Thrombocytopenia and increased risk of adverse outcome in COVID-19 patients

Study design and patient enrollment

A retrospective cohort study was conducted in four designated hospitals for COVID-19 in Hubei province in China, including Wuhan First Hospital, Renmin Hospital of Wuhan University, Tongji Hospital, Tongji Medical College in Wuhan city, and Huangmei hospital in Huanggang city. All the patients were confirmed by positive SARS-COV-2 RNA detection on nasopharyngeal swab samples that were collected for real-time reverse transcriptase polymerase chain reaction (RT-PCR) assay (Chinese Center for Disease Control and Prevention, 2020).

Sample collection and data extraction

All of the electronic medical records of the hospitals were reviewed by trained medical staff, from which demographic information, epidemiological history, underlying comorbidities, clinical manifestations, laboratory test, CT scan, treatment history and outcome were extracted and summarized by standardized EpiData data collection form. The outcome included respiratory failure, intensive care unit (ICU) entrance, acute heart injury, acute kidney injury, septic shock and disseminated intravascular coagulation (DIC). The plasma with 0.109 mol/L sodium citrate was collected for the detection of coagulation parameters. The serum was collected for the detection of cytokines. The study was approved by the Research Ethics Commission of Tongji Hospital, Tongji Medical College (TJ-IRB20200102), and the requirement for informed consent was waived by the Ethics Commission.

Definition

Thrombocytopenia was defined as PLT < 100 × 10⁹/L and non-thrombocytopenia was defined as the PLT ≥ 100 × 10⁹/L which were evaluated on admission. Respiratory failure was defined as arterial oxygen partial pressure <60 mmHg with or without carbon dioxide partial pressure >50 mmhg under the condition of breathing air at rest. According to the KDIGO Clinical Practice Guideline, acute kidney injury was defined as any of the following: an increase in serum creatinine levels by ≥0.3 mg/dl (≥26.5 µmol/l) within 48 h; increase in serum creatinine to ≥1.5 times baseline, which was known or presumed to have occurred within the prior 7 days; urine volume <0.5 ml/kg/h for 6 h. Acute heart failure was defined as a clinical syndrome of current or previous symptoms and/or signs caused by structural and/or functional abnormalities of the heart, combined with elevated levels of natriuretic peptides or objective evidence of cardiogenic pulmonary or systemic congestion (Bauersachs et al., 2022). Septic shock was defined as persistent hypotension requiring vasopressors to maintain MAP ≥ 65 mm Hg and serum lactate levels > two mmol/L (18 mg/dL) on the basis of adequate volume resuscitation (Singer et al., 2016). The International Society on Thrombosis and Haemostasis (ISTH) DIC 8-point score was applied for DIC calculation (Taylor Jr et al., 2001). A platelet count >100 × 10⁹/L accounts for 0 points, a score between 50–100 × 10⁹/L for 1 point and a platelet count <50 × 10⁹/L for 2 points. Prolonged prothrombin time, results in 0 points when <3 s, 1 point when 3–6 s and 2 points when ≥6 s. Fibrinogen levels ≥1 g/L result in 0 points, <1 g/L in 1 point. D-dimer levels of <0.4 µg/mL account for 0 points, levels between 0.4 and 4 µg/mL account for 2 points and D-dimer levels >4 µg/mL account for 3 points. A score ≥5 is compatible with overt DIC.

Statistical analysis

Propensity score matching (PSM) was performed (1:6 for thrombocytopenia versus non-thrombocytopenia patients), with age, sex and days from symptom onset to admission used for propensity score by the nearest-neighbor method with a caliper of 0.2 standard deviation. The standardized mean difference (SMD) was measured to determine the balance between the two groups before and after PSM, with its cut-off value of <0.2 regarded as indicating an adequate balance between two groups. The inter-group comparison of patients’ characteristics was made for the matched patients to ensure the two groups were well balanced.

Descriptive statistics were performed for categorical variables as frequencies and proportions or rates, as well as for continuous variables were expressed as medians and interquartile ranges. A Mann–Whitney U test or a Kruskal-Wallis test was used to compare the differences of continuous nonparametric variables between different groups. Pearson chi-square test or Fisher exact test was performed to compare the differences of categorical variables. Kaplan–Meier method was used to assess the survival rates, and the differences between groups with different platelet counts were analyzed by the log-rank test. Hazard ratio (HR) with 95% confidence interval (CI) were estimated using the Cox regression model, adjusting for baseline variables. Cox regression model was used to perform multivariable analysis between patients with early and late development of thrombocytopenia at 1–28 days and 28–56 days, respectively. A generalized estimating equation (GEE) was used to perform the comparisons on the levels of the parameters among two groups due to the repeated measurements in the same patient. Odds ratio (OR) with 95% CI were estimated by multivariable logistic regression model, to investigate the association between groups with different platelet counts and fatal outcome. Statistical analysis was performed using R software (version 3.6.3; R Core Team, 2020), and a two-sided P value of less than 0.05 was considered statistically significant.

Patient recruiting and demographic information

From 3 January to 10 March 2020, 2346 hospitalized COVID-19 patients were enrolled, among whom 137 patients were excluded because of incomplete or missing data on the platelet count or other important parameters during hospitalization. Among the remaining 2209 COVID-19 patients available, 127 (5.7%) were recorded with thrombocytopenia, including 34 females and 93 males, with median age of 67 (IQR 56–75) years old. Compared with 2082 (94.3%) patients with PLT ≥ 100 × 10⁹/L on admission, the thrombocytopenia patients were older, more likely to be male gender, and with higher presence of hypertension, clinical manifestation of fever, fatigue and anorexia (Table S1).

For the thrombocytopenia group, the median PLT counts were kept below the normal range throughout the observation, with the nadir level observed at three weeks after disease and increased thereafter until attaining near 100 × 10⁹/L at four weeks after disease. For the non-thrombocytopenia patients, the median PLT counts were kept within normal range during the whole observation, and with the peaking counts observed at 15 and 18 days after disease, respectively (Fig. 1A).

The PLT counts were higher in survival patients than in fatal patients by the GEE model, after adjusting the variables of age, sex, delay from symptom onset to admission and comorbidities (P < 0.001, Fig. S1A). However, the patients with comorbidities had comparable levels of PLT counts at the different stages to those without (Figs. S1B–S1E).

The comparison of clinical features based on the PSM cohort

We used PSM based on age, sex and days from symptom onset to admission to create a matched group with PLT within normal range on admission (729 patients) in Table 1. The age, sex, and days from symptom onset to admission were comparable after matching (all SMD < 0.2). Pre-existing hypertension, diabetes and coronary heart disease were non-significantly different between the two groups. The major symptoms were fever, cough, fatigue, anhelation, anorexia, and chest tightness, among which fever, fatigue and anorexia were reported with significantly higher frequencies in the thrombocytopenia group than the non-thrombocytopenia group (P = 0.019, P = 0.024 and P = 0.007, respectively).

Comparison of all-cause mortality and detrimental complications based on the PSM cohort

Patients with thrombocytopenia developed a higher prevalence of respiratory failure (41.9% vs. 22.6%), ICU entrance (25.6% vs. 11.5%), acute heart injury (9.3% vs. 5.8%), acute kidney injury (4.7% vs. 2.9%), septic shock (9.3% vs. 2.9%), and DIC (45.2% vs. 10.6%) compared to the matched non-thrombocytopenia group. The difference remained significant for respiratory failure, ICU entrance and DIC after adjusting the effect from age, sex, days from symptom onset to admission and comorbidities (Table 2). However, there was no significant difference in the risk of acute heart injury, acute kidney injury, and septic shock between thrombocytopenia group and non-thrombocytopenia group. These results were in line with those obtained from the database without PSM (Table S2).

The mortality rate was increased from 18.4% (134/729) in the non-thrombocytopenia group to 47.2% (60/127) in the thrombocytopenia group. After adjusting for age, sex, days from symptom onset to admission and comorbidities, the HR of the all-cause mortality in the thrombocytopenia group was significantly increased (HR 3.08, 95% CI [2.26–4.18], P < 0.001) compared with non-thrombocytopenia (Fig. 1B).

Comparison of haematological disorders based on the PSM cohort

The inter-group comparison was made at three clinical stages separately, i.e., 1–10 days, 11–20 days and 21–30 days after symptom onset, revealing that monocyte counts, monocyte percentage and basophil percentage were decreased in thrombocytopenia vs. non-thrombocytopenia patients at all three stages (Figs. 2E, 2F, 2I and Table S3). Moreover, significantly higher neutrophil counts and lower levels of lymphocyte percentage were observed for thrombocytopenia over non-thrombocytopenia group at 11–20 days and 21–30 days after symptom onset (Figs. 2B and 2C). Eosinophil counts and percentage at both 1–10 days and 21–30 days after disease onset were significantly lower in the thrombocytopenia group than in the non-thrombocytopenia group (Figs. 2H and 2G). The remaining indicators showed no inter-group difference, or only difference at the last clinical stage, which was less likely to be related to thrombocytopenia (Figs. 2A, 2D and 2J).

Comparison of platelet morphology indexes and coagulation perturbation based on the PSM cohort

During the whole hospitalization, we evaluated platelet morphological indexes, which reflect the etiology of thrombocytopenia, and there were significant differences in two indexes between the two groups (Figs. 3A–3B). Mean platelet volume (MPV) increased, whereas the thrombocytocrit decreased, which corresponded with the reduction of platelet count.

We also evaluated clotting profile and found a series of important coagulation factors with significant differences, including prothrombin time (PT), activated partial thromboplastin time (APTT), thrombin time (TT), fibrinogen (FIB), and fibrinogen degradation products (FDP) (Figs. 3C–3G). The thrombocytopenia group had higher levels of APTT and FDP at 1–10 days and 11–20 days after symptom onset, longer PT at 1–10 days, 11–20 days and 21–30 days, and higher level of TT at 11–20 days when compared with the non-thrombocytopenia group. Lower level of FIB was observed at 1–10 days and 11–20 days after symptom onset in the patients with thrombocytopenia.

Comparison of inflammatory markers and biochemical parameters based on the PSM cohort

Elevated levels of inflammation markers, including interleukin 6 (IL-6), tumor necrosis factor α (TNF-α), C-reactive protein (CRP) and procalcitonin, were observed at 1–10 days or 11–20 days after symptom onset in the patients with thrombocytopenia (Fig. 4). Especially for IL-6, the patients with thrombocytopenia had persistently elevated levels at 1–10 days, 11–20 days and 21–30 days after symptom onset, with their difference from the non-thrombocytopenia group further enlarged, which were significant at 1–10 days and 11–20 days after symptom onset. Higher level of CRP was demonstrated in patients with thrombocytopenia at 1–10 days, 11–20 days and 21–30 days after symptom onset, and with significant difference at 11–20 days compared to the non-thrombocytopenia group. The TNF-α and procalcitonin were elevated to higher levels in the thrombocytopenia group, with the difference attaining significance at 1–10 days compared to the non-thrombocytopenia group.

The levels of albumin (ALB) and total protein (TP) were reduced below normal range in the thrombocytopenia patients at 11–20 days or 21–30 days after symptom onset. The lactate dehydrogenase (LDH) was augmented above normal range for both groups, and to higher levels in the thrombocytopenia patients at 1–10 days and 11–20 days after symptom onset. The urea was exclusively elevated in the thrombocytopenia group, to a significantly higher level than in the non-thrombocytopenia group at 11–20 days and 21–30 days after symptom onset. Lower level of creatine kinase-MB (CK-MB) was observed at 21–30 days after symptom onset in the patients with thrombocytopenia (Fig. 5).

Demographic information, outcomes and laboratory findings of patients with late development of thrombocytopenia

Altogether 61 patients had late development of thrombocytopenia (the first observation of thrombocytopenia seen beyond 14 days post-symptom) (Table S4). Altogether 184 developed early thrombocytopenia (the first observation of thrombocytopenia seen within 14 days post-symptom). These two groups added up to a total frequency of 11.1% during the whole disease progress. Comparison between the two groups revealed older age with non-significant difference, significantly longer delay between disease and hospital admission, as well as longer duration of hospitalization in the late-thrombocytopenia group. The patients with late development of thrombocytopenia demonstrated higher levels of neutrophil counts at 21–30 days, CRP at 11–20 days and 21–30 days after symptom onset compared with patients with early development of thrombocytopenia (Figs. S2A and S2E). Lower levels of lymphocyte counts, INR and procalcitonin were demonstrated from the patients with late development of thrombocytopenia (Figs. S2B, S2D and S2F). Thrombocytocrit had a higher level in late development of thrombocytopenia group at 1–10 days and 11–20 days after symptom onset (Fig. S2C).

The nadir of PLT counts was observed at three weeks after disease for early thrombocytopenia and late thrombocytopenia (Fig. S3A). The all-cause mortality rate was comparable between early thrombocytopenia and late thrombocytopenia; however, survival analysis revealed a lower mortality rate at 21 days post symptom onset in the patients with late thrombocytopenia than early thrombocytopenia (21.3% vs. 33.7%), which trend, however, was reversed from 28 to 56 days post symptom onset (30.8% and 5.4% for patients with late and early thrombocytopenia, respectively) (Fig. S3B). This difference remained significant after adjusting the effect from age, sex, days from symptom onset to admission and comorbidities at 1–28 days (HR 0.43, 95% CI [0.24–0.75], P < 0.001) and 28–56 days (HR 4.89, 95% CI [1.91–12.55], P = 0.001).