It is known that some COVID-19 patients have poor outcomes and the factors responsible for determining these unfavorable evolution are still unsuccessfully understood. However, age and some comorbidities such as diabetes and hypertension were the first conditions thought to be risk factors for severe COVID-19.1

Since SARS CoV-2 is a novel virus causing infection in humans, there is substantial uncertainty about how early clinical or laboratory findings could be related to infectivity and disease severity, besides how these parameters could contribute to increase mortality. In this context, studies have shown that the exacerbated inflammatory response (cytokine storm) can directly impair organ function in COVID-19 patients with moderate to severe disease, leading to decompensation, organ dysfunction and death.2

In this report, we aimed to study the relationship between serum ferritin levels, measured at the moment of hospitalization, and general mortality among COVID-19 patients admitted to a high complexity university hospital. In addition, we also investigated if there is a ferritin cut-off value that could be predictive of death as the final outcome, which could be very useful in clinical practice for following up moderate to severe COVID-19 cases.

During the months of May through July 2020, we received respiratory tract samples of hospitalized patients to assess COVID-19 suspected cases. A total of 97 SARS-CoV-2 positive patients were diagnosed. The overall mean age (±SD) was 59.9±16.3 years, 57/97 (58.8%) were male; 42.3% self-declared to be white, 9.3% black, 23.7% brown, 1.0% Asian, and that was no information for 23.7%. According to clinical severity, patients were admitted at the emergence room, infectious disease unit or intensive care unit. General symptoms presented at the time of hospital admission were mainly fever, cough, sore throat, sneezing, loss of taste, and eventually diarrhea or abdominal pain. In general, prostration and prominent dyspnea with O2 saturation <95% associated to a ground glass pattern on thoracic tomography (>50%) were also frequently observed (cases were moderate to severe), constituting a typical presentation of hospitalized cases in our center. Overall, 44 (45.4%) patients died due to COVID-19. Biochemistry tests were performed within the first 24–48h of hospitalization. We carefully checked the exact day of respiratory samples collection for SARS-CoV-2 detection and the day of symptoms onset. In our institution we have a CMV screening routine for laboratory monitoring in immunosuppressive patients and these results were also included, but with no statistical significance.

As shown in Table 1, the first part of the evaluation was a univariate analysis comparing the two groups (survival vs. death). Overall, there were statistical differences in age (p<0.001), serum levels of ferritin (p<0.05), C-reactive protein (p<0.001), albumin (p<0.05), and creatinine (p<0.05). In addition, admission to intensive care unit (ICU) was also significantly associated (p<0.0001). Importantly, serum ferritin levels at admission were not statically different between patients with and without immunosuppressive conditions (2661±3188ng/mL vs. 3028±3644ng/mL, respectively; p=0.9).

We also performed an analysis of demographic, clinical and laboratory parameters of COVID-19 hospitalized patients according to quartiles of serum ferritin obtained at admission. Patients with the highest ferritin levels (Q4 – ferritin of 3345–14,660ng/mL) also presented significant higher levels of C-reactive protein (p<0.005) and serum creatinine (p=0.04). These patients also had significantly lower levels of hemoglobin (p=0.03) and serum albumin (p=0.04). Finally, we observed an increased frequency of death by COVID-19 in the highest ferritin quartile (p=0.003), where 75% of patients in the highest quartile evolved to death. These data are presented in Table 2. Of note, we performed the analysis of other clinical parameters such as higher frequency of comorbidities according to quartiles of ferritin, but no significant differences were observed (data not shown).

Next, to further test the relationship between each evaluated variable and mortality, we performed a series of ROC curves analyses as seen in Table 1. For non-significant variables (as LDH, neutrophil and lymphocytes count, for example, the ROC curves had low AUC values and a p-value less than 0.05). For the variables with a significant p-value, (as age, ferritin, C-reactive protein, serum albumin and creatinine), we identified that only age and ferritin presented a value of AUC greater than 0.7 (0.72 and 0.79; respectively) using cut-offs of 60.0 years for age, and 1873.0ng/mL for ferritin, respectively. Using this cut-off for ferritin, we observed sensitivity of 68.4% and specificity of 79.3%. Fig. 1 shows the display of ROC curve for ferritin.

Fig. 1.

ROC curve for ferritin, with the grouping outcome of being alive or dead at the end of the study.

(0.04MB).

Lastly, we performed a logistic regression model for the analysis of age and serum ferritin as independent predictors of mortality. In this regard, based on a p<0.1 in univariate analysis seen in Table 1, the following variables were included in a step-wise logistic regression model: age, ferritin, C-reactive protein, serum albumin, and creatinine. In the final model, only age ≥60.0 years-old (OR: 10.5; 95% CI=1.8–59.5; p=0.008) and ferritin ≥1873.0ng/mL (OR: 6.0; 95% CI: 1.4–26.2; p=0.016) remained independently associated with mortality. Table 3 shows the final logistic regression model.

In December 2019, a new coronavirus was identified as responsible for cases of severe pneumonia, which was later called SARS-CoV-2.3 The virus spread quickly worldwide and subsequently, the World Health Organization (WHO) declared its associated disease, COVID-19, as an international public health emergency.4 Genome sequencing of this virus was carried out5 and led to the development of specific diagnostic tests based on reverse transcription real time-PCR (RT-qPCR).6

The acute phase reaction of an inflammatory process consists of a series of physiological and metabolic changes that begin immediately after tissue injury.7 Among the numerous systemic manifestations of this acute phase reaction is the variation in the concentrations of various plasma proteins, which are called “acute phase proteins”. Among these, the best known in clinical practice are C-reactive protein, amyloid substance serum A, haptoglobin, fibrinogen and ferritin.8 Serum ferritin has been long studied as a marker of iron metabolism,9 however, its application as biomarker of inflammation has far presented high importance in the context of COVID-19 progression, as demonstrated by previous studies in the field.10 Ferritin is an acute phase reactant, and as such, is generally elevated in inflammatory responses of any type. An initial assessment to identify the laboratory findings most commonly associated with the cytokine storm syndromes include a complete blood count, serum levels of ferritin, liver function tests, and others. These tests are readily available in the majority of health care facilities.

We studied, at the first days of hospital admission, a series of SARS-CoV-2 infected patients and we observed a strong correlation between serum ferritin and overall mortality, independently of age. A worse prognosis frequently is associated with a more rapid evolution to intensive and respiratory care or even dialysis. This is a concise laboratory-based report, which sought to answer whether we are able to identify any relevant role for risk prognosis using complementary tests routinely requested in positive SARS-CoV-2 hospitalized patients. The magnitude of the inflammation the patient already has at admission when the RT-PCR for SARS-CoV-2 is performed, seems to have a prognostic value.

There is increasing evidence that circulating ferritin levels may not only reflect an acute phase response but may play a critical role in inflammation.11 For most clinicians dealing with inflammatory diseases, serum ferritin levels are a rather non-specific marker of the acute phase response, which is often ignored or not measured when the patient is acutely ill. In some diseases, ferritin levels may be extremely high and, while not specific, these very high levels may be helpful in identifying patients at risk.12 In some previous studies, ferritin levels were more elevated in older and/or hypertensive participants, which was also associated with increased mortality.13

We emphasize that the profile of COVID-19 patients in our hospital is different from the general population, as they have advanced age, cardiovascular disease and diabetes, all correlated with higher mortality. We are a reference center for high complexity cases in our region. We highlight that 48.6% of our population had onco-hematological diseases and 22.0% had chronic kidney disease. A very relevant point in this study is the searching for a complementary routine exam that could be used as an independent predictor of mortality at the time of hospital admission.

In our study, CMV viral load at admission was not significantly correlated with final outcomes. CMV reactivation and risk for CMV disease possibly related with a subjacent immunosuppressive status may occur as in systemic lupus erythematous or in organ transplant14,15; however, characteristics of latency and reactivation of the herpes family should be better monitored in the critically ill COVID-19 patients. For example, many of the reported cases and series of CMV-associated to cytokine storm syndromes involve viral reactivation during immune suppressed states.16 The rate of CMV seroprevalence increases with age. CMV has been shown to affect peripheral T cell phenotypes, increase inflammatory mediated cytokines such as IL-6 and play a role in immune dysregulation. The role of CMV in those with severe COVID-19 disease merits exploration.17

A study focusing on early hyperinflammation in COVID-19 patients evaluated the high levels of serum ferritin in the first seven days of hospitalization as a predictor for the cytokine storm syndrome,18 and a meta-analysis including 21 studies (3377 patients and 33 laboratory parameters) also demonstrated serum ferritin as a biomarker for potential progression to critical illness.10 Nevertheless, recent studies have also demonstrated that anti-inflammatory biomarkers could also be elevated during the acute phase of COVID-19.19,20 This highlights the importance of future studies focusing on the balance between pro- and anti-inflammatory mediators and how this could impact on disease progression.

In conclusion, this study was designed to answer whether one could identify any relevant role for prognostic risk using routine complementary exams. The magnitude of the inflammation in the first days after hospitalization, represented by a hyperferritinemic syndrome, may help to identify patients at higher risk for early clinical decisions regarding the preventive measures directed at such patients.

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, #001), Conselho Nacional de desenvolvimento Científico e Tecnológico (CNPq), Ministério da Educação (UFF) and The Brazilian Innovation Agency (FINEP) through Ministério da Ciência, Tecnologia e Inovações – REDE VÍRUS/Laboratórios de Campanha.

The authors declare no conflicts of interest.