The sources utilized to conduct this systematic review and meta-analysis were articles published in five English databases, namely, PubMed, EMBASE, Web of Science, Science Direct, and Google Scholar and four Chinese databases, namely, the Wanfang Database, Chongqing VIP Database, Chinese National Knowledge Infrastructure (CNKI), and SinoMed. To widen the search coverage, the International Clinical Trials Registry Platform and the Chinese Clinical Trial Registry were also utilized to identify unpublished trials in this context. Literature published up to 4 September 2021, with an updated search performed on 20 March 2022, was selected. The additional relevant trials identified from the references of systematic reviews and eligible studies were also retrieved. The comprehensive search strategy was conducted using two categories of keywords for PubMed, and the search terms were modified to suit other databases: (“Shenfu injection” OR “SFI” OR “Shenfu” OR “Shenfu decoction” OR “SF injection” OR “SF” OR “ginsenoside” OR “aconite total alkaloids” OR “Panax ginseng C. A. Meyer” OR “Radix Aconitum carmichaeli”) AND (“coronavirus disease 2019” OR “COVID-19” OR “severe acute respiratory syndrome coronavirus 2” OR “SARS-CoV-2” OR “coronavirus” OR “novel coronavirus” OR “nCoV” OR “2019-nCoV” OR “severe pneumonia” OR “Severe Acute Respiratory Syndrome”). The Chinese databases were also searched by the abovementioned search terms, translated into Chinese. Search strategies for selecting the keyword, title, or abstract fields were different, referring to the specific databases. There were no language, date, or publication status limitations for the inclusion of eligible studies to reduce the risk of publication bias. Detailed search strategies were presented in the additional file.

The study screen and the selection process were performed independently by two investigators (SS and DL). After removing duplicate articles, the titles, abstracts, and full texts were screened, applying the selection criteria to select potentially eligible articles. The differences in data screening and selection were resolved through panel discussions or settled by a third reviewer (FW) to resolve doubt or disagreement. Given that randomized controlled trials (RCTs) on the optimal treatments for older patients with severe COVID-19 were lacking and clinical trials investigating SFI for severe pneumonia including COVID-19 may be limited, RCTs and quasi RCTs were both considered to be included in our study, due to the urgent nature of the COVID-19 pandemic. Finally, RCTs or quasi RCTs studying the effects of SFI on the elderly with severe pneumonia were included. Table 1 presented the Participants, Intervention, Comparators, Outcomes, and Study Design (PICOS) criteria.

The primary outcome measures of interest were the total effective rate and the Acute Physiology and Chronic Health Evaluation (APACHE) II score predicting disease severity, mortality, and safety profiles. The clinical efficacy was defined based on Infectious Diseases Society of America/American Thoracic Society Consensus Guidelines on the Management of Community-Acquired Pneumonia in Adults (Mandell et al., 2007); markedly effective: the clinical symptoms such as fever, cough, and chest tightness were significantly alleviated, and all vital indicators returned to normal levels; effective: clinical symptoms were relieved, and most indicators returned to normal levels; the total effective rate = [(number of markedly effective cases + number of effective cases)/number of cases] × 100%. Safety concerns mainly included dyspnea, nausea, vomiting, adverse cardiovascular events (arrhythmia, unstable angina, abnormal blood pressure, and myocardial infarction), and coma that occurred in severe pneumonia. In addition, predictors associated with COVID-19 disease progression, namely, BNP, CK, stroke volume (SV), cardiac output (CO), cardiac index (CI, CO indexed to body surface area), left ventricular ejection fraction (LVEF), soluble endothelial selectin (sE-selectin), APTT, vWF, D-Dimer, platelet counts, C-reactive protein (CRP), white blood cell (WBC) count, PCT, partial pressure of arterial oxygen (PaO₂), and lactic acid accumulation were the secondary outcome measures of interest.

Included studies were supposed to meet inclusion criteria as follows: 1) RCTs or quasi RCTs including prospective cohort studies; 2) elderly patients: the trials reported that the patients included were elderly patients with average age ≥ 60; 3) patients diagnosed with severe pneumonia with one of the vital criteria (septic shock or mechanical ventilation) or at least three of the secondary criteria including arterial oxygen pressure/fraction of inspired oxygen (PaO₂/FiO2) ratio < 250 mmHg, respiratory rate > 30 breaths/min, WBC count < 4 × 10⁹, blood urea nitrogen level ≥ 30 mg/dl, multilobar infiltrates, platelet count < 10 × 10⁹, confusion, hypotension requiring aggressive fluid resuscitation, or core temperature < 36°C (Mandell et al., 2007); 4) comparison of SFI as an adjuvant with standard control alone; SFI could be delivered via an intravenous drip, intravenous injection, or pumps; 5) trials that investigated at least one of the outcomes of interest.

Exclusion criteria were as follows: duplicated studies; articles with unavailable data; retrospective cohort studies, crossover trials, cluster-randomized control studies, and the RCTs with factorial design or sequential design; the effects of SFI as an adjuvant could not be evaluated because of the involvement of other TCM treatments.

For each included article, two scholars (SS and ZW) independently and simultaneously collected information on participant characteristics, study procedure, and primary outcomes using a standardized, pilot-tested data collection form and then crosschecked. The following items were collected: 1) first author’s name and publication date; 2) study location; 3) sample size; 4) subject characteristics (patient age, gender); 5) intervening measure and the source; 6) dose of SFI; 7) treatment duration; 8) outcome measures. A third researcher (FW) checked the collected data and resolved any discrepancies.

The risk of bias in the included studies through Cochrane Collaboration’s tool was independently assessed by two reviewers (SS and BC) in the following seven aspects: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias (Higgins et al., 2011). A third reviewer (WL) independently repeated the data quality assessment to settle the disagreements when necessary. The statistical analyses were conducted using RevMan 5.4 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen).

To analyze the effect size for each outcome parameter, continuous variables were shown as weighted mean difference (WMD) with 95% confidence intervals (95% CI), whereas dichotomous data were calculated as the risk ratio (RR) with 95% CI. The data utilized the standard mean difference (SMD) to reduce the difference, if there was a difference in the measurement unit (Riley et al., 2011). The random-effect model or fixed-effect model was adopted to analyze the pooled statistics. The selection of the effect model was according to the heterogeneity test: the random-effect model was utilized if statistical heterogeneity was identified (I² > 50% or p < 0.1); conversely, a fixed-effect model was adopted in the absence of substantial heterogeneity (I² ≤ 50% and p ≥ 0.1) (Higgins et al., 2003). p values < 0.05 were identified to be statistically significant. Subgroup analysis was performed based on the study design. Sensitivity analysis was carried out to identify possible outlier studies with a high risk of bias or industry funding, which may affect our results by excluding one study each time from the pooled analysis. Publication bias was measured by the Eggers test and analyzed using funnel plots when ≥ 10 studies were included (Egger et al., 1997; Sterne and Egger, 2001). Rev Man 5.4. was used to conduct the data analysis. The certainty of the evidence was assessed by GRADEprofiler (version 3.6).

After an initial search from the five English and four Chinese databases was performed, 4,824 articles were identified. Of these, 3,111 articles were eliminated due to duplicate collections. About 1,573 publications were further precluded since they were non-clinical studies. After screening the titles and abstracts of studies for potential trials, a total of 38 possibly relevant studies were obtained for full-text screening. Then, 18 articles were eliminated for the following reasons: not elderly patients, the intervention in the SFI group or control group did not meet the inclusion criteria, or have inappropriate outcome measures (Additional file). Finally, 20 clinical trials published from 2013 to 2022 were eligible (Figure 1).

The included studies were published in Chinese or English, with a total of 1,909 patients. There were 977 patients in the experimental group and 932 patients in the control group among these 20 clinical trials. Patients were divided into the SFI group (SFI plus standard control) and control group (standard control) with no significant difference between these two groups in general information, such as age, gender, body temperature, respiratory rate, heart rate, the inflammatory mediator, course of the disease, or comorbidities. The SFI, prepared according to the Chinese pharmacopeia, was obtained from Ya’an Sanjiu Pharmaceutical Co., Ltd. or Shenzhen Huarun Sanjiu Pharmaceutical Trade Co. Ltd., with chemical analysis based on previous high-performance liquid chromatography (HPLC) research (Table 2), and no studies were funded by the industry.

To evaluate the risk of bias in the involved studies, Cochrane Collaboration’s risk of bias assessment tool was utilized and the results were shown in Figure 2. There were seven trials describing the detailed stochastic methods used, and all used completely random number tables to generate the allocation sequence (Li et al., 2015a; Li et al., 2015b; Chen et al., 2016; Wang et al., 2017; Sun et al., 2018; Li, 2019; Liu, 2020a); while eight studies mentioned random sequence generation without the specific random method (Lin et al., 2013; Zhi et al., 2014; Dong et al., 2017; Lv et al., 2017; Qu and Shao, 2018; Li et al., 2019; Shi et al., 2021a; Kong et al., 2022). The blinding of subjects, researchers, or outcome assessment was not reported in all eligible articles. Whether prior protocols were designed was not reported in most trials. All involved articles showed a low risk of bias on incomplete outcome data.

Sensitivity analysis was implemented for the outcome with heterogeneity, and relevant results have been demonstrated in the individual section above. In addition, we tested the robustness of all results. As a result, the conclusions for most outcomes, including total effective rate, BNP, APACHE II score, sE-selectin, Vwf, APTT, platelet counts, PaO₂, lactic acid accumulation, WBC, CRP, and PCT, were stable because the combined RR, WMD, or SMD of overall risk estimates were consistent, without reversion after deleting a single study each time. Therefore, the sensitivity analyses based on those outcomes gave results very close to the main results, verifying that SFI may be beneficial for improving total effective rate, BNP, APACHE II score, sE-selectin, Vwf, APTT, platelet counts, PaO₂, lactic acid accumulation, WBC count, CRP, and PCT. Nonetheless, the sensitivity analysis of D-Dimer demonstrated that those results were not robust enough and needed to be further verified since the individual study significantly affected pooled results.

Given that none of the above meta-analyses involved more than ten studies and the inclusive studies were inadequate to generate funnel plots, publications bias was not assessed to screen publication bias.

According to the GRADE approach, the certainty of findings of the effect of SFI on LVEF, SV, CO, and CI based on RCTs was high. The certainty of findings relevant to the effective rate, APACHE-II, mortality rate, safety profiles, lactic acid accumulation, CRP, WBC, CK, and platelet counts based on RCTs was moderate. The certainty of evidence about the effect of SFI on APACHE II, mortality rate, safety profiles, PaO₂, lactic acid accumulation, PCT, CRP, LVEF, SV, CO, CI, sE-selectin, and APTT based on quasi RCTs and PaO₂, PCT, BNP, sE-selectin, Vwf, APTT, and D-Dimer based on RCTs was low (Supplementary Table S3).

Managing severe pneumonia in older adults is critical for the treatment development of COVID-19 in the elderly as the elderly are disproportionately affected by severe pneumonia. Given that there are currently no evidence-based recommendations for elderly severe COVID-19 and studies on the optimal treatments for older patients with severe or fatal pneumonia are lacking, the present study is necessary. It may give some hints on these issues. Here, we briefly reviewed the SFI aimed at the older population, focusing on the evidence of safety and efficacy, and various laboratory biomarkers associated with COVID-19 disease progression. In China, SFI has been applied routinely in treating critically ill patients with fatal diseases such as shock, coronary heart disease, congestive heart failure, and trauma, with a usage history of more than 30 years for end-stage diseases (Zhang et al., 2012; Huang and Cao, 2014). SFI, which mainly consists of ginsenoside and aconite total alkaloids, has been registered in China’s State Food and Drug Administration and has undergone rigorous scientific and clinical scrutiny to establish efficacy and safety. In the present study, we have provided a complete chain of evidence to determine the effectiveness and safety of SFI on elderly patients with severe pneumonia in related clinical and experimental profiles. First, SFI might improve the total effective rate and the APACHE II score and reduce safety concerns and mortality rate. Second, improving oxygenation played a vital role in decreasing mortality, and SFI could improve PaO₂ and decrease lactic acid accumulation levels based on the present study. The primary clinical presentation of severe COVID-19 was an acute respiratory failure with extreme hypoxemia, which ultimately required mechanical ventilation (Berlin et al., 2020). A recent study revealed that elderly patients with severe pneumonia including COVID-19 were at risk for acute respiratory distress syndrome (Grasselli et al., 2020). Among elderly patients with COVID-19, severe hypoxia and lactic acid accumulation resulting from respiratory failure may increase the risk of death, contributing to poor outcomes (Peng et al., 2020). The therapeutic effect of SFI on reducing mortality and improving clinical outcomes including total effective rate and the APACHE II score may be achieved through improving gas exchange during severe pneumonia, which deserves to be investigated further on COVID-19. Third, PCT played a principal role in severe cases of COVID-19 (Bhargava et al., 2020; Hu et al., 2020), especially in COVID ICU patients (Grasselli et al., 2020). SFI might decrease PCT levels based on our studies including patients with severe pneumonia, providing indirect evidence for further trials on COVID-19. Fourth, the effect of SFI was verified from the cardiovascular system, endothelial function, and coagulation function. While elderly patients with severe pneumonia primarily suffered from lung injury with interstitial pneumonitis and severe acute respiratory distress syndrome, the cardiovascular system was also affected, especially in elderly COVID-19 patients (Guzik et al., 2020). Mechanistically, after proteolytic cleavage of its S protein by a serine protease, SARS-CoV-2 binds to the transmembrane angiotensin-converting enzyme 2 (ACE2) to enter type 2 pneumocytes, macrophages, perivascular pericytes, and cardiomyocytes, contributing to microvascular and endothelial dysfunction, plaque instability, myocardial dysfunction and damage, and myocardial infarction (Guzik et al., 2020). There has been increasing evidence that severe pneumonia, including COVID-19, may lead to severe cardiac involvement with cardiac dysfunction and elevation of myocardial injury markers including BNP and CK levels (Zheng et al., 2020), which were demonstrated to be related to poor prognosis (Rath et al., 2020). CO and cardiac SV were considered determinants of epidemic fatality (Stevens, 1976), and low SV and CO conveyed the information about the patients’ bad hemodynamic status and cardiovascular system function (Kamoi et al., 2014). Previous studies have also illustrated higher D-Dimer levels in critically ill or fatal cases, especially in patients with SARS-COV-2 infection (Zhou et al., 2020). In addition, markers of endothelial perturbation, namely, sE-selectin and vWF have been reported to be proportional to the severity of pneumonia (Cugno et al., 2021). COVID-19 could cause a hypercoagulable state and coagulation dysfunction, which were more likely to occur in severe and critically ill patients. The levels of markers of coagulation activation played a significant role in predicting the severity and prognosis of COVID-19, and APTT might be applied as indicators in predicting the mortality of COVID-19 (Long et al., 2020). SFI may alleviate cardiac injury by decreasing the BNP level, CK level, sE-selectin, vWF, APTT, and D-Dimer level and improving LVEF, SV, CO, and CI during severe pneumonia with fewer safety concerns according to the present pooled results. Although ideal treatments for severe COVID-19 have not been available, the end-stage of fatal COVID-19 is similar to the end-stage of severe pneumonia since the imminent tasks for rescuing the patients clinically are improving gas exchange, strengthening heart function, and reducing inflammation, endothelial perturbation, and coagulation dysfunction. Whether SFI might be a potentially promising choice in rescuing critical or fatal COVID-19 patients deserves further clinical trials.

The widespread use of SFI is the application of “WE” medicine. Shenfu decoction, the origin of SFI, was first recorded in “The Yan Family’s Recipe for Health” (Yanshi Jisheng Fang) in the Southern Song Dynasty (1253 AD) and was suggested in emergency treatments as the classic Chinese formula for centuries. Pneumonia in the elderly was a severe problem with clinical presentations that differed from those in younger patients. Based on TCM, the elderly people with insufficient healthy-qi and visceral deficiency were vulnerable to depletion of yin and yang (Zhang et al., 2021), which was the pathological basis of severe pneumonia and played a vital role in the development and prognosis of COVID-19, tending to result in sudden death. Since the ability of elderly patients to resist the pathogenetic factors declined, invigorating yang-qi promptly for elderly severe pneumonia patients with loss of yang was clearly warranted. With the development of integrated traditional and western medicine (Qiu et al., 2020), it was believed that SFI might be more suitable for elderly severe pneumonia patients than decoction since older patients complained of more difficulty swallowing than younger patients and were more susceptible to coma with the inability to take oral medications (J., 2000). SFI was the empirical treatment of patients with the sudden exhaustion of yang-qi in clinical practice. The elderly severe pneumonia patients might present with few respiratory symptoms and signs. Instead, they could be manifest as delirium, worsening chronic confusion, and even coma (Yoshikava and Marrie, 2000), which could be rescued by SFI, especially those with sweating, cold limbs, purplish-gray ash tongue, and pulse that is floating, large, and weak. Cardiac SV could determine the epidemic fatality (Stevens, 1976), and low SV was prevalent in critical care (Kamoi et al., 2014). The SV value in the SFI arm may be dramatically higher than that of the control arm since the severe pneumonia patients’ response to SFI therapy was positive based on our results, which might also explain why SFI improved the hemodynamic profiles.

Age-related decline such as immunosenescence and inflammaging was able to heighten the vulnerability of elderly patients to severe COVID-19 (Chen et al., 2021). Understanding the theory from modern medicine, there are antiaging active ingredients in SFI that have beneficial effects on anti-oxidation, restoration of function, and reducing the deleterious contribution of senescent cells in aged diseases (Li et al., 2016). Ginsenoside, the main component of SFI, could inhibit the activation of the Wnt/β-catenin signaling pathway and decrease the senescence of neural stem cells (NSCs) (Xiang et al., 2019).

SFI, designed for the specific clinical presentation in severe and fatal cases, has been used widely in treating critically ill patients with respiratory failure and elderly severe pneumonia. Integrating the concepts of TCM with modern medicine could be an effective approach toward the overall management of elderly severe COVID-19 in the absence of specific drugs to treat this pandemic (Huang et al., 2020; Qiu et al., 2020). The protective effects of SFI against severe pneumonia have been researched by clinical studies for multiple indications and have been extensively studied through different experimental models. Indeed, SFI involved multiple targets, and the potential mechanism might be relevant to improving antioxidant capacity and energy metabolism and delaying cell senescence (Xiang et al., 2019). SFI infusion or the main component of SFI could enhance the activity of superoxide dismutase, Ca²⁺-ATPase, Na⁺-K⁺-ATPase, and reduce the malondialdehyde content of lung tissue (Zhang et al., 2012). Recent studies have also reported that SFI may inhibit the activity of IL-6 and TNF-α and alleviate acute lung injury via reducing the expression of NF-kB (Wang et al., 2008). Furthermore, SFI can protect lung tissue, inhibit heart failure, reduce the histological damage, and play an anti-inflammatory effect by altering the balance of microRNAs involved in activating and inhibiting apoptosis, regulating the production of cytokines, and inhibiting the expression of NF-kB (Huang and Cao, 2014; Yan et al., 2018). Given that the lung was responsible for pulmonary gas exchange, these pulmonary protective effects of SFI may simultaneously contribute to the recovery of vital organs including the heart and brain, thereby improving severe and fatal cases (Zhang et al., 2012). The molecular mechanisms by which SFI treats the elderly with severe COVID-19 remain elusive and merit exploration in the future.

A critical issue regarding the empirical treatment of SFI is that physicians ought to pay careful attention to safety and drug interactions. The most frequent safety concerns reported in included studies were adverse cardiovascular events (Ren, 2020). Although negative symptoms of the SFI group were dramatically less than those of the control group according to our results, the safety of the intervention should not be neglected. Based on the eighteen incompatibilities and nineteen medicaments of mutual restraint in TCM, SFI should not be used with Pinellia ternata (Thunb.) Makino, Trichosanthes rosthornii Harms, Trichosanthes kirilowii Maxim, Fritillaria usuriensis Maxim, Fritillaria cirrhosa D. Don, Fritillaria pallidiflora Schrenk, Fritillaria thunbergii Miq., Fritillaria monantha Migo, Fritillaria unibracteata P. K. Hsiao & K. C. Hsia, Ampelopsis japonica (Thunb.) Makino, or Bletilla striata (Thunb.) Rchb. F. Nevertheless, toxicological data on the drug interactions are inadequate, and the mechanisms are still largely unknown, which may lead to harming patients (Heinrich et al., 2020). Therefore, high-quality findings on drug interactions are warranted for the future benefit of severe pneumonia patients and for the development of SFI.

Considering that older patients generally have dysfunction of multiple organs due to the aging process and comorbidity, physicians are supposed to pay attention to drug dosages and optimal total duration in this group. Apart from giving SFI based on patient illness states and physical conditions by convention, 7–14 days was the usual duration of therapy based on the present systematic review. However, this review did not identify the best dosages and optimal total duration of SFI for elderly patients with severe COVID-19, which has not been explored in available prospective studies. Clinical trials are supposed to be conducted to determine if a longer duration is necessary or beneficial for elderly patients. In addition, severe pneumonia in elderly patients is often combined with multiple diseases. Subgroup analysis may be conducted in the future to assess the therapeutic effect of SFI on a single disease. Moreover, the impact of risk of bias on the certainty of evidence should not be underestimated, which was considered and interpreted in the table generated by the GRADEprofiler, leading to downgrading of the evidence for corresponding outcomes. Additionally, the heterogeneity of a few outcomes was an unavoidable issue, and definitive conclusions about heterogeneity were hard to be made. We have investigated the main sources of heterogeneity, which may be study design, based on sensitivity analysis. For instance, the heterogeneity of PCT and BNP brought by Qu’s RCT (Qu and Shao, 2018) was probably because the age range was different from that in other included studies, and the heterogeneity of PCT brought by Kong’s study (Kong et al., 2022) was probably due to the different conventional pharmacotherapy, and the heterogeneity of CRP may come from the risk of bias according to the opinions from clinical experts in our team (WL and FW). However, further trials specifically assessing those points were needed to confirm whether those factors contributed to the differences in the meta-analyses results. In addition, international guidelines regarding the use of SFI in elderly patients with severe pneumonia from different regions and ethnicities are not available. The cooperation of various countries is required to investigate the effect of SFI since the metabolism of SFI might be different in various races. Furthermore, the results and conclusion about a few outcomes with limited sample size might be changed in the future since the understanding of SFI treating severe pneumonia has been evolving continuously. Therefore, we intend to explore these unstable outcomes in our future research. According to the National Health Commission of the People’s Republic of China (PRC) guidance (The Chinese National Health Commission, 2020), SFI plus standard control in severe pneumonia was recommended based on clinical practice but without an evidence-based reference for the rational use of SFI. The present study may fill the gap in evidence-based medicine. Nonetheless, much remains to be learned about the treatment for elderly severe COVID-19, including the minimal clinically important difference (MCID), which is unknown for most outcomes, and whether the magnitude of these effects represents MCID has not been investigated. MCID, an approach to assess the clinical significance (Bennett, 2016), has not been widely used in severe pneumonia research, affecting the results’ external validity, including the markers of endothelial perturbation and clinical relevance. We hope that more data about recommended levels for the MCID based on changes during severe pneumonia, especially COVID-19, will be available to verify the relevant clinical benefits.