 15, with Dunn's post-hoc test results in Table 16.  A further link between data was assessed using Spearman's rank correlation test. A total of 77 (31.69%) out of 243 patients had available data on all the studied variables, as seen in Table 17. Considering that age did not show a difference in the Kruskal-Wallis test, it was omitted from this analysis. In this group, CK differences were observed in the previous test and were included, however, no correlation was statistically significant. Regarding 25-OH vitamin D, no differences were observed in the previous test in elders, yet there were significant correlations with CRP (weak and inverse) and ICU admission (moderate and inverse).  Other important correlations that were observed: a moderate direct link between CRP and LDH; a moderate direct link between CRP and ferritin; a moderate inverse relationship between CRP and HDL cholesterol; a moderate direct relationship between CRP and ICU admission; a weak direct relationship between CRP and deceased; a moderate direct link between LDH and ferritin, ICU admission and deceased; a weak inverse relationship between ferritin and HDL cholesterol; a moderate direct relationship with ICU admission and a weak direct link with deceased. HDL cholesterol also correlated with moderately and inversely with ICU admission; ICU admission was linked moderately and directly with deceased.

Risk for ICU Admission
Details from the testing using the ROC curve analysis with regard to ICU admission are shown in Table 18. CK was excluded from this analysis, as, across all testing, its AUC (area under curve) was not statically significant. Graphically, this data can be seen in Supplementary Material, Figures S1-S6. Observed differences are between the 25-OH vitamin D AUC of elders and both adults and children, being significantly smaller (0.631 vs. 0.838, 0.778, respectively) and ferritin where the AUC of children was significantly higher than both adults and elders (0.949 vs. 0.807, 0.735, respectively). Regarding 25-OH vitamin D, this might be due to lower levels in elders overall, considering that their median was also lower than that of both adults and children (19.21 vs. 25.96, 23.44, respectively). Ferritin might be a better marker in children than in adults or elders, due to its high AUC.

Risk for Death
Details from the testing using the ROC curve analysis with regard to death are shown in Table 19. CK was excluded from this analysis, as, across all testing, its AUC was not statically significant. Graphically, this data can be seen in Supplementary

Discussion
The emergence of the global crisis due to the COVID-19 pandemic has led to measures such as social distancing, self-isolation, quarantine. Children and the elderly are important, as they present somewhat polarizing sides with regard to risk and severity. While children have been less studied, especially in the early stages, elders were observed to have a higher risk of severe illness and mortality. This effect has led to social disfunction in both cases [24,25], and has also affected the psychosocial institutions to which some of these patients cater to [26] In order to manage COVID-19 patients and avoid complications, quick identification and the supply of the right treatments are essential. On the other hand, in its inception, the disease affected predominantly elders and patients with comorbidities [27] Regarding clinical presentation, the proportion of asymptomatic children fell within the range previously reported by studies such as Ylimaz et al. [9], Say et al. [28] or Jat et al. [29], which was approximately one-third of total pediatric patients (36.75%). Contrarily, as age progressed, so did the severity of the cases, with only 16.12% being asymptomatic and even fewer (9.47%) patients older than 65.
It was documented that CRP values have a predictive role with regard to severity in adults, due to the immune system responding to the SARS-CoV-2 infection [30,31]. This was also observed in our study, as all three types of analysis showed a link between CRP levels and severity, ICU admission or death.
CRP is known to trigger a cytokine storm, which results in a massive release on proinflammatory cytokines, which can cause acute lung injury, resulting in an unfavorable prognosis, especially in elders [32]. In our studied elders, those with higher CRP were in the severe or critical categories. Generally, CRP was associated with ICU admission across all age groups; however, it only correlated with mortality for adults and elders.
In a previous study, it was established that CRP was higher in older children and adolescents, whose immune system is more developed, and, hence, more akin to that of adults' and elders' [33]. Following a retrospective analysis of a group of 3424 children in Indonesia, it was concluded that CRP levels may be a good indicator for the treatment and early identification of the disease in children [34].
The levels of CRP in relation to ICU admission presented a good AUC for all three age groups, with no significant differences.  [35].
Information on LDH implication was contradictory [34,36]. In our study, its levels correlated well with severity, ICU admission and death. For ICU admission, AUCs were similar, with the predictive values being: 288 for children, who exhibited sensitivity 80% and specificity 67.74%; 302 for adults, who exhibited sensitivity 78.69% and specificity 74.53%; and 184 for elders, who exhibited sensitivity 78.35% and specificity 56.48%.
In the case of mortality, where the AUC of children was higher than of both adults' and elders' (0.936 vs. 0.729, 0.740, respectively). The cutoff values were: 354 for children, who exhibited sensitivity 100% and specificity 87.12%; 294 for adults, who exhibited sensitivity 74.47% and specificity 68.1%; and 195 for elders, who exhibited sensitivity 85.45% and specificity 55.33%. LDH levels have been associated with clinical severity since the beginning of the pandemic, and many authors agree on and encourage its use as a clinical biomarker [37,38].
Although CK levels have been associated with COVID-19 in adults, its use as a potential biomarker remains controversial [39][40][41]. In our study, no association could be observed between this biomarker and the severity of cases. As depicted in the supplementary material; its AUC value was not statistically significant.
The modulation of immunological response is significantly influenced by 25-OH vitamin D [15]. Children with inadequate 25-OH vitamin D may experience worse clinical outcomes in Omicron subvariant BA 2 infections, according to research by Peng et al. [42]. In a study by Michael et al., supplementation of 25-OH vitamin D in adults and elders proved to decrease the risk of mechanical ventilation and clinical severity in COVID-19 [43]. The usefulness of 25-OH vitamin D supplementation was also observed in a randomized controlled trial done by Zurita-Cruz et al. [44]. Studies by di Filippo et al. showed a strict association of VD levels with blood GLU and BMI in COVID-19 patients on one hand, and that 25-OHD levels at admission strongly predicted the occurrence of worsening outcomes in COVID-19 independently of the disease severity at presentation. [45,46].
Our research demonstrated that patients' levels of 25-OH vitamin D were inversely related to the presence and severity of clinical symptoms, as well as ICU admission and death. A difference in AUC's was observed when analyzing ICU admission, with both adults' and children' being greater than the elders' (0.631 vs. 0.838, 0.778, respectively). Cutoff values were: 17.07 for children, who exhibited sensitivity 81.82% and specificity 76.35%; 20.05 for adults, who exhibited sensitivity 83.05% and specificity 86.81%; and 22.44 for elders, who exhibited sensitivity 86.81% and specificity 35%.
25-OH vitamin D levels were observed to be similar across all elder severity groups, while in the adult and child groups there were differences according to severity. As such, for the risk of death, elder AUC was lower than that of children's and adults' (0.613 vs. 0.803, 0.883, respectively). Cutoff values were: 14.11 for children, who exhibited sensitivity 100% and specificity 78.98%; 17.34 for adults, who exhibited sensitivity 85% and specificity 84.74%; and 17.88 for elders, who exhibited sensitivity 54% and specificity 67.38%.
Ferritin is another biomarker that has been associated with severity in adults and is able to predict poor outcomes and prognosis, although its use is still somewhat contested [47][48][49]. In our patients, ferritin proved to be a good predictor of severity, ICU admission and death. Risk for ICU admission by studying ROC curves showed a higher AUC for children than for adults and elders (0.949 vs. 0.807, 0.735, respectively), with the cutoff values being: 211 for children, who exhibited sensitivity 100% and specificity 86.02%; 529 for adults, who exhibited sensitivity 60.78% and specificity 89.35%; and 133 for elders, who exhibited sensitivity 80.85% and specificity 60.4%.
The last biomarker that was studied was HDL cholesterol. A modified metabolic aspect has been observed in many COVID-19 patients, and this includes blood lipid levels. HDL cholesterol has been associated with poorer outcomes and more severe cases [21,50]. It has been suggested that it is involved in the regulation of innate immune response through interaction with ABCA1 or ABCG1, which negatively regulates T-cell activation and the expression of inflammatory mediators [49]. In our study, HDL correlated inversely with severity, ICU admission and death. It proved a good predictor for ICU admission, with all AUCs being similar. The following cutoff points were established: 47.75 for children, who exhibited sensitivity 85.71% and specificity 56.11%; 31.81 for adults, who exhibited sensitivity 71.6% and specificity 59.31%; and 33.17 for elders, who exhibited sensitivity 48.91% and specificity 80.68%.
With regard to its role as a mortality predictor, a difference was observed between the AUC of children and that of elders' (0.858, 0.623), with adults' in-between, without significant differences. Cutoff points were: 41.35 for children, who exhibited sensitivity 100% and specificity 67.4%; 32.03 for adults, who exhibited sensitivity 69.23% and specificity 60.04%; and 30.99 for elders, who exhibited sensitivity 39.13% and specificity 81.21%.
No differences due to sex, except physiological, were observed across all studied groups. The sex distribution was similar, with a slight increase in males for children and adults (55.64%, 52.12%, respectively) and a slight increase in females in the elder groups (57.61%).

Limitations
As in all retrospective studies, this article presents several limitations. First and foremost: data distribution was heterogeneous, as not all patients had available data for all studied biomarkers. For example, only a small number of pediatric patients (3.94%) presented data on all studied biomarkers. Data was also missing with regard to multi system inflammatory syndrome, which is a serious condition sometimes linked to COVID-19, or comorbidities, which are usually more common among older patients.
Another possible limitation is the small sample of critical and deceased pediatric patients, which might skew the results. In some cases, severity was assessed after admission and this time difference may affect interpretation of results. Lastly, data on children was collated, which might not account for differences within the children group (i.e., infants vs. children vs. teenagers).

Conclusions
It the present study, several biomarkers proved useful for the analysis of severity, ICU admission and mortality. CRP is one of the most important biomarkers and correlated well with outcomes across all ages, with the calculated cutoff values as a mortality predictor higher for adults than elders. LDH correlated well with outcomes across all ages, especially in adults and elders. CK values were mostly within normal ranges, and cannot be considered as a biomarker in the assessment of the health of pediatric patients with COVID-19 and did not offer any correlation results. 25-OH vitamin D is also an important marker of severity, ICU admission and death, especially in adults, but less so in elders. Ferritin correlated the best among all studied markers with outcomes across all ages; however, it would seem more useful for pediatric patients and somewhat less useful for elders. HDL cholesterol also inversely correlated well across all studied groups with regard to severity, ICU admission and death, being similarly useful as a predictor.