NLR is defined as the absolute neutrophil count (ANC)/absolute lymphocyte count (ALC) (27–29). d-NLR has a similar definition to NLR, calculated as ANC/(White blood cells (WBC) total count − ANC) (30). If we consider monocyte, basophil, and eosinophil levels as negligible (which are mostly not), the definition of these two markers would be equal. Earlier studies found links to higher NLR or d-NLR in chronic conditions with low-grade inflammatory nature, such as obesity, hypertension, diabetes mellitus, metabolic syndrome, atherosclerotic events of the heart and brain, and various cancers, although previous literature studied NLR more than d-NLR (27–29, 31–33). These underlying diseases are considered as risk factors for severe COVID-19 (2, 34, 35).

Since the beginning of the pandemic, studies have investigated the role of NLR in COVID-19 prognostication and its utility as a biomarker. NLR has been reported to prognosticate mortality, progression to severe disease, risk of intubation, risk of severe disease in intubated patients, days intubated, ICU admission, and longer intensive care unit (ICU) admission (30, 36–57). Two meta-analyses of n = 19 and n = 13 studies found significant associations between higher NLR and COVID-19 severity and mortality (58, 59). Furthermore, patients with higher NLR appear to have more comorbidities and, therefore, are more prone to severe COVID-19 (36). Even in patients with comorbidities, NLR might maintain its predictive ability for COVID-19 severity. For instance, NLR significantly predicted COVID-19 severity and survival in hospitalized patients with different types of cancers (60, 61). It has been suggested that each increased NLR unit resulted in an 8% higher mortality in COVID-19 patients (45).

A temporal analysis showed that on-admission NLR correlates well with the need for ICU and poor outcomes, and can be a potential risk-stratification tool. However, the clinical utility of NLR was lost in week 3 post-admission (62). The best predictive value of NLR can be achieved at its peak compared with its on-admission values (63). Concurrently, another study on the temporality of NLR found that day 7 measurement of NLR could significantly predict those requiring invasive mechanical ventilation and mortality, while measurement of day 1 NLR could not (64). In summary, on-admission NLR could predict COVID-19 prognosis. This predictive ability increases for a few days after admission, when NLR reaches its peak. However, NLR gradually loses its predictive ability as the patient recovers from COVID-19 and an associated reduction in inflammation. Finally, at week 3 post-admission, NLR loses its clinical utility to prognosticate severe COVID-19 outcomes.

Five studies proposed the ability of NLR to assist COVID-19 diagnosis (49, 65–68). They defined assisting COVID-19 diagnosis as significantly higher levels of NLR in individuals with COVID-19 compared with healthy controls. However, none of the studies mentioned how and due to what situations NLR can be integrated into COVID-19 diagnosis. Two other markers have been reported to be predictive for COVID-19 disease severity and mortality: granulocyte to lymphocyte ratio (69) and d-NLR (30, 70, 71).

PLR could help in diagnosing COVID-19. Four studies found a significant difference in the PLR of patients with positive SARS-CoV-2 compared to healthy individuals (49, 65–67), while only one concluded against this (68).

Two meta-analyses confirmed the effectiveness of higher PLR on predicting COVID-19 severity (59, 83). Higher PLR also correlated with an increased risk of severe disease in intubated COVID-19 patients (49). This ability to predict disease severity seemed to be optimal at its peak. PLR at peak could predict disease severity in the multivariate regression analysis; however, PLR at admission did not correlate with disease severity in the univariate analysis (84). Two studies studied the ability of PLR to predict ICU admission, and they produced conflicting results on this matter (53, 75).

Although PLR could predict disease severity in most of the studies, it was not able to predict mortality (61, 66, 77, 78), one specifically in cancer patients (61). Similarly, another study concluded that PLR is only slightly prognostic in predicting mortality in the univariate analysis among hospitalized patients (p < 0.001), but not in the multivariate analysis (p = 0.154) (82).

Owing to all the strengths of PLR in predicting various COVID-19-related parameters, it is a potentially suitable marker to triage COVID-19 patients. However, it seems to lack potentials to predict mortality and have a lower ability than NLR to predict several parameters.

Unlike neutrophil and monocyte count, a decrease in lymphocyte count correlated to multiorgan injury in COVID-19 patients (85). This was shown by Kazancioglu et al recently. However, in their study, monocyte count only correlated to SARS-CoV-2 infection but not severity (69). These studies provide the hypothetical bases for the prognostic value of LMR in COVID-19, as well as NLR and PLR.

Two studies compared the effectiveness of LMR in COVID-19 diagnosis, both finding a significant relationship between LMR and testing positive for SARS-CoV-2. In one of them, MLR (AUC: 0.892) was the best predictor compared with NLR and PLR, and 0.23 was declared the best MLR cutoff point (65). In another, significantly lower LMR was observed in intubated COVID-19 patients compared with healthy controls (49).

LMR did not correlate with disease severity in most studies (30, 71, 73, 74) except two (53, 72). Liu et al. showed that higher LMR could only significantly predict disease severity in the univariate analysis and also did not correlate with a longer hospital stay, higher hospital costs, and longer time to negative PCR (72). LMR could prognosticate progression to ARDS (65).

In two studies examining the ability to predict ICU admission, LMR did not correlate in one (53), and correlated but was inferior to NLR and PLR in the other (75). Data are limited regarding the ability of LMR to predict COVID-19 mortality; however, a study concluded the ineffectiveness of LMR in prognosticating this parameter (77).

LMR might have limited benefits in prognosticating COVID-19 (86), but its abilities seem to be lower than NLR and PLR, especially in predicting disease severity, ICU admission, and mortality. However, we encourage future studies to pursue the ability of LMR to recognize SARS-CoV-2 positive patients, as it demonstrated promises.

A limited number of studies examined this marker. The most important article on this is perhaps a meta-analysis on the role of LCR in predicting disease severity. They found a significantly lower LCR in patients with severe disease (SMD = −0.912, p < 0.001); however, it was less predictive compared with NLR (SMD = 2.404, p = 0.001). They based their results on five studies for each marker (80).

Higher LCR on day 1 predicted the need for ICU admission (adjusted OR: 3.1, p = 0.003) and invasive mechanical ventilation (adjusted OR: 2.5, p = 0.009), but could not predict in-hospital mortality (p = 0.60) and requiring dialysis (p = 0.44). Nevertheless, lower LCR on day 7 only correlated with an increased in-hospital mortality risk (adjusted OR: 0.1, 95% CI = 0.01–0.30, p < 0.0001) but not with the other factors (64).

Similar to most of the discussed markers, studies demonstrate a prognostic role for FPR and AFR in some cancers and other diseases with inflammatory pathophysiology (87–89). Nevertheless, only a few articles studied them to determine COVID-19 disease severity.

Lower AFR correlated with severe COVID-19 in univariate analysis (p < 0.0001), but not multivariate analysis (p = 0.079) (71). However, fibrinogen to albumin ratio (FAR), remained significant in predicting disease severity in the multivariate analysis in another study (HR = 4.058, 95% CI = 1.246–13.222, p = 0.020) (90).

CAR could predict disease severity in two studies (79, 91): one comparing it with NLR and finding a higher OR for CAR (OR = 17.652, p = 0.001) than NLR (OR = 1.512, p = 0.007) (79). However, the other study did not find as large an OR for CAR (1.264, p = 0.037) (91).

hsCAR and hsCPAR differ from their counterparts as they utilize high-sensitivity CRP (92). Xue et al. found that on-admission hsCAR, hsCPAR, and PNI significantly correlated with severe COVID-19 in the multivariate analysis among several other markers (NLR, LMR, FPR, PLR, SII, AFR). Furthermore, among these markers, only hsCPAR and hsCAR correlated with hospital stay length (71). In the Oh et al. study, hsCAR could also predict in-hospital mortality in adults older than 65 years of age after adjusting for confounders (92).

Taken together, CAR and CPAR seemed promising in predicting disease severity, mortality, and length of hospital stay in all the studies, although we only identified four studies. Furthermore, a small study hailed prealbumin and CRP as potential markers to effectively triage patients in the early stages, and prealbumin seemed to be more effective (93).

GPS constitutes two main serum components, CRP and albumin levels, both having a potential of 0 or 1 score. CRP >10 mg/L and albumin <3.5 mg/dl receive one point each, and the score classifies the patients into three total scores of 0, 1, or 2. mGPS does not allocate a score to hypoalbuminemia without a rise in CRP to above 10 mg/L. hs-mGPS provides a similar classification to mGPS with a lower CRP threshold (>3 mg/L) (94, 95). These markers also have proven roles in predicting various cancers (94, 95).

In a study of 397 patients with COVID-19, no deaths occurred among 40 patients with hs-mGPS score of 0, while 10/263 (3.80%) and 24/94 (25.53%) of patients scoring 1 and 2 died, respectively (70). Concurrently, Dettorre et al. found that mGPS was able to foretell the overall survival of hospitalized cancer patients infected with COVID-19 (11.4%, 30.4%, and 50.6% for mGPS = 0 to mGPS = 2, respectively; p < 0.0001) (61). These two studies found promising results for mGPS and hs-mGPS and set the grounds for future research to better identify its effectiveness.

PI is similar to the GPS, only differing in the WBC component. WBC >11,000/µl and CRP >10 mg/L contribute to this scoring system of 0 to 2 (22).

We could only find one study discussing this marker in COVID-19. In that study, PI was able to predict the survival of cancer patients infected with COVID-19 (9.1%, 40%, and 50%, for scores of 0 to 2), similar to NLR, PNI, and mGPs. PI seemed superior to PLR in the study. Patients categorized in the poor-risk group (PI = 2) had 23 days median overall survival, while patients with favorable scores did not reach the required follow-up duration (all p-values less than 0.0001) (61).

Onodera et al. proposed PNI as an immune-nutritional risk score for malnourished cancer patients undergoing for gastrointestinal surgery, formulizing it as serum albumin concentration (g/L) + 0.005 × total lymphocyte count (per mm³ of peripheral blood) (96, 97). This marker later demonstrated its effectiveness in prognosticating several types of cancer (98).

Three studies investigated the relationship between COVID-19 mortality and PNI, all showing significant correlations (61, 70, 82). Two studies were on hospitalized COVID-19 patients (70, 82) and another on patients with cancer (61). PNI also predicted disease severity better than other markers in the study of Xue et al., alongside hsCAR, and hsCPAR. Together, they were put into a nomogram that could predict disease severity well (C-index = 0.873) (71).

PNI successfully predicted disease severity and mortality in all the four studies examining it; therefore, it can be a suitable candidate for follow-up studies.

SII is defined as platelet count × NLR (99). SII remained a reliable predictor in most of the conducted studies on COVID-19 so far. Usul et al. found its superior predictive ability than NLR and PLR in COVID-19 diagnosis, as its values were significantly different in SARS-CoV-2-positive and -negative individuals. The proposed SII for helping in the COVID-19 diagnosis was 479.1 (68).

Xue et al. studied the relationship between several markers and disease severity. They found that SII could significantly predict disease severity in the univariate but not the multivariate analysis, inferior to hsCAR, hsCPAR, and PNI, but still better than several markers, such as NLR (71).

Two studies studied SII in predicting mortality, both finding significant correlations (70, 81). One of them found that SII was the only significant marker in the multivariate analysis, superior to NLR and d-NLR, but with a slight hazard ratio (HR = 1.0001, p = 0.029) (81).

IFI27 is a part of the innate immune system highly induced by interferon (IFN)-α (100). High expression of IFI27 may also induce cell proliferation and invasion and reduce apoptosis, making it a possible oncogene (100, 102).

Type I IFN deficiency can be a marker of severe COVID-19 (103). Some of the IFN-stimulated genes (ISGs) like IFI27 were upregulated in patients with COVID-19 (104–107) and later downregulated in the recovery process (108, 109). IFI27 was overexpressed in various cell lineages of SARS-CoV-2-infected patients compared with healthy controls (110). In another study, IFI27 demonstrated a higher than two-fold upregulation in A549 and normal human bronchial epithelial (NHBE) cells infected with SARS-CoV-2 (111). Shaath et al. analyzed the bronchoalveolar lavage of 10 individuals. IFI27 was among the genes upregulated in the peripheral blood mononuclear cells (PBMC) of severe and mild COVID-19 patients, compared with two healthy controls (112).

IFI27 might also help distinguish COVID-19 from other acute respiratory illnesses and some viral diseases such as Ebola, SARS, MERS, and H1N1, as SARS-CoV-2, even at low loads, induced IFI27 more than other viruses (113, 114).

There is a lack of robust clinical evidence concerning IFI27-related prognostic value for COVID-19; however, ISGs and, particularly, IFI27 seem to be interesting for conducting further studies.