Skip to main content
Download PDF

Predictive ability of hypotension prediction index and machine learning methods in intraoperative hypotension: a systematic review and meta-analysis

Journal of Translational Medicine volume 22, Article number: 725 (2024) Cite this article

Abstract

Introduction

Intraoperative Hypotension (IOH) poses a substantial risk during surgical procedures. The integration of Artificial Intelligence (AI) in predicting IOH holds promise for enhancing detection capabilities, providing an opportunity to improve patient outcomes. This systematic review and meta analysis explores the intersection of AI and IOH prediction, addressing the crucial need for effective monitoring in surgical settings.

Method

A search of Pubmed, Scopus, Web of Science, and Embase was conducted. Screening involved two-phase assessments by independent reviewers, ensuring adherence to predefined PICOS criteria. Included studies focused on AI models predicting IOH in any type of surgery. Due to the high number of studies evaluating the hypotension prediction index (HPI), we conducted two sets of meta-analyses: one involving the HPI studies and one including non-HPI studies. In the HPI studies the following outcomes were analyzed: cumulative duration of IOH per patient, time weighted average of mean arterial pressure < 65 (TWA-MAP < 65), area under the threshold of mean arterial pressure (AUT-MAP), and area under the receiver operating characteristics curve (AUROC). In the non-HPI studies, we examined the pooled AUROC of all AI models other than HPI.

Results

43 studies were included in this review. Studies showed significant reduction in IOH duration, TWA-MAP < 65 mmHg, and AUT-MAP < 65 mmHg in groups where HPI was used. AUROC for HPI algorithms demonstrated strong predictive performance (AUROC = 0.89, 95CI). Non-HPI models had a pooled AUROC of 0.79 (95CI: 0.74, 0.83).

Conclusion

HPI demonstrated excellent ability to predict hypotensive episodes and hence reduce the duration of hypotension. Other AI models, particularly those based on deep learning methods, also indicated a great ability to predict IOH, while their capacity to reduce IOH-related indices such as duration remains unclear.

Introduction

Every year more than 300 million surgeries are conducted worldwide [1], resulting in a significant number of patients experiencing possible intraoperative complications. One of the most common complications associated with both cardiac and non-cardiac surgeries is intraoperative hypotension (IOH) [2]. In non-cardiac surgeries, IOH has been associated with myocardial injury [3], acute kidney injury [4], and death [5]. Additionally, in cardiac surgeries hypotension during cardiopulmonary bypass (CPB) has been associated with a significant risk of ischemic stroke, with the risk increasing as more time is spent in a hypotensive state. As a consequence, it is imperative that the duration of IOH be minimized during surgeries. The current model of managing IOH is mostly reactive, and the treatment often occurs with delay [6]. However, the perilous ramifications of IOH have recently pushed researchers towards more proactive approaches to its treatment.

Even though efforts have been made to identify the epidemiological factors predisposing patients to IOH in order to estimate its risk during surgeries [7, 8], they are not helpful in reducing IOH in clinical settings [9]. However, AI models have demonstrated high efficiency in predicting IOH in real-time and providing the clinician with enough time to act before the onset of a hypotensive episode. Furthermore, AI model have proven useful in predicting and preventing hypotension-induced complications, including sepsis [10] and acute kidney injury [11]. These real-time IOH prediction models have been proven so valuable that one logistic regression (LR) model in particular_ the Hypotension Prediction Index (HPI) [12]_ is now commercially available for use, and has prompted many clinical trials and validation studies testing its efficiency in predicting IOH.

As studies assessing the efficiency of AI models including HPI have been accumulating in the literature, we set out to compile the existing evidence regarding the efficacy of such models and HPI by area under the receiver operating curve (AUROC). This review aims to examine the current state of AI research in IOH prediction. Further, it aims to quantify and compare the efficiency and predictive ability of both HPI and non-HPI models.

Methods

This systematic review and meta-analysis was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [13] in search of papers developing or validating artificial intelligence methods for the prediction of intraoperative hypotension. The protocol for this review was prospectively registered on PROSPERO (CRD42024504636). Due to the high number of studies evaluating the hypotension prediction index (HPI), those studies were extracted and analyzed separately.

Search strategy

An online database search of Pubmed, Scopus, Embase, and Web of Science was conducted on December 9th, 2023 with MeSH terms and keywords synonymous with “artificial intelligence” and “intraoperative hypotension”. No publication date or language limitations were defined. A manual citation search of the included studies was also performed after finalization of the screening process.

Eligibility criteria

For the purposes of our study IOH was defined as hypotension occurring after the induction of anesthesia, regardless of the specific blood pressure cut-off that was utilized (e.g. MAP < 65 mm Hg, or MAP < 55 mm Hg) or timeframe during which IOH was detected by different studies (e.g. from tracheal intubation to incision, after incision, etc.).

Studies were included if they complied with the following PICOS:

Population: patients undergoing cardiac or non-cardiac surgeries with intraoperative blood pressure monitoring.

Intervention: predictive models utilizing artificial intelligence which are either being developed or being validated.

Comparator: standard intraoperative care or non-AI models, if applicable.

Outcomes: area under the receiver operating characteristics curve for prediction of intraoperative hypotension as defined by the study using AI models that have been developed or are being validated. Duration of hypotension as defined by the study or time-weighted average of hypotension (TWA-MAP < 65 mmHg) or area under the threshold for hypotension (AUT-MAP < 65 mmHg) for AI models under validation.

Study design: AI development papers or controlled studies validating AI models.

Papers were excluded if they did not include human participants, only used AI models for feature selection, or if they were conference abstracts.

Study selection

Studies were screened by two independent reviewers in two phases. An initial title abstract screening followed by full-text retrieval and evaluation. Discrepancies between the two were resolved by an independent third reviewer through discussion.

Data extraction

The following data were extracted into one of two pre-constructed Excel spreadsheets by two independent reviewers under the supervision of a third reviewer. The first spreadsheet contained the following data regarding studies not utilizing the hypotensive prediction index: author, year of publication, country, type of surgery (cardiac or non-cardiac), type of the source of dataset, population size, population age, population gender, predicted outcome, number of variables used in the model, mode of validation, best-performing algorithm, number of hypotensive events, number of hypotensive patients, and area under the receiver operating curve (AUROC).

The second spreadsheet included the following data regarding studies employing the hypotension prediction index: author, year of publication, country, type of study, type of surgery, type of comparator, group size, age and gender of the groups, number of intraoperative hypotensive events, number of hypotensive patients, duration of intraoperative hypotension, time-weighted average of hypotension (MAP < 65 mmHg), area under the threshold of MAP < 65 mmHg (mmHg*min), and AUROC for the hypotensive prediction index.

Outcomes

The following were chosen as the main outcomes of this review: AUROC for HPI and non-HPI studies and TWA-MAP < 65, duration of intraoperative hypotension, and AUT-MAP < 65 for HPI studies only.

Quality assessment

Quality assessment of non-HPI studies was conducted using the PROBAST tool [14], while quality assessment of studies utilizing HPI was done using the Jadad scale [15] or Newcastle-Ottowa scale [16] for controlled trials and observational studies respectively.

Statistical analysis

Meta-analysis of the duration of hypotension, TWA-MAP < 65, and AUT-MAP < 65 of the HPI studies was conducted using the Comprehensive Meta-Analysis software (CMA, version 3, NJ, USA) utilizing a random effects model with standardized mean difference (SMD) and its corresponding 95% confidence interval (CI) as the effect size. Means and standard deviations were the only accepted data entry form and medians with interquartile ranges were converted to means and standard deviations using methods outlined by Wan et al. [17] and Luo et al. [18].

Meta-analysis of the AUROC for both the HPI and non-HPI studies was done using Stata version 18 (StataCorp. 2023. Stata Statistical Software; College Station, TX, USA), employing a random effects model with the restricted maximum-likelihood method. AUROC of the HPI studies was sub-group meta-analyzed by the time before prediction of intraoperative hypotension, while the AUROC of the non-HPI studies was sub-group analyzed by algorithm type, and the definition of IOH used to construct the models. In addition, subgroup analyses were conduceted for all of the primary outcomes in both types of studies according to the quality of the studies.

Heterogeneity was assessed using the I2 statistic, with an I2 > 50% signifying substantial heterogeneity [19]. A p-value of < 0.05 was considered statistically significant. Sensitivity analysis was conducted using the leave-one-out method and publication bias was assessed using Egger’s regression test (p-value < 0.05) and funnel plot symmetry if at least 10 studies were included in the meta-analysis.

Results

The initial search yielded 1705 records, with 997 studies eligible for screening after duplicate removal. A total of 67 studies were then chosen for full-text evaluation and 43 of them were included in our review. 22 of the included studies were HPI studies, while 21 studies developed non-HPI models to predict IOH (Fig. 1).

Fig. 1
figure 1

PRISMA flowchart made using the Shiny app by Haddaway et al. [20]

Full size image

HPI studies

The 22 HPI studies were published between 2018 and 2023, 13 of which were cohort studies [21,22,23,24,25,26,27,28,29,30,31,32,33], 8 were controlled trials [34,35,36,37,38,39,40,41] and one was the model development study [12]. Almost all of the HPI studies investigated the algorithm’s efficacy in non-cardiac surgeries, with the exception of Davies et al. [29] who did not exclude cardiac surgeries and Shin et al. [32] who only included cardiac surgeries. The number of participants in each study ranged between 20 [24] to 874 [12] and the mean/median age of participants ranged between 36 [27] to around 70 [35] years old, with most studies including participants with a mean/median age between 60 and 70 years old. 4 of the HPI studies investigated obstetric surgeries and only included females [26,27,28, 34] but the rest of the studies included males and females in relatively equal proportions, except for two studies [24, 32]. (Table 1; Additional file 1: Table S1)

Table 1 Characteristics of HPI studies
Full size table

Duration of IOH

12 HPI studies [23, 26, 28, 29, 31, 32, 34,35,36,37,38,39] reported the duration of IOH for 575 HPI-guided participants compared to 574 participants in standard in-house protocols, the pooled results of which shows a significant reduction in the duration of IOH in the HPI guided participants (SMD, (95CI), p-value; -0.70, (-0.92, -0.48), < 0.001; Fig. 2) with low heterogeneity (I2 = 0%). Sensitivity analysis shows our findings to be stable (Additional file 2: Figure S1) yet significant publication bias is present (Egger’s test p-value = 0.007; Additional file 1: Figure S2).

Fig. 2
figure 2

Forest plot for the meta-analysis of hypotension prediction index studies comparing the duration of intraoperative hypotension between hypotension prediction index guided participants and participants receiving standard in-house protocols. a [33]. b [27]

Full size image

TWA-MAP < 65 mmHg

11 HPI studies [25, 28, 30, 31, 33, 34, 36, 38,39,40,41] evaluated the TWA-MAP < 65 mmHg in 550 HPI-guided individuals compared to 550 individuals receiving standard in-house protocols. Our meta-analysis of these studies shows a significant reduction in the TWA-MAP < 65 mmHg in the HPI-guided groups (SMD, (95CI), p-value; -0.63, (-0.85, -0.42), < 0.001; Additional file 1: Fig. 3) with low heterogeneity (I2 = 0%). Sensitivity analysis shows our findings to be stable (Additional file 1: Figure S3), however, significant publication bias is present (Egger’s test p-value = 0.03; Additional file 2: Figure S4).

Fig. 3
figure 3

Forest plot for the meta-analysis of hypotension prediction index studies comparing the time-weighted average of hypotension (MAP < 65 mmHg) between hypotension prediction index guided participants and participants receiving standard in-house protocols. a [33]. b [27]

Full size image

AUT-MAP < 65 mmHg

9 HPI studies [28, 30, 33, 34, 36, 38,39,40,41] investigated changes in AUT-MAP < 65 mmHg in 364 HPI-guided participants compared to 364 participants receiving standard in-house protocols. After pooling the studies together, we found a significant reduction in AUT-MAP < 65 mmHg in the HPI-guided participants (SMD, (95CI), p-value; -0.57, (-0.86, -0.29), < 0.001; Fig. 4) with low heterogeneity (I2 = 0%). Sensitivity analysis shows our results to be stable (Additional file 2: Figure S5) and publication bias was not assessed due to less than 10 studies being included in the meta-analysis.

Fig. 4
figure 4

Forest plot for the meta-analysis of hypotension prediction index studies comparing the are under the threshold for hypotension (MAP < 65 mmHg) between hypotension prediction index guided participants and participants receiving standard in-house protocols. a [33]. b [27]

Full size image

AUROC

7 of the HPI studies assessed the performance of this algorithm in the prediction of IOH 5 to 15 min before the event [12, 22,23,24, 26, 29, 32]. After pooling the studies, we found the HPI algorithm has a considerable AUROC of 0.89 (95CI: 0.88, 0.92) between 5 and 15 min prior to the event. When stratified by time to event prediction, the algorithm performs insignificantly better 5 min before the event (AUROC: 0.92; 95CI: 0.88–0.95) in comparison to 10 min (AUROC: 0.88; 95CI: 0.82–0.93) and 15 min prior to IOH (AUROC: 0.86; 95CI: 0.80–0.93; Fig. 5). Significantly high heterogeneity is present in this meta-analysis (I2 = 99%) and inside each subgroup (I2 = 99%). Sensitivity analysis shows our findings to be stable (data not shown) and publication bias was not assessed due to less than 10 studies included in the meta-analysis.

Fig. 5
figure 5

Meta-analysis of the AUROC of the hypotension prediction index at 5, 10, and 15 minutes before the occurrence of intraoperative hypotension. The overall AUROC of the hypotension prediction index for predicting intraoperative hypotension is presented at the bottom of the figure. a [25]

Full size image

Non-HPI studies

A total of 21 studies developed original AI models to predict IOH using a wide host of algorithms, including random forest (RF) utilized in 7 studies [42,43,44,45,46,47,48], logistic regression (LR) used in 7 studies [44, 45, 49,50,51,52,53], convolutional neural networks (CNN) utilized in 5 studies [9, 46, 49, 54, 55], regular neural networks used in 4 studies [45, 56,57,58], gradient boosting machine (GBM) used in 4 studies [43, 45, 46, 53], deep neural networks (DNN) used in 4 studies [46, 52, 59, 60], artificial neural networks (ANN) utilized in 3 studies [44, 50, 51], and naive Bayesian used in 2 studies [44, 45]. Other algorithms were employed in one study only and include support vector machine [45], k-nearest neighbor [45], Linear Discriminant Analysis [45], recurrent neural networks (RNN) [49], Gaussian Process [61], Bayesian Neural Network [61], and Bayesian Generalized Linear Model [61]. These algorithms used between 4 [9] and 130 [43] features. (Additional file 1: Table S2)

The non-HPI studies were published between 2008 [51] and 2024 [61] and mostly based their models on datasets obtained from local electronic medical records (EMR) [9, 42,43,44,45,46,47,48,49,50,51,52,53, 55, 56, 59, 61], with only 4 studies utilizing a validated registry [54, 57, 58, 60]. The mean age of their datasets ranged from 31 [56] to 62 [48] years old and the mean age of most datasets hovered around 50 years old. Excluding the works by Dong et al. [61] and Feld et al. [53], which only used females or included mostly males respectively, other non-HPI studies used a gender-equal dataset (Table 2).

Table 2 Characteristics of the non-HPI AI development studies
Full size table

AUROC

10 of the 21 non-HPI studies [9, 41, 44, 45, 48, 49, 52,53,54, 60] had reported the AUROC of their AI models and were subgroup meta-analyzed by the algorithm employed in their models. Our meta-analysis found an AUROC of 0.79 (95CI: 0.74, 0.83) when all algorithms were pooled together, with substantial heterogeneity (I2 = 99%; Fig. 6). Sensitivity analysis showed our findings to be stable (data not shown) yet significant publication bias was present (Egger’s test p-value < 0.01; Additional file 1: Figure S6). In addition, subgroup analyses based on the definition of hypotension revealed that significantly higher AUROC values are achieved when IOH is defined as MAP < 65, while other definitions did not differe significantly in comparison to one another. Additionally high heterogeneity was present in each subgroup (Additional file 2: Figure S7).

Fig. 6
figure 6

Meta-analysis of the AUROC of non-hypotension prediction index artificial intelligence models for the prediction of intraoperative hypotension stratified by the type of algorithm utilized. The overall AUROC of all the models for predicting intraoperative hypotension is presented at the bottom of the figure

Full size image

Quality assessment

Among the HPI studies, there were 13 cohort studies, 4 of which were judged to be of high quality, and the remaining 9 were assessed to be of unclear quality due to the absence of control groups. Out of the 8 randomized controlled trials, 4 had excellent quality according to the Jadad scale while there were some concerns about 3 studies and one had low quality. The HPI development study was also of unclear quality due to lack of information about the analyses conducted to develop the model (Fig. 7). Subgroup analyses of the HPI studies revealed that none of the measured outcomes _namely 5-, 10-, and 15-minute AUROCs, TWA-MAP < 65, duration of intraoperative hypotension, and AUT-MAP < 65_ were significantly different among studies with different qualities (Additional file 2: Figures S8-S13).

Fig. 7
figure 7

The quality of the included studies accoding to (A) the Jadad scale (RCTs), (B) the Newcastle-Ottawa scale (Cohort studies), and (C) the PROBAST tool (AI development studies)

Full size image

Among the non-HPI studies, 9 were judged to be high quality, 10 were of unclear quality and 2 had low quality. The main cause of unclear quality among the studies was the unclear status they received in the analysis domain of the PROBAST tool (Fig. 7). Subgroup analyses revealed that high-quality studies obtained significantly lower AUROCs compared to those of unclear quality (Additional file 2: Figure S14).

Discussion

Intraoperative hypotension is a common phenomenon and has been associated with major postoperative complications such as major adverse renal [62], neurological [63], and cardiac [64] events and a high rate of mortality [65]. Despite these remarkable negative sequelae, the current management of IOH is predominantly reactive and often results in losing valuable time and higher IOH periods [6, 35, 49]. To remedy this issue, proactive approaches such as prediction models_ specifically those employing AI_ have recently been developed and tested in the relevant literature. Our study is the first systematic review and meta-analysis of the performance of AI models for the prediction of IOH. We found that one specific AI model, HPI, has been studied in 21 RCTs and cohort studies that externally validated it, and hence decided to evaluate it separately. In total, HPI was shown to significantly outperform the in-house protocol for predicting hypotension by reducing the cumulative duration of IOH per patient, TWA-MAP < 65, and AUT-MAP < 65. Furthermore, the pooled AUROC of HPI among the studies was 0.89, with the model demonstrating a statistically similar performance 5, 10, and 15 min before each hypotension event. In addition, 22 studies reported on the development of AI models other than HPI. Collectively, the non-HPI models achieved an AUROC of 0.79, with models incorporating recurrent, convolutional, and deep neural networks performing the best, and SVM models performing the worst.

Hypotension prediction index

Regarding HPI, the results of our analyses are in accordance with a previous systematic review of RCTs evaluating the performance of HPI during non-cardiac surgeries. Li et al. included 5 studies, representing 461 patients, and ultimately found that the median differences of medians indicated an improvement in hypotension-related endpoints such as duration, incidence, percentage, TWA-MAP < 65, and AUT-MAP < 65 [66]. In contrast, our study was not limited to RCTs in order to gather the limited evidence in this novel subject matter more effectively, and as a result, we were also able to meta-analyze the AUROCs obtained by external validation studies of HPI, demonstrating its excellent predictive ability, particularly compared to other LR-based models. In addition, while HPI was originally designed based on invasive arterial line waveforms, many of the included reports have utilized it in combination with non-invasive arterial pressure waveforms and achieved comparable results [22, 23, 27, 28], expanding its applicability to potentially any patient undergoing surgery. It is also noteworthy that while the original HPI model was developed to predict IOH in noncardiac surgeries, Shin et al. conducted a cohort study testing the feasibility of its use in patients undergoing cardiac surgery requiring cardiopulmonary bypass, and concluded that HPI predicted IOH with a high degree of sensitivity and specificity [32]. As these results are promising, we recommend further research be conducted to examine HPI’s efficacy in cardiac surgeries.

Non-HPI AI models

With respect to the non-HPI studies, we observed an excellent prediction ability among the deep learning algorithms, such as RNN, DNN, and CNN, with the highest AUROC achieved by STEP-OP [49], a model incorporating CNN and RNN (AUROC = 0.96). In general, deep learning has shown the potential to achieve more precise predictions than traditional ML in many areas of medical research [67], as it employs layers of neurons_as opposed to manual feature extraction in traditional ML_ and thus detects more abstract and generalized connections in the data [68]. As a result, it has been speculated that deep-learning algorithms may be able to detect subtle heralding changes in the arterial waveform which could be overlooked when represented as features in traditional ML models such as HPI [49]. Overall, although our meta-analysis demonstrated that HPI is very effective and indeed a formidable opponent for other prediction models, the model has been developed using LR [12], and non-HPI deep learning models hold the promise of even more accurate projections in the future.

In addition, our subgroup analyses revealed that studies employing a cut-off of MAP < 65 mm Hg for the definition of IOH performed significantly better in comparison to reports using other definitions. This finding is consistent with previous findings, as HPI also uses this definition of IOH [12] and has achieved excellent results by doing so, as evident by our analyses. Consequently, it seems that MAP < 65 mm Hg can be deemed as the best definition for IOH for the design of future IOH prediction models.

Limitations

Our study has several limitations. While the analyses of the HPI studies were not heterogeneous, a high degree of heterogeneity was observed among the analyses of non-HPI models, and could not be explained with sub-group analyses based on AI model types or the definitions of IOH. We believe this heterogeneity can be due to the diverse nature of surgeries included, population makeup variations, and _in the cases of traditional ML models_differences in feature extraction. Second, while we were able to perform qualitative synthesis on 10 of the non-HPI studies, 12 of them could not be included due to either not reporting their AUROC and/or its confidence interval. In addition, many of the non-HPI studies were not of high quality, mainly due to underreporting in the analyses they performed to develop the model. We suggest that moving forward, all AI development studies meticulously mention the steps they take to develop the model, and also report AUROC in order to be eligible for inclusion in quantitative syntheses of meta-analysis studies. Further, to aid the design of future studies, it is advisable that features with high predictive values be discussed by the authors more extensively. Finally, while the results of our analyses regarding HPI are promising, the number of patients included in the studies is low and the study designs are far from robust. Additionally, Li et al. [66] found, the grade of evidence regarding the associated outcomes is poor. We recommend more RCTs with larger sample sizes and higher quality be conducted to remedy this issue.

Conclusion

Artificial intelligence prediction models of intraoperative hypotension hold the potential to fundamentally change the way we treat this perilous condition. HPI, the first commercially available AI prediction model of IOH, demonstrated an excellent ability to predict hypotensive episodes and hence reduce the duration of hypotension. Other AI models, particularly those based on deep learning methods also demonstrated a great ability to predict IOH, while their capacity to reduce IOH-related indices such as duration remains unclear. This systematic review provides a comprehensive overlook of these models, serving as a stepping stone for future studies that may be conducted in this field.

Data availability

All the supporting data are included in the study.

Abbreviations

IOH:

Intraoperative hypotension

AI:

Artificial intelligence

HPI:

Hypotension prediction index

TWA-MAP < 65:

Time weighted average of mean arterial pressure < 65

AUT-MAP:

Area under the threshold of mean arterial pressure

AUROC:

Area under the receiver operating characteristics curve

CPB:

Cardiopulmonary bypass

PRISMA:

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

SMD:

Standardized mean difference

CI:

Confidence interval

NP:

Not provided

NA:

Not applicable

NN:

Neural networks

DNN:

Deep neural network

ANN:

Artificial neural network

CNN:

Convolutional neural network

RNN:

Recurrent neural network

LR:

Logistic regression

RF:

Random forest

XGBoost:

Extreme gradient boosting

GBM:

Gradient boosting machine

LGBM:

Light gradient boosting machine

SVM:

Support vector machine

LDA:

Linear discriminant analysis

KNN:

K-nearest neighbor

NB:

Naïve Bayes

GP:

Gaussian process

BGLM:

Bayesian generalized linear models

BNN:

Bayesian neural network

BNNMR:

Bayesian neural network with multivariate mixed response

ABP:

Arterial blood pressure

ECG:

Electrocardiogram

EEG:

Electroencephalogram

EMR:

Electronic medical records

References

  1. Weiser TG, Haynes AB, Molina G, Lipsitz SR, Esquivel MM, Uribe-Leitz T, et al. Estimate of the global volume of surgery in 2012: an assessment supporting improved health outcomes. Lancet. 2015;385(Suppl 2):S11.

    Article  PubMed  Google Scholar 

  2. Guarracino F, Bertini P. Perioperative hypotension: causes and remedies. J Anesth Analg Crit Care. 2022;2(1):17.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Sessler DI, Khanna AK. Perioperative myocardial injury and the contribution of hypotension. Intensive Care Med. 2018;44(6):811–22.

    Article  PubMed  Google Scholar 

  4. Sun LY, Wijeysundera DN, Tait GA, Beattie WS. Association of intraoperative hypotension with acute kidney injury after elective noncardiac surgery. Anesthesiology. 2015;123(3):515–23.

    Article  PubMed  Google Scholar 

  5. Mascha EJ, Yang D, Weiss S, Sessler DI. Intraoperative Mean arterial pressure variability and 30-day mortality in patients having noncardiac surgery. Anesthesiology. 2015;123(1):79–91.

    Article  PubMed  Google Scholar 

  6. Szrama J, Gradys A, Bartkowiak T, Woźniak A, Kusza K, Molnar Z. Intraoperative hypotension Prediction-A proactive perioperative hemodynamic Management-A literature review. Medicina. 2023;59(3).

  7. Südfeld S, Brechnitz S, Wagner J, Reese P, Pinnschmidt H, Reuter D, et al. Post-induction hypotension and early intraoperative hypotension associated with general anaesthesia. Br J Anaesth. 2017;119(1):57–64.

    Article  PubMed  Google Scholar 

  8. Mayhew D, Mendonca V, Murthy B. A review of ASA physical status–historical perspectives and modern developments. Anaesthesia. 2019;74(3):373–9.

    Article  CAS  PubMed  Google Scholar 

  9. Lee S, Lee HC, Chu YS, Song SW, Ahn GJ, Lee H, et al. Deep learning models for the prediction of intraoperative hypotension. Br J Anaesth. 2021;126(4):808–17.

    Article  PubMed  Google Scholar 

  10. Chen C, Chen B, Yang J, Li X, Peng X, Feng Y, et al. Development and validation of a practical machine learning model to predict sepsis after liver transplantation. Ann Med. 2023;55(1):624–33.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Zhang Y, Yang D, Liu Z, Chen C, Ge M, Li X, et al. An explainable supervised machine learning predictor of acute kidney injury after adult deceased donor liver transplantation. J Transl Med. 2021;19(1):321.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Hatib F, Jian Z, Buddi S, Lee C, Settels J, Sibert K, et al. Machine-learning Algorithm to Predict Hypotension based on high-fidelity arterial pressure Waveform Analysis. Anesthesiology. 2018;129(4):663–74.

    Article  PubMed  Google Scholar 

  13. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Int J Surg. 2021;88:105906.

    Article  PubMed  Google Scholar 

  14. Wolff RF, Moons KGM, Riley RD, Whiting PF, Westwood M, Collins GS, et al. PROBAST: A Tool to assess the risk of Bias and Applicability of Prediction Model studies. Ann Intern Med. 2019;170(1):51–8.

    Article  PubMed  Google Scholar 

  15. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials. 1996;17(1):1–12.

    Article  CAS  PubMed  Google Scholar 

  16. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2000.

  17. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol. 2014;14(1):135.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Luo D, Wan X, Liu J, Tong T. Optimally estimating the sample mean from the sample size, median, mid-range, and/or mid-quartile range. Stat Methods Med Res. 2018;27(6):1785–805.

    Article  PubMed  Google Scholar 

  19. Higgins JPTJ, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, editors. (s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.2 (updated February 2021). Cochrane, 2021 2021 [updated 2021. Available from: training.cochrane.org/handbook.

  20. Haddaway NR, Page MJ, Pritchard CC, McGuinness LA. PRISMA2020: an R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and open synthesis. Campbell Syst Rev. 2022;18(2):e1230.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Kouz K, Monge García MI, Cerutti E, Lisanti I, Draisci G, Frassanito L, et al. Intraoperative hypotension when using hypotension prediction index software during major noncardiac surgery: a European multicentre prospective observational registry (EU HYPROTECT). BJA Open. 2023;6:100140.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Maheshwari K, Buddi S, Jian Z, Settels J, Shimada T, Cohen B, et al. Performance of the Hypotension Prediction Index with non-invasive arterial pressure waveforms in non-cardiac surgical patients. J Clin Monit Comput. 2021;35(1):71–8.

    Article  PubMed  Google Scholar 

  23. Wijnberge M, Van Der Ster BJP, Geerts BF, De Beer F, Beurskens C, Emal D, et al. Clinical performance of a machine-learning algorithm to predict intra-operative hypotension with noninvasive arterial pressure waveforms: a cohort study. Eur J Anaesthesiol. 2021;38(6):609–15.

    Article  CAS  PubMed  Google Scholar 

  24. Yang SM, Cho HY, Lee HC, Kim HS. Performance of the Hypotension Prediction Index in living donor liver transplant recipients. Minerva Anestesiol. 2023;89(5):387–95.

    Article  PubMed  Google Scholar 

  25. Grundmann CD, Wischermann JM, Fassbender P, Bischoff P, Frey UH. Hemodynamic monitoring with Hypotension Prediction Index versus arterial waveform analysis alone and incidence of perioperative hypotension. Acta Anaesthesiol Scand. 2021;65(10):1404–12.

    Article  PubMed  Google Scholar 

  26. Frassanito L, Giuri PP, Vassalli F, Piersanti A, Longo A, Zanfini BA, et al. Hypotension prediction index with non-invasive continuous arterial pressure waveforms (ClearSight): clinical performance in gynaecologic oncologic surgery. J Clin Monit Comput. 2022;36(5):1325–32.

    Article  PubMed  Google Scholar 

  27. Frassanito L, Sonnino C, Piersanti A, Zanfini BA, Catarci S, Giuri PP, et al. Performance of the Hypotension Prediction Index with Noninvasive arterial pressure waveforms in Awake Cesarean Delivery patients under spinal anesthesia. Anesth Analg. 2022;134(3):633–43.

    CAS  PubMed  Google Scholar 

  28. Frassanito L, Giuri PP, Vassalli F, Piersanti A, Zanfini BA, Catarci S, et al. Noninvasive hypotension Prediction Index versus continuous blood pressure monitoring and intraoperative hypotension. Minerva Anestesiol. 2023;89(6):603–5.

    Article  PubMed  Google Scholar 

  29. Davies SJ, Vistisen ST, Jian Z, Hatib F, Scheeren TWL. Ability of an arterial waveform analysis-derived hypotension Prediction Index to predict future hypotensive events in Surgical patients. Anesth Analg. 2020;130(2):352–9.

    Article  PubMed  Google Scholar 

  30. Schenk J, Wijnberge M, Maaskant JM, Hollmann MW, Hol L, Immink RV, et al. Effect of Hypotension Prediction Index-guided intraoperative haemodynamic care on depth and duration of postoperative hypotension: a sub-study of the Hypotension Prediction trial. Br J Anaesth. 2021;127(5):681–8.

    Article  CAS  PubMed  Google Scholar 

  31. Runge J, Graw J, Grundmann CD, Komanek T, Wischermann JM, Frey UH. Hypotension prediction index and incidence of Perioperative Hypotension: a single-center propensity-score-matched analysis. J Clin Med. 2023;12(17).

  32. Shin B, Maler SA, Reddy K, Fleming NW. Use of the Hypotension Prediction Index during Cardiac surgery. J Cardiothorac Vasc Anesth. 2021;35(6):1769–75.

    Article  PubMed  Google Scholar 

  33. Solares GJ, Garcia D, Garcia MIM, Crespo C, Rabago JL, Iglesias F, et al. Real-world outcomes of the hypotension prediction index in the management of intraoperative hypotension during non-cardiac surgery: a retrospective clinical study. J Clin Monit Comput. 2023;37(1):211–20.

    Article  PubMed  Google Scholar 

  34. Frassanito L, Giuri PP, Vassalli F, Piersanti A, Garcia MIM, Sonnino C, et al. Hypotension Prediction Index guided goal Directed therapy and the amount of hypotension during major gynaecologic oncologic surgery: a Randomized controlled clinical trial. J Clin Monit Comput. 2023;37(4):1081–93.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Murabito P, Astuto M, Sanfilippo F, La Via L, Vasile F, Basile F et al. Proactive management of Intraoperative Hypotension Reduces Biomarkers of Organ Injury and oxidative stress during elective non-cardiac surgery: a pilot randomized controlled trial. J Clin Med. 2022;11(2).

  36. Maheshwari K, Shimada T, Yang D, Khanna S, Cywinski JB, Irefin SA, et al. Hypotension Prediction Index for Prevention of Hypotension during Moderate- to high-risk noncardiac surgery. Anesthesiology. 2020;133(6):1214–22.

    Article  CAS  PubMed  Google Scholar 

  37. Schneck E, Schulte D, Habig L, Ruhrmann S, Edinger F, Markmann M, et al. Hypotension Prediction Index based protocolized haemodynamic management reduces the incidence and duration of intraoperative hypotension in primary total hip arthroplasty: a single centre feasibility randomised blinded prospective interventional trial. J Clin Monit Comput. 2020;34(6):1149–58.

    Article  PubMed  Google Scholar 

  38. Šribar A, Jurinjak IS, Almahariq H, Bandić I, Matošević J, Pejić J et al. Hypotension prediction index guided versus conventional goal directed therapy to reduce intraoperative hypotension during thoracic surgery: a randomized trial. BMC Anesthesiol. 2023;23(1).

  39. Tsoumpa M, Kyttari A, Matiatou S, Tzoufi M, Griva P, Pikoulis E et al. The use of the hypotension prediction index integrated in an algorithm of goal directed hemodynamic treatment during moderate and high-risk surgery. J Clin Med. 2021;10(24).

  40. Wijnberge M, Geerts BF, Hol L, Lemmers N, Mulder MP, Berge P, et al. Effect of a machine learning-derived early warning system for Intraoperative Hypotension vs Standard Care on depth and duration of intraoperative hypotension during elective noncardiac surgery: the HYPE randomized clinical trial. JAMA - J Am Med Association. 2020;323(11):1052–60.

    Article  Google Scholar 

  41. Yoshikawa Y, Maeda M, Kunigo T, Sato T, Takahashi K, Ohno S, et al. Effect of using hypotension prediction index versus conventional goal-directed haemodynamic management to reduce intraoperative hypotension in non-cardiac surgery: a randomised controlled trial. J Clin Anesth. 2023;93:111348.

    Article  PubMed  Google Scholar 

  42. Greenbaum NR, White ST. Machine learning predicts intraoperative hypotension from end-tidal Carbon Dioxide Measurement. Anesth Analg. 2020;130.

  43. Inada R, Doi C, Yamasaki Y, Shigeno H, Seki H, editors. Intraoperative Hypotension Prediction System by Considering Personal Lifestyle and Medical History. 2021 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events, PerCom Workshops 2021; 2021.

  44. Kang AR, Lee J, Jung W, Lee M, Park SY, Woo J, et al. Development of a prediction model for hypotension after induction of anesthesia using machine learning. PLoS ONE. 2020;15(4):e0231172.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Kendale S, Kulkarni P, Rosenberg AD, Wang J. Supervised machine-learning Predictive Analytics for Prediction of Postinduction Hypotension. Anesthesiology. 2018;129(4):675–88.

    Article  PubMed  Google Scholar 

  46. Lee J, Woo J, Kang AR, Jeong YS, Jung W, Lee M, et al. Comparative analysis on machine learning and deep learning to predict post-induction hypotension. Sens (Basel). 2020;20(16):1–21.

    CAS  Google Scholar 

  47. Lee S, Lee M, Kim SH, Woo J. Intraoperative hypotension prediction model based on systematic feature Engineering and Machine Learning. Sens (Basel). 2022;22(9).

  48. Li XF, Huang YZ, Tang JY, Li RC, Wang XQ. Development of a random forest model for hypotension prediction after anesthesia induction for cardiac surgery. World J Clin Cases. 2021;9(29):8729–39.

    Article  PubMed  PubMed Central  Google Scholar 

  49. Choe S, Park E, Shin W, Koo B, Shin D, Jung C, et al. Short-term event prediction in the operating room (STEP-OP) of five-Minute Intraoperative Hypotension using Hybrid Deep Learning: Retrospective Observational Study and Model Development. JMIR Med Inf. 2021;9(9):e31311.

    Article  Google Scholar 

  50. Lin CS, Chang CC, Chiu JS, Lee YW, Lin JA, Mok MS, et al. Application of an artificial neural network to predict postinduction hypotension during general anesthesia. Med Decis Mak. 2011;31(2):308–14.

    Article  Google Scholar 

  51. Lin CS, Chiu JS, Hsieh MH, Mok MS, Li YC, Chiu HW. Predicting hypotensive episodes during spinal anesthesia with the application of artificial neural networks. Comput Methods Programs Biomed. 2008;92(2):193–7.

    Article  PubMed  Google Scholar 

  52. Lee C, Cannesson M, Baldi P. Prediction of Postinduction Hypotension with Deep Learning. Anesth Analg. 2019;128.

  53. Feld S, Hippe DS, Miljacic N, Polissar NL, Newman SF, Nair BG, et al. A Machine Learning Approach for Predicting Real Time Risk of Intraoperative Hypotension in Traumatic Brain Injury. Anesth Analg. 2021;132(5SSUPPL):514–5.

    Google Scholar 

  54. Jo YY, Jang JH, Kwon JM, Lee HC, Jung CW, Byun S, et al. Predicting intraoperative hypotension using deep learning with waveforms of arterial blood pressure, electroencephalogram, and electrocardiogram: retrospective study. PLoS ONE. 2022;17(8 August):e0272055.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Lu F, Li W, Zhou Z, Song C, Sun Y, Zhang Y, et al. editors. A Composite Multi-Attention Framework for Intraoperative Hypotension Early Warning. Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023; 2023.

  56. Gratz I, Baruch M, Takla M, Seaman J, Allen I, McEniry B, et al. The application of a neural network to predict hypotension and vasopressor requirements non-invasively in obstetric patients having spinal anesthesia for elective cesarean section (C/S). BMC Anesthesiol. 2020;20(1):98.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Shi M, Zheng Y, Wu Y, Ren Q. Multitask attention-based neural network for intraoperative hypotension prediction. Bioengineering. 2023;10(9).

  58. Hwang E, Park YS, Kim JY, Park SH, Kim J, Kim SH. Intraoperative hypotension prediction based on features automatically generated within an interpretable deep learning model. IEEE Trans Neural Netw Learn Syst. 2023;Pp.

  59. Yoshimura M, Shiramoto H, Koga M, Morimoto Y. Preoperative echocardiography predictive analytics for postinduction hypotension prediction. PLoS ONE. 2022;17(11 November).

  60. Kim S, Kwon S, Bovik AC, Markey MK, Cannesson M. Machine learning predicting hypotension in operating room: application of deep neural networks on continuous physiologic waveforms for early prediction of intraoperative hypotension. Anesth Analg. 2023;136(4):90–1.

    Google Scholar 

  61. Dong Z, Chen X, Ritter J, Bai L, Huang J. American society of anesthesiologists physical status classification significantly affects the performances of machine learning models in intraoperative hypotension inference. J Clin Anesth. 2024;92:111309.

    Article  PubMed  Google Scholar 

  62. Shaw AD, Khanna AK, Smischney NJ, Shenoy AV, Boero IJ, Bershad M, et al. Intraoperative hypotension is associated with persistent acute kidney disease after noncardiac surgery: a multicentre cohort study. Br J Anaesth. 2022;129(1):13–21.

    Article  CAS  PubMed  Google Scholar 

  63. Wongtangman K, Wachtendorf LJ, Blank M, Grabitz SD, Linhardt FC, Azimaraghi O et al. Effect of intraoperative arterial hypotension on the risk of Perioperative Stroke after noncardiac surgery: a retrospective Multicenter Cohort Study. Anesth Analg. 2021;133(4).

  64. van Waes JA, van Klei WA, Wijeysundera DN, van Wolfswinkel L, Lindsay TF, Beattie WS. Association between Intraoperative Hypotension and Myocardial Injury after vascular surgery. Anesthesiology. 2016;124(1):35–44.

    Article  PubMed  Google Scholar 

  65. Wijnberge M, Schenk J, Bulle E, Vlaar AP, Maheshwari K, Hollmann MW et al. Association of intraoperative hypotension with postoperative morbidity and mortality: systematic review and meta-analysis. BJS Open. 2021;5(1).

  66. Li W, Hu Z, Yuan Y, Liu J, Li K. Effect of hypotension prediction index in the prevention of intraoperative hypotension during noncardiac surgery: a systematic review. J Clin Anesth. 2022;83:110981.

    Article  PubMed  Google Scholar 

  67. Garg A, Mago V. Role of machine learning in medical research: a survey. Comput Sci Rev. 2021;40:100370.

    Article  Google Scholar 

  68. Yang S, Zhu F, Ling X, Liu Q, Zhao P. Intelligent Health Care: applications of Deep Learning in Computational Medicine. Front Genet. 2021;12.

Download references

Acknowledgements

None.

Funding

There are no financial grants supporting this study.

Author information

Authors and Affiliations

  1. Cardiovascular Surgery Research and Development Committee, Iran University of Medical Sciences (IUMS), Tehran, 14665-354, Iran

    Ida Mohammadi, Shahryar Rajai Firouzabadi, Melika Hosseinpour, Mohammadhosein Akhlaghpasand, Bardia Hajikarimloo, Roozbeh Tavanaei & Sam Zeraatian-Nejad

  2. Department of Surgery, Surgery Research Center, School of Medicine, Rasool-E Akram Hospital, Iran University of Medical Sciences, Tehran, Iran

    Mohammadhosein Akhlaghpasand, Amirreza Izadi, Sam Zeraatian-Nejad & Foolad Eghbali

Authors
  1. Ida Mohammadi
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Shahryar Rajai Firouzabadi
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Melika Hosseinpour
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Mohammadhosein Akhlaghpasand
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Bardia Hajikarimloo
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Roozbeh Tavanaei
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Amirreza Izadi
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Sam Zeraatian-Nejad
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Foolad Eghbali
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

IM: Resources, data curation, formal analysis, and writing original draft; SRF: Resources, data curation, formal analysis, and writing original draft; MH: Resources, data curation, formal analysis, and writing original draft; MA: Conceptualization, project administration, writing-review and editing, and supervision; BH: Supervision, writing-review and editing; RT: writing-review and editing; AI: Writing-review and editing; SZN: Supervision, writing-review and editing; FE: Supervision, writing-review and editing.

Corresponding author

Correspondence to Mohammadhosein Akhlaghpasand.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

All the authors declare no competing interest.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Additional file 1: Table S1:

Inclusion and exclusion criteria of HPI studies. Table S2: Characteristics of the non-HPI studies.

12967_2024_5481_MOESM2_ESM.docx

Additional file 2: Figure S1: Sensitivity analysis for the meta-analysis of the duration of intraoperative hypotension between hypotension prediction index-guided participants and participants receiving standard in-house protocols. Figure S2: Funnel plot for assessing the publication bias of our meta-analysis of the duration of intraoperative hypotension between hypotension prediction index-guided participants and participants receiving standard in-house protocols. The plot is asymmetric. Figure S3: Sensitivity analysis for the meta-analysis of the time-weighted average of hypotension (MAP < 65 mmHg) between hypotension prediction index-guided participants and participants receiving standard in-house protocols. Figure S4: Funnel plot for assessing the publication bias of our meta-analysis of the time-weighted average of hypotension (MAP < 65 mmHg) between hypotension prediction index-guided participants and participants receiving standard in-house protocols. The plot is asymmetric. Figure S5: Sensitivity analysis for the meta-analysis of the area under the threshold for hypotension (MAP < 65 mmHg) between hypotension prediction index-guided participants and participants receiving standard in-house protocols. Figure S6: Funnel plot for assessing the publication bias of our meta-analysis of the area under the receiver operating curve for non-hypotension prediction index artificial intelligence models for the prediction of intraoperative hypotension. The plot is asymmetric. Figure S7: Sub-group analysis of the AUROC meta-analysis for non-HPI studies based on the cut-off used to define hypotension. Figure S8: Sub-group analysis of the AUROC meta-analysis for HPI studies 5 minutes prior to intraoperative hypotension based on study quality and risk of bias. Figure S9: Sub-group analysis of the AUROC meta-analysis for HPI studies 10 minutes prior to intraoperative hypotension based on study quality and risk of bias. Figure S10: Sub-group analysis of the AUROC meta-analysis for HPI studies 15 minutes prior to intraoperative hypotension based on study quality and risk of bias. Figure S11: Sub-group analysis of the AUT-MAP < 65 meta-analysis for HPI studies based on study quality and risk of bias. Figure S12: Sub-group analysis of the duration of intraoperative hypotension meta-analysis for HPI studies based on study quality and risk of bias. Figure S13: Sub-group analysis of the TWA-MAP < 65 meta-analysis for HPI studies based on study quality and risk of bias. Figure S14: Sub-group analysis of the AUROC meta-analysis for non-HPI studies based on study quality and risk of bias.

Supplementary Material 3

Supplementary Material 4

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Mohammadi, I., Firouzabadi, S.R., Hosseinpour, M. et al. Predictive ability of hypotension prediction index and machine learning methods in intraoperative hypotension: a systematic review and meta-analysis. J Transl Med 22, 725 (2024). https://doi.org/10.1186/s12967-024-05481-4

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12967-024-05481-4

Keywords

Journal of Translational Medicine

ISSN: 1479-5876

Contact us