Abstract
Objective
To review, inventory and compare available diagnostic tools and investigate which tool has the best performance for prehospital risk assessment in patients suspected of non-ST-segment elevation acute coronary syndrome (NSTE-ACS).
Methods
Systematic review and meta-analysis. Medline and Embase were searched up till 1 April 2021. Prospective studies with patients, suspected of NSTE-ACS, presenting in the primary care setting or by emergency medical services (EMS) were included. The most important exclusion criteria were studies including only patients with ST-elevation myocardial infarction and studies before 1995, the pretroponin era. The primary end point was the final hospital discharge diagnosis of NSTE-ACS or major adverse cardiac events (MACE) within 6 weeks. Risk of bias was evaluated by the Quality Assessment of Diagnostic Accuracy Studies Criteria.
Main outcome and measures
Sensitivity, specificity and likelihood ratio of findings for risk stratification in patients suspected of NSTE-ACS.
Results
In total, 15 prospective studies were included; these studies reflected in total 26 083 patients. No specific variables related to symptoms, physical examination or risk factors were useful in risk stratification for NSTE-ACS diagnosis. The most useful electrocardiographic finding was ST-segment depression (LR+3.85 (95% CI 2.58 to 5.76)). Point-of-care troponin was found to be a strong predictor for NSTE-ACS in primary care (LR+14.16 (95% CI 4.28 to 46.90) and EMS setting (LR+6.16 (95% CI 5.02 to 7.57)). Combined risk scores were the best for risk assessment in an NSTE-ACS. From the combined risk scores that can be used immediately in a prehospital setting, the PreHEART score, a validated combined risk score for prehospital use, derived from the HEART score (History, ECG, Age, Risk factors, Troponin), was most useful for risk stratification in patients with NSTE-ACS (LR+8.19 (95% CI 5.47 to 12.26)) and for identifying patients without ACS (LR−0.05 (95% CI 0.02 to 0.15)).
Discussion
Important study limitations were verification bias and heterogeneity between studies. In the prehospital setting, several diagnostic tools have been reported which could improve risk stratification, triage and early treatment in patients suspected for NSTE-ACS. On-site assessment of troponin and combined risk scores derived from the HEART score are strong predictors. These results support further studies to investigate the impact of these new tools on logistics and clinical outcome.
Funding
This study is funded by ZonMw, the Dutch Organisation for Health Research and Development.
Trial registration number
This meta-analysis was published for registration in PROSPERO prior to starting (CRD York, CRD42021254122).
Keywords: myocardial infarction, primary care, coronary heart disease, cardiology
Strengths and limitations of this study.
This systematic review and meta-analysis include all currently available prospective data on the subject of prehospital risk assessment in patients with non-ST-segment elevation acute coronary syndrome.
Available patient characteristics and diagnostic tools (including recently developed combined risk scores and point-of-care biomarkers) are reviewed and compared in both the setting of primary care as well as emergency medical care.
Despite strict inclusion and exclusion criteria, heterogeneity was present in our review.
For none of the included prehospital diagnostic tools, there were prospective intervention and/or randomised controlled studies available.
Introduction
The suspicion of an acute coronary syndrome (ACS) is one of the main reasons for consultation of emergency medical care.1–4 Depending on national and regional protocols, patients who are suspected of having an ACS are first seen in primary care by a general practitioner (GP) or by the emergency medical services (EMS). The decision to admit the suspected patient with ACS directly to a tertiary intervention centre, or to transport the patient to a local hospital for observation results from varying risk stratification methods. In general, this prehospital risk assessment and triage is mainly based on the patient’s medical history and physical examination. None or only a limited number of diagnostic tools are used in this process. Often, only an ECG is made by the EMS to rule out a ST-segment elevation myocardial infarction (STEMI). Consequently, no adequate prehospital risk assessment and triage is performed in those without ST-segment elevation on their ECG (NSTE-ACS) and are transferred to a local hospital without onsite interventional options for further diagnostic evaluation.2
In the recent years, in-hospital diagnostic algorithm for suspected NSTE-ACS has been developed and improved. The implementation of combined risk scores (CRSs) such as the History, ECG, Age, Risk factors, Troponin (HEART) and the Thrombolysis in Myocardial Infarction (TIMI) scores and new sensitive biomarker assays have substantially improved risk stratification, reduced the delay to diagnosis, shortened stays in the emergency departments and lowered healthcare costs.5 6 Further improvement of care and cost reduction could be possible when these diagnostic algorithms would be applicable in the prehospital phase. The recent development of point-of-care (POC) analysers for on-site biomarker assessment has accelerated these possibilities of prehospital risk assessment in suspected patients with NSTE-ACS. Improving early risk stratification by GP or ambulance paramedics is important for several medical and economic reasons. Most importantly, due to the absence of reliable prehospital diagnostic tools, an NSTE-ACS is not always recognised, leading to a missed diagnosis in around 2%–5% of consultations.7 Additionally, the latest European guidelines recommend that high-risk patients with NSTE-ACS receive an early invasive strategy within 24 hours.8 It is well known that this is associated with shorter ischaemic times and subsequent improved clinical outcomes.9–11
On the other hand, the use of additional diagnostic tools could help to definitively rule out an NSTE-ACS in the prehospital setting in chest pain patients with a low suspicion for NSTE-ACS. This is important as only 11%–21% of these patients presenting to an emergency department are finally diagnosed with an NSTE-ACS, in the remaining majority generally no life-threatening condition is found.12–14 Early risk assessment in these patients with low suspicion could significantly reduce the need of unnecessary diagnostics, referrals to hospitals, and therefore reduce healthcare cost.
Recently, several studies have been performed to investigate several known and new diagnostic tools for prehospital risk assessment in suspected patients with ACS.15–17 We performed a systematic review and meta-analysis of available prospective data to summarise and compare current available and new diagnostic tools for prehospital risk assessment in patients suspected of NSTE-ACS.
Methods
Study selection and data extraction
This meta-analysis was published for registration in PROSPERO prior to starting (CRD York, CRD42021254122).
We performed searches in Medline and Embase for all published articles up till 1 April 2021. After identifying articles, we reviewed references from appropriate articles to identify additional references for this systematic review. Full search is shown in online supplemental eAppendix 1.
bmjopen-2021-057305supp001.pdf (775.7KB, pdf)
Titles and abstract for all articles were screened by the primary author (JD) and decisions were checked by a second reviewer (P-JV). If the author identified the article as potentially suitable for inclusion, the full text of the article was reviewed in detail by two authors (JD and P-JV) and selected whether the inclusion criteria were met. Disagreements between reviewers were resolved by the decision of a third independent reviewer. If data were sufficient to generate a 2×2 table, the data were independently extracted, and the methodological quality and eligibility were determined. Bias was evaluated by the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) Criteria.18
We included studies that met the following inclusion criteria: (1) patients suspected for NSTE-ACS; (2) prospective study; (3) original data; (4) presenting prehospital (primary care, EMS); (5) prehospital risk assessment or triage; (6) outcome data on in-hospital ACS or major adverse cardiac event (MACE) within 6 weeks after presentation; (7) article published in English, French or German.
The objective of this review was to investigate which patient characteristics and diagnostic tools were used for early risk assessment in an undifferentiated prehospital population suspected of NSTE-ACS, we excluded studies enrolling a specific subpopulation from the general ACS population, studies with less than 100 patients, and studies enrolling only patients with STEMI. Further, we excluded studies from which the data were incomplete, or the endpoints were not focused on diagnostic performance. Studies before 1995, the pretroponin era, were excluded as well.
End points and definitions
The primary end point was the final hospital discharge diagnosis of NSTE-ACS by the treating physician or MACE within 6 weeks. NSTE-ACS refers to non-STEMI or unstable angina pectoris (UAP). MACE was defined in the studies and included at least myocardial infarction.
Data and statistical analysis
For included studies individual data of sample size, number of true positive, true negative, false positive and false negative were extracted separately by two authors (JD and P-JV). When possible, we excluded the patients with STEMI from the data extraction. Sensitivity, specificity, likelihood ratios (LRs) and diagnostics OR (dOR) were calculated from the extracted data by two authors (JD and JZ). The statistical analysis was checked by a third author (MvhV). Extracted data included symptoms, physical examination, risk factors, ECG, biomarkers and CRSs.
Since LRs are an appropriate way to measure and express diagnostic accuracy, our main focus was on this diagnostic measurement.19 20 An LR around 1.0 has no proven diagnostic value. A positive LR (LR+) between 1 and 2 as a minimal predictor for making ACS more likely, a LR+ of >2 as a weak predictor for ACS, a LR+ >5 as a moderate to strong predictor and a LR>10 as a very strong predictor. A negative LR (LR−) between 0.5 and 1 as a minimal predictor to make ACS less likely, a LR− of <0.5 as a weak predictor, LR−<0.2 as a moderate to strong predictor and a LR−<0.1 as a very strong predictor.21
Summary data for dichotomous findings are reported as weighted average when a finding was evaluated in two studies. When findings were reported in three or more studies, the analyses were performed using R V.4.0.4 (R Foundation for Statistical Computing, Vienna, Austria). Sensitivity, specificity and the dOR analyses were performed using the ‘meta’ package as was heterogeneity. LRs and bivariate analyses were performed using the ‘mada’ package.22
Statistical heterogeneity was reported as the I2 for the sensitivity and specificity when findings were evaluated in three studies or more. We considered values between 0% and 30% as low heterogeneity, between 30% and 60% as moderate heterogeneity and 60% or more as considerable heterogeneity.23
Patient and public involvement
Patient and public were not involved.
Statement of ethics approval
It was not necessary to engage an ethics committee for this study-level systematic review and meta-analysis.
Data sharing statement
Not applicable.
Results
Of the 1817 unique articles, a total of 15 prospective studies met our inclusion criteria and were included in this systematic review (online supplemental eFigure 1).15–17 24–35 These studies reflected a total of 26 083 patients. Five studies were performed in setting of primary care by the GP and 10 in the setting of the EMS. Rates of final NSTE-ACS diagnosis ranged in primary care from 5% to 22% (median 11% (IQR, 6%–20%). In the EMS setting, the rates of final ACS diagnosis ranged from 8% to 57% (median 17% (IQR, 12%–20%).
Risk of bias for each included study was scored with the QUADAS tool (online supplemental eAppendix 2). Data on 92 variables related to symptoms, physical examination, risk factors, ECG, biomarkers and risk scores were extracted from the articles. The total number of studies providing data, the total number of patients, sensitivity, specificity, LR+, LR− and dOR are listed in the online supplemental material. All data relevant to the systematic review and meta-analysis are uploaded as online supplemental material (online supplemental eAppendies 3–10).
Symptoms
Symptoms were not predictive for ACS in primary care or in the EMS setting (online supplemental eTable 1). The absence of having chest pain in the ambulance was a minimal predictor for making ACS less likely (LR− 0.61 (0.58–0.63)). Duration of symptoms in primary care until presentation was not appropriate for risk assessment in ACS.
Physical examination
Pain not reproducible by palpation (LR+1.10 (1.02–1.20)) and pallor (LR+1.17 (1.00–1.38)) were minimally predictive for making ACS more likely (online supplemental eTable 2). Reproducing the pain by palpation is a weak predictor for making ACS less likely (LR− 0.47 (0.43–0.51)) in EMS setting.
Cardiovascular risk factors
Of all the cardiovascular risk factors in EMS setting, history of prior coronary disease (LR+2.07 (1.58–2.71)) was the best predictor to make ACS more likely (online supplemental eTable 3). Male gender (LR+1.28 (1.12–1.45), history of angina pectoris (LR+1.27 (1.08–1.49)) or previous acute myocardial infarction (AMI) (LR+1.83 (1.21–2.80)), CABG (LR+1.60 (1.14–2.24) or percutaneous coronary intervention (PCI) (LR+1.49 (1.24–1.78)) were minimal predictors for ACS. For identifying patients less likely to have an ACS, no risk factor when absent conferred an LR− of 0.5 or lower. In primary care, male gender (LR+1.61 (1.34–1.93)) was the only risk factor for making ACS more likely (online supplemental eTable 4).
ECG
A total of three prospective studies investigated the value of ECG, one in primary care and two in EMS.24 25 34 In all three studies, the ECG was categorised in ischaemic ECG versus abnormal and/or non-ischaemic ECG. In only one study, the exact definition of an ischaemic ECG was mentioned: ST-deviation≥800 μV.24 Interpreting an ECG as an ischaemic ECG is predictive for ACS in both the EMS setting (LR+2.23 (1.11–5.30)) as in primary care (LR+3.4 (2.15–5.37)). Classifying the ECG as non-ischaemic is a weak predictor for making ACS less likely in primary care (LR− 0.33 (0.17–0.67)) and a minimal predictor in the EMS setting (LR− 0.57 (0.33–0.95)) (table 1). In terms of individual characteristics of the ECG, a Q-wave is a very strong predictor for ACS (LR+24.62 (3.38–179.20)). ST-depression (LR+3.85 (2.58–5.76)) is a better predictor for ACS than T-wave inversion (LR+2.56 (1.50–4.39)) (table 1).
Table 1.
Performance of the ECG for risk assessment in non-ST-elevation acute coronary syndrome
| Test | Setting | Studies | Patients | Sensitivity, % (95 CI) | I2, %* | Specificity, % (95 CI) | I2, %* | LR+ (95 CI) | LR− (95 CI) | dOR (95 CI) |
| Ischaemic ECG†24 25 | EMS | 2 | 1959 | 58 (45 to 71) |
– | 74 (59 to 85) |
– |
2.23(1.11 to 5.30) |
0.57 (0.33 to 0.95) |
4.03 (0.70 to 23.42) |
| Ischaemic ECG†34 | GP | 1 | 243 | 74 (53 to 88) |
– | 78 (69 to 84) |
– | 3.4 (2.15 to 5.37) |
0.33 (0.17 to 0.67) |
10.20 (3.55 to 29.28) |
| Q-wave25 | EMS | 1 | 541 | 11 (8 to 15) |
– | 100 (98 to 100) |
– | 25.36 (3.49 to 184.47) |
0.89 (0.12 to 6.50) |
28.39 (3.90 to 209.01) |
| ST-depression25 | EMS | 1 | 541 | 42 (36 to 48) |
– | 89 (84 to 93) |
– | 3.85 (2.58 to 5.76) |
0.65 (0.43 to 0.97) |
5.91 (3.68 to 9.49) |
| T-wave inversion25 | EMS | 1 | 541 | 18 (14 to 23) |
– | 93 (89 to 96) |
– | 2.56 (1.50 to 4.39) |
0.88 (0.52 to 1.51) |
2.91 (1.62 to 5.22) |
*When the summary measure was from less than three studies, I2 was not calculated.
†Definition of ischaemic ECG given in one study only; ST-deviation ≥800 μV.
dOR, diagnostic odds ratio; EMS, emergency medical services; GP, general practitioner; LR−, negative likelihood ratio; LR+, positive likelihood ratio.
Biomarkers
A total of nine prospective studies investigated the value of biomarkers, four in primary care and five in EMS.17 24–29 33 34 All blood samples for biomarkers were drawn in the prehospital setting. Most samples were analysed directly prehospital by POC analysers, but in one study the blood sample was analysed after arrival in the hospital.27 In EMS setting, median time of onset of complaint to biomarker assessment was 80 min (IQR 59–135 min) and in primary care the median was 600 (IQR 180–600 min). Further details on the prehospital analysed biomarkers are shown in online supplemental eTable 5.
Of the analysed biomarkers in primary care, the heart-type fatty acid binding protein (H-FABP) (LR+7.84 (3.39–20.76)) was moderate to strong for predicting ACS (table 2). POC-troponin (LR+14.16 (4.28–46.90)) was a very strong predictor for ACS. Both were not useful to make an ACS less likely (LR− 0.68 (0.31–1.91) versus LR− 0.71 (0.49–1.02)) (table 3).
Table 2.
Performance of the biomarkers for risk assessment in non-ST-elevation acute coronary syndrome
| Assay | Setting | Studies | Patients | Sensitivity, % (95 CI) | I2, %* | Specificity, % (95 CI) | I2, %* | LR+ (95 CI) | LR− (95 CI) | dOR (95 CI) |
| H-FABP26 28 | GP | 2 | 601 | 35 (12 to 52) |
– | 96 (90 to 99) |
– | 7.84 (3.39 to 20.76) |
0.68 (0.31 to 1.90) |
11.53 (3.76 to 31.91) |
| Hs-troponin T27 | GP | 1 | 115 | 83 (36 to 100) |
– | 76 (67 to 84) |
– | 3.49 (2.40 to 5.09) |
0.22 0.15 to 0.32) |
15.97 (1.79 to 142.88) |
| Myoglobin25 | EMS | 1 | 536 | 8 (5 to 12) |
– | 96 (93 to 98) |
– | 2.07 (0.98 to 4.38) |
0.96 (0.45 to 2.02) |
2.17 (0.99 to 4.74) |
*When the summary measure was from less than three studies, I2 was not calculated.
dOR, diagnostic odds ratio; EMS, emergency medical services; GP, general practitioner; H-FABP, heart-type fatty acid binding protein; Hs, high-sensitive; LR−, negative likelihood ratio; LR+, positive likelihood ratio.
Table 3.
Performance of the point-of-care (POC)-troponin for risk assessment in non-ST-elevation acute coronary syndrome
| Setting | Studies | Patients | Sensitivity, % (95 CI) | I2, %* | Specificity, % (95 CI) | I2, %* | LR+ (95 CI) | LR− (95 CI) | dOR (95 CI) | |
| POC-troponin15 17 25 30 33 | EMS | 5 | 21 484 | 25 (14 to 40) |
96% | 96 (94 to 98) |
85% | 6.16 (5.02 to 7.57) |
0.75 (0.62 to 0.92) |
8.27 (7.42 to 9.21) |
| POC-troponin27 29 | GP | 2 | 311 | 31 (12 to 59) |
– | 98 (95 to 99) |
– | 14.16 (4.28 to 46.90) |
0.71 (0.49 to 1.02) |
20.01 (4.56 to 87.88) |
*When the summary measure was from less than three studies, I2 was not calculated.
dOR, diagnostic OR; EMS, emergency medical services; GP, general practitioner; LR+, positive likelihood ratio; LR−, negative likelihood ratio.
In the EMS setting, POC-troponin (LR+6.16 (5.02–7.57)) was a moderate to strong predictor for making ACS more likely that could be interpreted immediately. There was no difference between blood sampled by POC in the EMS and blood analysed in-hospital.
Combined risk scores
CRSs were investigated in eight studies, two in primary care28 34 and six in EMS.15 16 31–33 35 In the EMS setting, the used CRSs were all derived from the well-known HEART score.5 In total five different CRSs were evaluated in this systematic review (table 4). There were minimal differences between the risk scores. These were due to different positivity threshold, risk factors and troponin analysers. In two studies, data and blood were drawn prehospital but analysed after arrival in-hospital.
Table 4.
Performance of the combined risk scores in emergency medical services setting for risk assessment in non-ST-elevation acute coronary syndrome
| Risk level | Studies | Threshold | LR (95 CI) | Troponin analyser |
| High | ||||
| Core-lab PMHP15 | 1 | Positive troponin | 8.68 (5.04 to 14.93) | In-hospital laboratorium |
| preHEART35 | 1 | 7–10 | 8.19 (5.47 to 12.26) | POC |
| PMHP15 | 1 | Positive troponin | 6.70 (3.01 to 14.91) | POC |
| Modified HEART31 | 1 | 7–10 | 4.96 (3.62 to 6.79) | In-hospital laboratorium |
| HEART35 | 1 | 7–10 | 4.84 (3.41 to 6.89) | POC |
| Intermediate | ||||
| Core-lab PMHP15 | 1 | ≥4 | 0.61 (0.35 to 1.04) | In-hospital laboratorium |
| preHEART35 | 1 | 4–6 | 1.22 (0.95 to 1.57) | POC |
| PMHP15 | 1 | ≥4 | 1.12 (0.74 to 1.70) | POC |
| Modified HEART31 | 1 | 4–6 | 0.89 (0.66 to 1.18) | Lab |
| HEART35 | 1 | 4–6 | 0.69 (0.49 to 0.98) | POC |
| Low | ||||
| Core-lab PMHP15 16 | 2 | 0–3 | 0.08 (0 to 0.17) | In-hospital laboratorium |
| preHEART35 | 1 | 0–3 | 0.05 (0.02 to 0.15) | POC |
| PMHP15 | 1 | 0–3 | 0.26 (0.11 to 0.59) | POC |
| Modified HEART31 | 1 | 0–3 | 0.00 | In-hospital laboratorium |
| HEART32 35 | 2 | 0–3 | 0.14 (0.13 to 0.16) | POC |
HEART, History, ECG, Age, Risk factors, Troponin; LR, likelihood ratio; PMHP, prehospital modified HEART pathway; POC, point-of-care.
In the EMS setting, all the CRSs were strong predictors for ACS when classified in the high-risk group. Between the risk scores there was no statistically significant difference (table 4).
In the low-risk group, the modified HEART performed the best in excluding ACS (LR−0.00). Of the CRSs that can be interpreted directly prehospital, preHEART is a very strong predictor in making ACS less likely (LR− (0.05 (0.02–0.15)).
In all primary care studies, the CRSs classified patients in two risk groups (low vs intermediate/high).28 34 In the EMS setting, this was only the case with the HEAR score and was a weak predictor for making ACS less likely (LR− 0.40 (0.38–0.43)).33 The Marburg Heart Score (MHS) showed a significant better performance in making ACS less likely when the threshold value was lowered from 2 to 1 (LR− 0.57 (0.47–0.68) versus LR− 0.35 (0.32–0.38)) (table 5). Alternative decision tools used in primary care were the CRS. Scoring a positive answer to the elements resulted in a point that was calculated into a final score. This final score classified the patient into a low-risk group or intermediate-high risk group. CRS 1 had the best performance in making ACS less likely (LR− 0.24 (0.21–0.28)) and outperformed the MHS.
Table 5.
Performance of the combined risk scores (CRS) in primary care for risk assessment in non-ST-elevation acute coronary syndrome
| CRS | Studies | Threshold for intermediate-high risk group | LR+ (95 CI) | Threshold for low-risk group | LR− (95 CI) | Setting | Elements of CRS |
| CRS 128 | 1 | 2–6 | 1.82 (1.59 to 2.10) | 0–1 | 0.24 (0.21 to 0.28) | GP |
|
| CRS 228 | 1 | 2–5 | 1.69 (1.47 to 1.96) | 0–1 | 0.36 (0.31 to 0.42) | GP |
|
| CRS 328 | 1 | 2–4 | 1.75 (1.49 to 2.06) | 0–1 | 0.40 (0.34 to 0.46) | GP |
|
| CRS 428 | 1 | 2–3 | 1.55 (1.30 to 1.85) | 0–1 | 0.56 (0.47 to 0.67) | GP |
|
| MHS34 | 1 | ≥3 | 1.34 (1.11 to 1.61) | 0–2 | 0.57 (0.47 to 0.68) | GP | |
| MHS34 | 1 | ≥2 | 1.12 (1.04 to 1.22) | 0–1 | 0.35 (0.32 to 0.38) | GP | |
| GP probability assessment34 | 1 | ≥6 | 1.48 (1.29 to 1.70) | 0–5 | 0.32 (0.28 to 0.37) | GP | |
| Combined MHS and GP probability assessment34 | 1 | 1–2 | 1.30 (1.19 to 1.42) | 0 | 0 | GP |
GP, general practitioner; H-FABP, heart-type fatty acid binding protein; LR+, positive likelihood ratio; LR−, negative likelihood ratio; MHS, Marburg Heart Score; POC, point-of-care.
Compared with GP decision-making without use of a CRS (LR− 0.32 (0.00–0.46)), there is no improvement that makes ACS less likely (table 6).
Table 6.
Performance of general practitioner (GP) decision-making for risk assessment in non-ST-elevation acute coronary syndrome
| Test | Studies | Patients | Sensitivity, % (95 CI) | I2, %* | Specificity, % (95 CI) | I2, %* | LR+ (95 CI) | LR− (95 CI) | dOR (95 CI) |
| GP decision to refer28 29 | 2 | 371 | 79 (54 to 100) |
– | 67 (50 to 75) |
– | 2.37 (1.95 to 3.73) |
0.32 (0.00 to 0.46) |
7.50 (0 to 14.45) |
| GP decision to refer with POC-troponin29 | 1 | 128 | 71 (29 to 96) |
– | 78 (69 to 85) |
– | 3.20 (2.14 to 4.80) |
0.37 (0.25 to 0.55) |
8.70 (1.60 to 47.40) |
| GP immediately suspected a serious condition34 | 1 | 243 | 58 (42 to 72) |
– | 53 (45 to 60) |
– | 1.22 (0.98 to 1.51) |
0.80 0.65 to 1.00) |
1.51 0.79 to 2.91) |
*When the summary measure was from less than three studies, I2 was not calculated.
dOR, diagnostic odds ratio; LR+, positive likelihood ratio; LR−, negative likelihood ratio; POC, point-of-care.
Making an ACS more likely, decision-making by the GP to refer to the hospital, without using a POC-troponin (LR+2.37 (1.95–3.73)) and using POC-troponin (LR+3.20 (2.14–4.80)) outperformed the CRSs.
Alternative diagnosis
In our systematic review, three studies in the EMS setting described which alternative diagnoses were possible. From these, two studies noted the predictive value of POC-troponin for other diseases,17 30 one study noted which other causes for chest pain could be classified as low risk.31 POC-troponin is a moderate to strong predictor for other cardiac-related diseases like myocarditis (LR+8.89 (8.29–9.54)), decompensated heart failure (LR+5.26 (1.88–6.03)), valve disorders/endocarditis (LR+3.29 (2.93–3.69)) and cardiomyopathy (LR+4.80 (4.36–5.29)). For non-cardiac diseases, POC-troponin is a weak to moderate predictor to make pulmonary embolism (LR+3.81 (1.38–4.46)), pneumonia (LR+2.41 (2.13–2.74)) and diseases of the genitourinary system more likely (LR+4.07 (2.66–4.51)). A normal POC-troponin value makes a myocarditis less likely (LR− 0.40 (0.37–0.42)) but is not a strong predictor for other diagnoses. In the modified HEART, 403 patients were classified as low risk for ACS. In this group, two serious non-cardiac diseased (pulmonary embolism, acute pancreatitis) were found.
Discussion
In this systematic review, we evaluated 15 prospective studies including 26.083 patients to identify patient characteristics and diagnostic tools for prehospital risk assessment in suspected patients with NSTE-ACS. The main finding is that in the prehospital setting several new diagnostic tools are available which could improve early triage and risk assessment in suspected patients with NSTE-ACS, in particular on-site assessment of biomarkers and CRSs derived from the HEART score were very strong for risk assessment in patients with NSTE-ACS.
Symptoms, physical examination and risk factors
In accordance with previous analyses, in the present study no specific variables related to symptoms and physical examination were relevant predictors for an ACS diagnosis. The predictive value of some cardiovascular risk factors in the prehospital setting was as a single variable weak at best. A possible explanation is that in the prehospital setting less objective information is present on previous history of a patient. This was indeed corroborated in a recent analysis comparing prehospital and in-hospital calculated HEART scores.16 In this study, disagreement between hospital and prehospital HEART score risk classifications was found in approximately 25% of patients, mainly by different scoring of history and cardiovascular risk factors.
ECG
In patients with suspected ACS, the resting 12-lead ECG is the first-line diagnostic tool. Our results indicate that in the prehospital setting only a minority of the studies provided detailed description of the ECG criteria and used definitions. Moreover, the ECG was only a moderate predictor for NSTE-ACS. We expect that a more detailed analysis of ECGs, training of EMS personnel or GPs could improve the performance of ECG in the prehospital setting.
Biomarkers
The advent of POC analysers facilitates the use of established biomarkers in the prehospital setting. Current guidelines state that measurement of the biomarker troponin is mandatory in all suspected ACS patients.36 Although the included studies used normal troponin I and T assays, the first high-sensitivity assays also became available recently for POC analysers. This can increase diagnostic accuracy and reduce the time to diagnosis.37 Depending on the type of assay used, assessment of this biomarker is already reliable 3–6 hours after onset of symptoms. The duration of symptoms could differ between different settings and populations. We found that duration of symptoms often is much longer in the primary care setting compared with the EMS setting. This explains why the negative predictive value of POC troponin is much higher in the primary care setting than in EMS setting. Consequently, symptom duration could indeed be too short to rule out an ACS based on a single troponin assessment in the EMS setting. However, the use of POC troponin in a CRS could offer a solution to lower the risk of false negatives. This is supported by the results from our analyses which clearly show that the CRSs are superior to a single troponin for risk assessment in a patient suspected for NSTE-ACS.
Two primary care studies26 28 examined the diagnostic value of H-FABP, a small protein that is common in cardiac muscle and is released within 2 hours after the onset of ischaemia.38 Although there are no prehospital studies that compare troponin with H-FABP, several in-hospital studies have compared these biomarkers. Despite of the early release, these studies showed that the predictive value of the H-FABP is moderate when assessed within the time frame of 6 hours after onset of symptoms. Compared with POC-troponin, stand-alone H-FABP showed inferior diagnostic performance.39–41
Combined risk scores
International guidelines advise the use of CRSs at emergency departments.36 Several scores are available that, generally combine elements of history, cardiovascular risk factors, ECG and the biomarker troponin. The advent of POC analysers makes it possible to use these risk scores in the prehospital setting. We found mainly risk scores derived from the well-known HEART score, which were validated for prehospital usage. The results tend to be very promising; for identifying high-risk patients for NSTE-ACS, but especially for excluding ACS for low-risk patients. Compared with a previous systematic review by Fanaroff et al42 in 2015, in-hospital use of the original HEART score shows higher predictive value for high-risk patients in diagnosing ACS, and a similar strong negative predictive value is found for patients classified as low risk. The better predictive value for the high-risk group in the original, in-hospital HEART score can be explained by the presence of more objective information and the longer time between the onset of symptoms in relation to blood analyses. In this patient population, there exists a considerable probability that the in-hospital troponin value will be positive more often compared with the prehospital setting as explained before.
In the present analysis, the performance of CRSs in the primary care setting was lower as compared with the EMS setting. This is due the fact that different scores were used, and in none a POC troponin. In addition, important differences exist in patient selection and prevalence of ACS, which also could have impacted the results. Finally, the performance of diagnostic tools (in particular ECG) and CRSs depends on the experience and support of each individual physician. In primary care, there are large difference between regions and countries to what extent additional diagnostic tools are used or necessary. In particular in remote areas with no nearby hospitals, the implementation of CRSs could improve acute medical care for suspect patients with ACS.
When evaluating a patient with suspected ACS, life-threatening diagnoses such as aortic dissection and pulmonary embolism should be part of the clinical evaluation and differential diagnosis. These alternative diagnoses are often associated with increased levels of biomarkers or ECG changes as well. So, the question is if the prehospital tested biomarkers and CRS are sufficient to also rule out chest pain due to aortic dissection or pulmonary embolism. Our meta-analysis included three studies that reported life-threatening diagnoses in the different risk groups. In the group of patients classified as low risk for ACS, one pulmonary embolism and one acute pancreatitis occurred. No information regarding aortic dissections was available. These data indicate that although a CRS is a very strong predictor for excluding ACS, the possibility for a severe alternative diagnosis should not be ignored.
Further perspective and upcoming studies
Before using these new diagnostic tools in routine practice, more data on safety, efficacy and performance are necessary. More data are needed when these CRSs are used to determine a different triage and/or treatment strategy. Some up-coming studies that will examine whether this early risk stratification and associated referral decisions based on CRS are feasible, are the preHEART3 study (Trialregister.nl, NL7866), PRe-hospital Evaluation of Senstive TrOponin (PRESTO) study,43 ARTICA trial,44 FamouS Triage 3 study45, TRIAGE-ACS study (Clinicaltrials.gov, NCT05243485), and POPular HEART (Clinicaltrials.gov, NCT04851418).
Study limitations
Although we included only prospective studies to improve the overall study quality, this analysis suffers from several limitations inherent to this study design and patient population.
An important limitation of the study was verification bias. There was a variable definition for ACS between the studies, for example, not all studies included UAP and in some studies some patients were included who developed an STEMI. Although this reflects daily practice, it could have impacted the results.
In the analysis of some variables, heterogeneity was considerable in our meta-analysis (eg, cardiovascular risk factors and POC-troponin measurement in the EMS setting). In POC-troponin, heterogeneity is the result of variation in thresholds between studies. Although this is expected in systematic reviews of diagnostic test accuracy, it is possible this influences our main findings.46 However, given the small contribution we do not think that the heterogeneity in our meta-analysis alternates the influence of stand-alone risk factors in early risk stratification for ACS.
Another limitation is the lack of well-described patient and additional diagnostic tools data. For example, only one of three studies described their definition of an ischaemic ECG. Therefore, it is difficult to draw conclusions about the diagnostic value of the ECG in this particular population in the prehospital setting. On the other hand, these limited available data are not uncommon for these kind of studies. It is the result of the acute setting in which these patients presents themselves to the GP or the EMS.
Additionally, many of the diagnostic tools were examined in a limited number of studies and study populations. Given the large differences in patient presentation and prevalence of ACS, the performance of the different diagnostic tools could differ when used in a clinical setting or population with different prevalence of ACS.
Finally, not for all diagnostic tools information on false negative results and alternative life-threatening diagnoses were given. Before changing triage and consequent treatment based on these tools, more prospective data are necessary to assess the safety in these rare but important selection of patients.
Conclusion
In the prehospital setting, several new diagnostic tools are available which could improve risk stratification, triage and early treatment in suspected patients with ACS. On-site assessment of biomarkers and CRSs derived from the HEART score showed to be strong predictors for diagnosing patients with a NSTE-ACS and to identify patients without a NSTE-ACS in the EMS setting (figure 1). These promising results underline the importance of new prospective studies in this field.
Figure 1.
Graphical abstract. It shows the indicative positive and negative likelihood ratio (LR) for each element and how it compares to its predictive value. POC, point-of-care.
Supplementary Material
Footnotes
Twitter: @DemandtJesse
Contributors: JD: planning, conducting, data collection, reporting, and check, statistical analysis, article writing, submission. JZ: data collection and reporting, statistical analysis, article writing. AK, GS, PvdH, PT, LD: article writing. MvhV: planning, conducting, date collection, reporting, and check, statistical check, article writing. P-JV: planning, conducting, data collection and reporting, article writing, submission, guarantor. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Funding: This study is funded by ZonMw, the Dutch Organisation for Health Research and Development, grant programme “Topspecialistische Zorg en Onderzoek”, grant number 10070012010001.
Competing interests: None declared.
Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review: Not commissioned; externally peer reviewed.
Supplemental material: This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Not applicable.
Ethics approval
Not applicable.
References
- 1.Hoorweg BB, Willemsen RT, Cleef LE, et al. Frequency of chest pain in primary care, diagnostic tests performed and final diagnoses. Heart 2017;103:1727–32. 10.1136/heartjnl-2016-310905 [DOI] [PubMed] [Google Scholar]
- 2.Pedersen CK, Stengaard C, Friesgaard K, et al. Chest pain in the ambulance; prevalence, causes and outcome - a retrospective cohort study. Scand J Trauma Resusc Emerg Med 2019;27:84. 10.1186/s13049-019-0659-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Thang ND, Karlson BW, Bergman B, et al. Characteristics of and outcome for patients with chest pain in relation to transport by the emergency medical services in a 20-year perspective. Am J Emerg Med 2012;30:1788–95. 10.1016/j.ajem.2012.02.014 [DOI] [PubMed] [Google Scholar]
- 4.Wibring K, Lingman M, Herlitz J, et al. Prehospital stratification in acute chest pain patient into high risk and low risk by emergency medical service: a prospective cohort study. BMJ Open 2021;11:e044938. 10.1136/bmjopen-2020-044938 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Backus BE, Six AJ, Kelder JC, et al. A prospective validation of the HEART score for chest pain patients at the emergency department. Int J Cardiol 2013;168:2153–8. 10.1016/j.ijcard.2013.01.255 [DOI] [PubMed] [Google Scholar]
- 6.Antman EM, Cohen M, Bernink PJ, et al. The TIMI risk score for unstable angina/non-ST elevation MI: a method for prognostication and therapeutic decision making. JAMA 2000;284:835–42. 10.1001/jama.284.7.835 [DOI] [PubMed] [Google Scholar]
- 7.Harskamp R, van Peet P, Bont J, et al. The conundrum of acute chest pain in general practice: a nationwide survey in The Netherlands. BJGP Open 2018;2:bjgpopen18X101619. 10.3399/bjgpopen18X101619 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Collet JP, Thiele H, Barbato E. ESC Guidelines for the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation. Rev Esp Cardiol 20202021;74:544. [DOI] [PubMed] [Google Scholar]
- 9.Bainey KR, Gafni A, Rao-Melacini P, et al. The cost implications of an early versus delayed invasive strategy in acute coronary syndromes: the TIMACS study. J Med Econ 2014;17:415–22. 10.3111/13696998.2014.911184 [DOI] [PubMed] [Google Scholar]
- 10.Nieuwets A, Poldervaart JM, Reitsma JB, et al. Medical consumption compared for TIMI and HEART score in chest pain patients at the emergency department: a retrospective cost analysis. BMJ Open 2016;6:e010694. 10.1136/bmjopen-2015-010694 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.de Koning E, Biersteker TE, Beeres S, et al. Prehospital triage of patients with acute cardiac complaints: study protocol of HART-c, a multicentre prospective study. BMJ Open 2021;11:e041553. 10.1136/bmjopen-2020-041553 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Goodacre S, Cross E, Arnold J, et al. The health care burden of acute chest pain. Heart 2005;91:229–30. 10.1136/hrt.2003.027599 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kienbacher CL, Fuhrmann V, van Tulder R, et al. Impact of more conservative European Society of Cardiology guidelines on the management of patients with acute chest pain. Int J Clin Pract 2021;75:e14133. 10.1111/ijcp.14133 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Boubaker H, Grissa MH, Beltaief K, et al. A new score for the diagnosis of acute coronary syndrome in acute chest pain with non-diagnostic ECG and normal troponin. Emerg Med J 2015;32:764–8. 10.1136/emermed-2013-203151 [DOI] [PubMed] [Google Scholar]
- 15.Stopyra JP, Snavely AC, Smith LM, et al. Prehospital use of a modified HEART pathway and point-of-care troponin to predict cardiovascular events. PLoS One 2020;15:e0239460. 10.1371/journal.pone.0239460 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.van Dongen DN, Fokkert MJ, Tolsma RT, et al. Accuracy of pre-hospital HEART score risk classification using point of care versus high sensitive troponin in suspected NSTE-ACS. Am J Emerg Med 2020;38:1616–20. 10.1016/j.ajem.2019.158448 [DOI] [PubMed] [Google Scholar]
- 17.Rasmussen MB, Stengaard C, Sørensen JT, et al. Predictive value of routine point-of-care cardiac troponin T measurement for prehospital diagnosis and risk-stratification in patients with suspected acute myocardial infarction. Eur Heart J Acute Cardiovasc Care 2019;8:299–308. 10.1177/2048872617745893 [DOI] [PubMed] [Google Scholar]
- 18.Whiting PF, Weswood ME, Rutjes AWS, et al. Evaluation of QUADAS, a tool for the quality assessment of diagnostic accuracy studies. BMC Med Res Methodol 2006;6:9. 10.1186/1471-2288-6-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.McGee S. Simplifying likelihood ratios. J Gen Intern Med 2002;17:647–50. 10.1046/j.1525-1497.2002.10750.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Honest H, Khan KS. Reporting of measures of accuracy in systematic reviews of diagnostic literature. BMC Health Serv Res 2002;2:4. 10.1186/1472-6963-2-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Grimes DA, Schulz KF. Refining clinical diagnosis with likelihood ratios. Lancet 2005;365:1500–5. 10.1016/S0140-6736(05)66422-7 [DOI] [PubMed] [Google Scholar]
- 22.Shim SR, Kim S-J, Lee J. Diagnostic test accuracy: application and practice using R software. Epidemiol Health 2019;41:e2019007. 10.4178/epih.e2019007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21:1539–58. 10.1002/sim.1186 [DOI] [PubMed] [Google Scholar]
- 24.Anroedh SS, Kardys I, Akkerhuis KM, et al. e-Transmission of ECGs for expert consultation results in improved triage and treatment of patients with acute ischaemic chest pain by ambulance paramedics. Neth Heart J 2018;26:562–71. 10.1007/s12471-018-1187-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Svensson L, Isaksson L, Axelsson C, et al. Predictors of myocardial damage prior to hospital admission among patients with acute chest pain or other symptoms raising a suspicion of acute coronary syndrome. Coron Artery Dis 2003;14:225–31. 10.1097/01.mca.0000063503.13456.0d [DOI] [PubMed] [Google Scholar]
- 26.Bruins Slot MHE, Rutten FH, van der Heijden GJMG, et al. Diagnostic value of a heart-type fatty acid-binding protein (H-FABP) bedside test in suspected acute coronary syndrome in primary care. Int J Cardiol 2013;168:1485–9. 10.1016/j.ijcard.2012.12.050 [DOI] [PubMed] [Google Scholar]
- 27.Andersson PO, Karlsson J-E, Landberg E, et al. Consequences of high-sensitivity troponin T testing applied in a primary care population with chest pain compared with a commercially available point-of-care troponin T analysis: an observational prospective study. BMC Res Notes 2015;8:210. 10.1186/s13104-015-1174-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Willemsen RT, Winkens B, Kietselaer BL, et al. Evaluating possible acute coronary syndrome in primary care: the value of signs, symptoms, and plasma heart-type fatty acid-binding protein (H-FABP). A diagnostic study. BJGP Open 2019;3:bjgpopen19X101652. 10.3399/bjgpopen19X101652 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Nilsson S, Andersson PO, Borgquist L, et al. Point-of-care troponin T testing in the management of patients with chest pain in the Swedish primary care. Int J Family Med 2013;2013:532093 10.1155/2013/532093 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Stengaard C, Sørensen JT, Ladefoged SA, et al. Quantitative point-of-care troponin T measurement for diagnosis and prognosis in patients with a suspected acute myocardial infarction. Am J Cardiol 2013;112:1361–6. 10.1016/j.amjcard.2013.06.026 [DOI] [PubMed] [Google Scholar]
- 31.Ishak M, Ali D, Fokkert MJ, et al. Fast assessment and management of chest pain patients without ST-elevation in the pre-hospital gateway (famous triage): ruling out a myocardial infarction at home with the modified HEART score. Eur Heart J Acute Cardiovasc Care 2018;7:102–10. 10.1177/2048872616687116 [DOI] [PubMed] [Google Scholar]
- 32.van Dongen DN, Tolsma RT, Fokkert MJ, et al. Pre-hospital risk assessment in suspected non-ST-elevation acute coronary syndrome: a prospective observational study. Eur Heart J Acute Cardiovasc Care 2020;9:5–12. 10.1177/2048872618813846 [DOI] [PubMed] [Google Scholar]
- 33.van Dongen DN, Fokkert MJ, Tolsma RT, et al. Value of prehospital troponin assessment in suspected non-ST-elevation acute coronary syndrome. Am J Cardiol 2018;122:1610–6. 10.1016/j.amjcard.2018.07.037 [DOI] [PubMed] [Google Scholar]
- 34.Schols AMR, Willemsen RTA, Bonten TN, et al. A nationwide flash-mob study for suspected acute coronary syndrome. Ann Fam Med 2019;17:296–303. 10.1370/afm.2401 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Sagel DC, Vlaar PJ, Waardenburg I. Pre-hospital risk stratification in patients with undifferentiated chest pain 2021.
- 36.Collet J-P, Thiele H, Barbato E, et al. 2020 ESC guidelines for the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation. Eur Heart J 2021;42:1289–367. 10.1093/eurheartj/ehaa575 [DOI] [PubMed] [Google Scholar]
- 37.Boeddinghaus J, Nestelberger T, Koechlin L, et al. Early diagnosis of myocardial infarction with point-of-care high-sensitivity cardiac troponin I. J Am Coll Cardiol 2020;75:1111–24. 10.1016/j.jacc.2019.12.065 [DOI] [PubMed] [Google Scholar]
- 38.Dekker MS, Mosterd A, van 't Hof AWJ, et al. Novel biochemical markers in suspected acute coronary syndrome: systematic review and critical appraisal. Heart 2010;96:1001–10. 10.1136/hrt.2009.189886 [DOI] [PubMed] [Google Scholar]
- 39.Lippi G, Mattiuzzi C, Cervellin G. Critical review and meta-analysis on the combination of heart-type fatty acid binding protein (H-FABP) and troponin for early diagnosis of acute myocardial infarction. Clin Biochem 2013;46:26–30. 10.1016/j.clinbiochem.2012.10.016 [DOI] [PubMed] [Google Scholar]
- 40.Bivona G, Agnello L, Bellia C, et al. Diagnostic and prognostic value of H-FABP in acute coronary syndrome: still evidence to bring. Clin Biochem 2018;58:1–4. 10.1016/j.clinbiochem.2018.04.021 [DOI] [PubMed] [Google Scholar]
- 41.Bruins Slot MHE, van der Heijden GJMG, Stelpstra SD, et al. Point-of-care tests in suspected acute myocardial infarction: a systematic review. Int J Cardiol 2013;168:5355–62. 10.1016/j.ijcard.2013.08.002 [DOI] [PubMed] [Google Scholar]
- 42.Fanaroff AC, Rymer JA, Goldstein SA, et al. Does this patient with chest pain have acute coronary syndrome?: the rational clinical examination systematic review. JAMA 2015;314:1955–65. 10.1001/jama.2015.12735 [DOI] [PubMed] [Google Scholar]
- 43.Alghamdi A, Cook E, Carlton E, et al. PRe-hospital evaluation of sensitive troponin (PRESTO) study: multicentre prospective diagnostic accuracy study protocol. BMJ Open 2019;9:e032834. 10.1136/bmjopen-2019-032834 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Aarts GWA, Camaro C, van Geuns R-J, et al. Acute rule-out of non-ST-segment elevation acute coronary syndrome in the (pre)hospital setting by HEART score assessment and a single point-of-care troponin: rationale and design of the ARTICA randomised trial. BMJ Open 2020;10:e034403. 10.1136/bmjopen-2019-034403 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.van Dongen DN, Tolsma RT, Fokkert MJ, et al. Referral decisions based on a prehospital HEART score in suspected non-ST-elevation acute coronary syndrome: design of the famous triage 3 study. Future Cardiol 2020;16:217–26. 10.2217/fca-2019-0030 [DOI] [PubMed] [Google Scholar]
- 46.Macaskill P, Gatsonis C, Deeks J, et al. Chapter 10: analysing and presenting results. In: Cochrane handbook for systematic reviews of diagnostic test accuracy Version 10, 2010. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
bmjopen-2021-057305supp001.pdf (775.7KB, pdf)
Data Availability Statement
All data relevant to the study are included in the article or uploaded as supplementary information.