Forty Postmortem Examinations in COVID-19 Patients

Two Distinct Pathologic Phenotypes and Correlation With Clinical and Radiologic Findings

Simona De Michele, MD; Yu Sun, MD, PhD; Mine M. Yilmaz, MD; Igor Katsyv, MD, PhD; Mary Salvatore, MD; Amy L. Dzierba, PharmD; Charles C. Marboe, MD; Daniel Brodie, MD; Nina M. Patel, MD; Christine K. Garcia, MD, PhD; Anjali Saqi, MD, MBA


Am J Clin Pathol. 2020;154(6):748-760. 

In This Article


In our cohort, three major pulmonary patterns were identified: ALI in 29 (73%) of 40, IFPAs in 36 (90%) of 40, and VCHL in 20 (50%) of 40. Second, there was a novel finding of non-ALI, which suggested two main pulmonary pathologic phenotypes—ALI and non-ALI. Each major finding was reviewed independently and in the context of the others to assess for correlation among them.

In patients with COVID-19, DAD has been described extensively[1–3,5,7,9,12–16] and AFOP in five (83.3%) of six cases in a single small series.[11] Meanwhile, an absence of DAD (non-ALI) has been described rarely[6,11,12] and infrequently correlated with clinical or imaging findings.[12]

ALI vs Non-ALI

For these reasons, while ALI and non-ALI were evaluated in parallel for similarities and differences, the primary focus was on the non-ALI phenotype. First and foremost, the presence of the two distinct pathologically identified phenotypes was corroborated with available data. Although imaging was limited to primarily chest radiographs and performed only for a subset (n = 6) of non-ALI decedents (due in large part to early death in that cohort), the overall concordance between radiology and pathology was 93.8%, which argues against undersampling of ALI. Furthermore, when present microscopically, ALI was evident in most H&E slides (mean, 78.4%), also making undersampling an unlikely cause for its absence in the non-ALI cohort. Although it can be unevenly distributed, DAD in COVID-19 has been described bilaterally and in all lobes.[16]

Clinical, laboratory, and radiologic data were assessed for distinct patterns that correspond to the microscopically identified ALI and non-ALI phenotypes. There was no difference in O2 saturation at presentation, dyspnea, and Tmax between ALI and non-ALI. There was a statistically significant difference in the length of the hospital course between ALI (mean, 13.2 days) and non-ALI (mean, 3.0 days) decedents—the latter were hospitalized, if at all, for fewer days. Radiographically, patients with ALI were significantly more likely to have higher grade findings, corresponding to consolidation, on chest radiographs compared with the non-ALI group. The abbreviated hospitalization in the non-ALI group resulted in limited clinical and laboratory data for comprehensive assessment of premortem parameters. With the exception of two cases, which showed a macroscopic pulmonary embolism or intracranial hemorrhage, a definitive pathologically confirmed cause of death for the non-ALI decedents was not identified. Neither myocarditis nor myocardial infarction, which typically causes microscopic change 4 to 12 hours following the insult,[17] was microscopically evident.

The presence of a non-ALI phenotype in our cohort and the rarity of its descriptions in the literature raise important questions. First, was mortality in these patients directly attributable to COVID-19 or an alternative cause? The cause of death in most non-ALI patients was attributed to cardiac arrest or cardiopulmonary failure. COVID-19 as the predisposing condition for cardiac arrest is plausible, explains the absence of microscopic cardiac findings, and is supported by a 10-fold increase in New York City[18] and a 52% increase in four provinces in Italy[19] in out-of-hospital cardiac arrests during the pandemic relative to similar timeframes in 2019.

Second, did the non-ALI decedents have minor microscopic pattern(s) contributing to the cause of death? They had minor findings, but these were improbable causes of death—one decedent had focal bronchopneumonia comprising less than 5% of sections; none had microorganisms or microscopically evident underlying interstitial or other predisposing lung disease. Third, is a non-ALI phenotype a precursor to ALI? While this remains a possibility, it is less likely based on the imaging and microscopic features. Most patients with ALI had consolidation on imaging at the time of presentation; only two patients with ALI demonstrated progression of disease between the first chest radiograph (no consolidation) and the last chest radiograph (marked consolidation) over a span of 27 and 29 days. Furthermore, additional microscopic features that are present in the (early) exudative phase of DAD, such as interstitial edema, were not evident. Fourth, the emergence of a second phenotype may reflect larger case numbers in our cohort. Last, the fact that this phenotype has been infrequently reported in the literature may be explained by a selection bias; in other centers, patients who were not hospitalized or only so for a short time may not have undergone postmortem examinations.

Major Microscopic Patterns

Of the three major findings, microscopic VCHL was the least common. VCHL was present in all cases of non-ALI, while it negatively correlated with ALI. Vascular congestion has been described in conjunction[14] with and without DAD[6,14] in a minority of lungs from patients with COVID-19. Relative to the images accompanying published reports, the extent of vascular congestion and accompanying alveolar wall thickening with hemangiomatosis-like change in our cohort was more prominent (Image 2). Angiogenesis has been seen in lungs of patients with COVID-19 in association with DAD.[15] The presence of thickened alveolar septa and congested capillary channels resembling a hemangiomatosis-like proliferation in our cases could be ascribed to a similar angiogenesis-type process. Other possibilities remain, such as microscopic IFPAs resulting in secondary capillary proliferation.

Overall, pulmonary IFPAs, small and large, represented the most common (90%) of the three microscopic patterns. There was no statistical correlation between the type of IFPAs (macroscopic or microscopic; pulmonary or extrapulmonary) and ALI or non-ALI phenotype. Increased macroscopic venous thromboemboli[12] and microscopic fibrin (thrombi)[7,10,13,14] have been detected in decedents with COVID-19. With the aid of ancillary studies, significantly greater numbers[15] of thrombi have been identified in lungs infected with COVID-19 relative to those with H1N1. Compared with the control cases, the COVID-19 lungs did not have increased IFPAs; however, our number of controls was relatively small and quantitated only on H&E stain. Similar to the identification of a non-ALI pattern, a key question is whether IFPAs are a manifestation of COVID-19 or related to other underlying comorbidities.

Macroscopic and microscopic pulmonary thrombi are well described in the DAD pathology literature. DAD, which is the histologic pattern ascribed to multiple possible etiologies with infection representing one, causes injury to the alveolar-capillary barrier and initiates multiple downstream effects, including proinflammatory and procoagulation pathways. In fact, fibrinogen and complement, among other proteins and cellular debris, are present in hyaline membranes of DAD,[20] and thromboemboli have been identified in 95% of lungs from 22 postmortem non–COVID-19 examinations using specimen arteriography,[21] as well as in our pre–COVID-19 postmortem controls. Therefore, pulmonary IFPAs may at least to some extent be a consequence of DAD than injury resulting from COVID-19 alone.

Nevertheless, COVID-19 as a culprit of IFPAs—either emboli or thrombi—remains a possibility, since all decedents without DAD/ALI in our study also had IFPAs. In their examination of COVID-19 cases without significant DAD, Magro et al[10] noted fibrin deposits in the microvasculature of the lung and skin and suggested activation of complement pathways. Wichmann et al[12] identified deep venous thrombosis (DVT) in both lower extremities in seven (58.3%) of 12 postmortem examinations of COVID-19 cases; six of these patients had DAD. Relatively small numbers of DVTs identified in other postmortem series, including ours, may reflect differences in prosection practice—rigorous search for DVTs in all cases vs only those with indicators on external examination. It remains to be determined if IFPAs represent thrombi or emboli. Most important, understanding the etiology of IFPAs has the potential to affect clinical management since in some scenarios, such as DAD, patients do no typically receive systemic anticoagulation. Three of five ALI decedents had macroscopic pulmonary or extrapulmonary IFPAs on systemic anticoagulation (Supplemental Table 3).

COVID-19: Mechanisms of Decompensation

Our findings support two phenotypes—ALI and non-ALI. In contrast to non-ALI, ALI is concordant with the literature and expected clinically. Non-ALI probes the understanding of COVID-19, including if it is related to COVID-19. Our data support that COVID-19 presents with primarily pulmonary, cardiac, or vascular manifestations Figure 3. Cardiac arrest is a consideration based upon no or minimal hospitalization, absence of microscopic changes, and the recently reported significant[18,19] increases in out-of-hospital cardiac deaths during the pandemic. The coexistence of VCHL and IFPAs in all (11/11) non-ALI cases (and only 7/29 ALI cases) suggests, rather than a nonspecific postmortem feature, the non-ALI phenotype is associated with COVID-19 infection. IFPAs raise the possibilities of microemboli from undetected DVTs, microthrombi in the pulmonary vasculature, and microthrombi secondary to coagulopathies that accompany infections. Interestingly, while the cause of death may be attributable to one or more of the above conditions, there were no significant differences in clinical parameters, such as O2 saturation at presentation, dyspnea, and Tmax, between the ALI and non-ALI patients. Understanding the specific mechanism(s) may potentially alter clinical management.

Figure 3.

Potential mechanisms of decompensation in non–acute lung injury phenotype that include a vascular or cardiac pathway.