Background:
The incidence of acute pulmonary embolism (APE) (especially early diagnosis) has increased annually in recent years, but the diagnosis of APE is a great challenge for every clinician. However, few studies have evaluated multiple diagnostic indicators simultaneously.
Methods:
A systematic search was performed using CNKI, Wan fang data, VIP, PubMed and Web of Science for studies on the diagnosis of pulmonary embolism published up to October 31, 2022. Using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2), we evaluated the risk of bias in included studies, and used a random-effects meta-analysis to obtain the summary sensitivity and specificity. The data that were extracted and calculated for this study included the first author, year of publication, country, study type, sample size, disease type, gold standard, diagnostic indicators and 4-compartment table data. We strictly followed the Preferred Reporting Items for Systematics reviews and Meta-Analysis (PRISMA) guidelines in this review.
Results:
This study included 30 articles with a total sample size of 8947 cases, involving 4 detection methods: D-dimer, Geneva rules, Wells rules, and lung imaging. The combined effect size showed that lung imaging had the highest diagnostic value (SEN = 0.95, SPE = 0.89), followed by D-dimer (SEN = 0.92, SPE = 0.60), Geneva rules (SEN = 0.78, SPE = 0.68), and Wells rules (SEN = 0.77, SPE = 0.67). The area of lung imaging was largest under the Summary Receiver Operator Characteristic (SROC) curve (AUC = 0.97), followed by Geneva rules (AUC = 0.80), Wells rules (AUC = 0.79), and D-dimer (AUC = 0.74).
Conclusion:
All 4 detection methods showed good ability to diagnose PE, and lung imaging was the best. Clinical trials are recommended to build an early decision-making model for the diagnosis of pulmonary embolism in order to increase the detection rate and improve prognosis.
Keywords: diagnosis, meta-analysis, pulmonary embolism
1. Introduction
Pulmonary embolism (PE) is a clinical condition in which the blood flow in the pulmonary arteries is obstructed by an embolus, resulting in pulmonary hypertension, right heart failure and impaired pulmonary oxygen uptake.[1] The embolus can be a thrombus, or can be made of fat, tumor, gas, etc, and can come from the vena cava system, the pulmonary artery itself, systemic arterial system where there is a left-right shunt, etc.[1] Acute pulmonary embolism (APE) is particularly prevalent in clinical practice.[2] The incidence of APE has increased each year in recent years.[2] This has been a topical issue of common concern in critical care, emergency, respiratory and cardiac disciplines. Some studies have shown that the average age of patients with PE is around 60 years.[3-5] Some studies even found 50 to be the age inflection point. There is a gradual increase in the proportion of middle-aged and older patients.[3,6,7]
In elderly patients with PE, the presence of a number of underlying conditions makes care difficult. Meanwhile, due to the low early diagnosis rate of PE, with 1 study showed that the early diagnosis and treatment rate of APE was only 7% for APE,[8] missed diagnosis makes PE more common.[9] Patients are often diagnosed at an advanced stage of the disease and cannot be cured, resulting in a high mortality rate. This means that early diagnosis of PE could provide great benefits to patients. Research and clinical practice have shown that specific treatments, such as early thrombolysis, could greatly improve the prognosis of patients with PE.[10]
The diagnosis of APE (especially early diagnosis) is a great challenge for every clinician. It lacks specific diagnostic methods such as ECG, myocardial enzyme profile for acute myocardial infarction or glutathione for liver impairment. Although computed tomography pulmonary angiography (CTPA) and pulmonary arteriography have a high confirmed diagnosis rate, they require that the patient is moved to the imaging department and contrast medium is intravenously received. they have the disadvantages of requiring patient movement, additional damage such as allergies and nephrotoxicity, and a diagnosis based on morphology, making it difficult to achieve dynamic and quantitative monitoring.[11] There is also a time lag. Several studies have shown that D-dimer, various clinical scales and lung imaging can be used to diagnose PE.[12] However, few studies have evaluated multiple diagnostic indicators simultaneously. Therefore, this paper provides a systematic review and meta-analysis of diagnostic methods for the early diagnosis of PE. We had published an outline of the protocol in the International Prospective Register of Systematic Reviews (PROSPERO) in 2021 (registration number: CRD42021254411).
2. Methods
2.1. Ethical issues
This systematic review does not require ethical approval or informed consent because there was no direct contact with individual patients, and only previously published data were included in the review.
2.2. Search criteria and strategy
Through electronic searches of different databases, including CNKI, Wanfang data, VIP, PubMed and Web of Science, relevant publications regarding the diagnosis of PE until December 31, 2022 were selected, no language limits. The search codes for each database are shown in supplement Table 1, http://links.lww.com/MD/J300.
2.3. Selection criteria
2.3.1. Inclusion criteria.
Prospective cohort studies or retrospective case-control studies investigating the diagnosis of PE.
Extraction or ability to calculate 4 cell table data from the provided data.
When multiple studies of the same population were included in the literature, the study with the largest sample size was included.
Diagnostic tests for D-dimer, Geneva rules, Wells rules, lung imaging.
2.3.2. Exclusion criteria.
In vitro and in vivo testing or animal studies literature.
Literature review or meta-analysis.
Literature with only an abstract or where the full text was not available.
Inadequate literature data.
2.3.3. Study selection.
Initially we searched the database (CNKI, Wanfang, VIP, PubMed, Web of Science) for documents, imported the collected documents into NoteExpress (v3.5.0.9054) and remove duplicates through the “check” function.
The initial screening was completed by quickly skimming the titles and abstracts of the collected literature through NoteExpress and eliminating studies that did not meet the inclusion criteria in terms of, for example, content and methods or study population.
We read the full text of the remaining literature and conduct a more rigorous screening process to eliminate literature with inadequate data or where the full text is not available, with 2 researchers independently screening the literature and discussing or seeking advice from the supervisor when there were inconsistent results.
2.4. Data extraction
At the end of the literature screening, the included literature was summarized and data were independently extracted from the literature by 2 researchers using Excel. The results discussed or an instructor advice asked promptly in case of inconsistency. Based on this extensive reading of the literature, the data that were extracted and calculated for this study included the first author, year of publication, country, study type, sample size, disease type, gold standard, diagnostic indicators and 4-compartment table data.
2.5. Quality assessment
Two researchers independently evaluated the quality of the literature using the QUADAS 2 scale, which mainly consists of 4 parts: case selection, trial to be evaluated, gold standard, case flow, and progress, and promptly discussed or asked the instructor for advice in the event of significant discrepancies.
2.6. Statistical analysis
We extracted the number of TP, FP, FN, and TN of each study to calculate the pooled sensitivity, specificity, PLR, NLR, DOR, and corresponding 95% confidence interval (CI). We also tested the pooled diagnostic value through the SROC curve and the area under the SROC curve (AUC). In the present study, Deeks’ funnel plot was also conducted to test publication bias. We assessed heterogeneity among the studies using the chi-squared and I2 tests. If P < .1 or I2 > 50%, heterogeneity was defined as significant. We also conducted meta-regression, subgroup and sensitivity analyses to identify potential sources of heterogeneity. We carried out all analyses using Stata 12.0 and Meta-DiSc 1.4, and a P value < .05 was considered statistically significant. The statistical analysis process is shown in Figure 1.
Figure 1.
Statistical analysis process.
3. Results
3.1. Description of literature inclusion
After a preliminary search of the database, a total of 6855 examples of the literatures were screened, and the title and abstract were read with the help of NoteExpress software for preliminary screening, followed by a reading of the full text for strict fine screening. Finally, 30 literature studies were included, with a total sample size of 8947 cases. The literature screening process is shown in Figure 2, and the basic information of the included literature is shown in Table 1.
Figure 2.
Flow diagram.
Table 1.
Study characteristics.
| First author (yr of publication) | Country | Sample size | Gold standard | Diagnostic indicators |
|---|---|---|---|---|
| Wang JC (2001)[13] | China | 30 | Pulmonary arteriography | Lung imaging |
| Marika (2002)[14] | Sweden | 115 | Clinical | Lung imaging |
| Reinartz (2004)[15] | Germany | 83 | Clinical | Lung imaging |
| Perrier (2005)[16] | Switzerland, France | 756 | Clinical | Geneva |
| Li J (2005)[17] | China | 42 | Clinical | D-dimer, Lung imaging |
| Wang Q (2005)[18] | China | 104 | Clinical | D-dimer, Lung imaging |
| Zhao HL (2005)[19] | China | 137 | Clinical | D-dimer, Lung imaging |
| Righini (2006)[20] | Switzerland | 922 | Clinical | Wells |
| Tang HP (2007)[21] | China | 352 | Clinical | Geneva, Wells |
| Wang F (2007)[22] | China | 53 | Clinical | Lung imaging |
| Gutte (2009)[23] | Denmark | 77 | Clinical | Lung imaging |
| Wang JG (2009)[24] | China | 688 | CTPA | Geneva, Wells |
| Hemnes (2010)[25] | USA | 298 | Clinical | Wells |
| Xiong GJ (2011)[26] | China | 243 | CTPA | Geneva, Wells |
| Bai Y (2013)[27] | China | 191 | Clinical | Lung imaging |
| Su F (2013)[28] | China | 79 | Clinical | Lung imaging |
| Zhao M (2013)[29] | China | 317 | Clinical | D-dimer |
| Dimarca (2015)[30] | Italy | 101 | CTPA | Geneva, Wells |
| Li WL (2015)[31] | China | 704 | CTPA | Geneva, Wells |
| Zhou XJ (2015)[32] | China | 85 | Clinical | D-dimer, Lung imaging |
| Lu X (2016)[33] | China | 156 | Clinical | D-dimer, Lung imaging |
| Cao DL (2019)[34] | China | 87 | CTPA | Wells |
| Jing WH (2019)[35] | China | 139 | CTPA | Wells |
| Li DY (2019)[36] | China | 313 | CTPA | D-dimer, Geneva |
| Zhu YW (2019)[37] | China | 112 | CTPA | Wells |
| Depooter (2021)[38] | France | 1082 | CTPA | D-dimer |
| Han J (2021)[39] | China | 514 | CTPA | Geneva |
| Liu J (2021)[40] | China | 946 | CTPA | Geneva, Wells |
| Shao ZB (2021)[41] | China | 85 | CTPA | D-dimer, Geneva |
| Zhang YJ (2021)[42] | China | 136 | CTPA | D-dimer, Geneva |
Except for 1 article, which is a retrospective case-control study,[34] the rest of the literatures is prospective cohort studies. The included literature contains 4 types of detection methods, namely D-dimer, Geneva rules, Wells rules, and lung imaging.
Except for 1 article, which is a retrospective case-control study,[34] the rest of the literatures is prospective cohort studies. The included literature contains 4 types of detection methods, namely D-dimer, Geneva rules, Wells rules, and lung imaging.
3.2. Assessment of risk of bias and applicability
Review Manager 5.3 was used to evaluate the quality of the included literature. The results of the QUADAS 2 scale evaluation showed that the overall quality of the literature included in this study was satisfactory, with 96.7% (29/30) of the studies being prospective and clinically applicable. In the field of case selection, more than half of the literature had a low risk of bias, although some of the literatures did not describe the baseline data of the included patients, meaning that there is a lack of bias in their judgment; in the field of diagnostic testing, most literatures determined the threshold before the test, and only one round of threshold selection formed after the test results,[13] but there was no clear description of whether the diagnostic test was performed under blinded method; in the field of gold standard, for the literature with clinical diagnosis as the gold standard, it was not possible to implement a blinded method for gold standard diagnosis, so there was a certain risk of bias; in the field of cases flow and appropriate time interval, most of the literature had accurate descriptions and the bias was small. The results of the quality evaluation are shown in Figure 3 and Table 2.
Figure 3.
Overall quality assessment of the included literature
Table 2.
Single article quality evaluation of the included literature.
| Yr | Author | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 2001 | Wang JC[13] | U | Y | U | U | - | Y | U | Y | Y | Y |
| 2002 | Marika[14] | Y | Y | Y | Y | - | Y | N | Y | Y | Y |
| 2004 | Reinartz[15] | U | Y | Y | Y | - | Y | U | Y | Y | Y |
| 2005 | Perrier[16] | Y | Y | Y | Y | Y | Y | N | Y | Y | Y |
| 2005 | Li J[17] | U | Y | U | U | Y | Y | U | Y | Y | Y |
| 2005 | Wang Q[18] | U | Y | U | Y | Y | Y | N | Y | Y | Y |
| 2005 | Zhao HL[19] | Y | Y | U | U | Y | Y | U | Y | Y | Y |
| 2006 | Righini[20] | Y | Y | Y | Y | Y | Y | U | Y | Y | Y |
| 2007 | Tang HP[21] | Y | Y | U | U | Y | Y | U | Y | Y | Y |
| 2007 | Wang F[22] | Y | Y | Y | Y | - | Y | N | Y | Y | Y |
| 2009 | Gutte[23] | Y | Y | Y | Y | - | Y | N | Y | Y | Y |
| 2009 | Wang JG[24] | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y |
| 2010 | Hemnes[25] | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y |
| 2011 | Xiong GJ[26] | U | Y | Y | U | Y | Y | U | Y | Y | Y |
| 2013 | Bai Y[27] | Y | Y | U | Y | - | Y | N | Y | Y | Y |
| 2013 | Su F[28] | Y | Y | U | Y | - | Y | U | Y | Y | Y |
| 2013 | Zhao M[29] | Y | Y | Y | U | Y | Y | U | U | Y | Y |
| 2015 | Dimarca[30] | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y |
| 2015 | Li WL[31] | Y | Y | Y | U | Y | Y | U | Y | Y | Y |
| 2015 | Zhou XJ[32] | Y | Y | U | Y | Y | Y | N | U | Y | Y |
| 2016 | Lu X[33] | Y | Y | Y | U | Y | Y | U | Y | Y | Y |
| 2019 | Cao DL[34] | Y | N | Y | N | Y | Y | Y | Y | Y | Y |
| 2019 | Jing WH[35] | Y | Y | Y | U | Y | Y | U | Y | Y | Y |
| 2019 | Li DY[36] | Y | Y | Y | U | Y | Y | U | U | Y | Y |
| 2019 | Zhu YW[37] | Y | Y | U | U | Y | Y | U | Y | Y | Y |
| 2021 | Depooter[38] | Y | Y | Y | Y | N | Y | N | Y | N | Y |
| 2021 | Han J[39] | Y | Y | Y | U | Y | Y | U | Y | Y | Y |
| 2021 | Liu J[40] | Y | Y | Y | Y | Y | Y | Y | Y | Y | Y |
| 2021 | Shao ZB[41] | Y | Y | Y | U | Y | Y | U | Y | Y | Y |
| 2021 | Zhang YJ[42] | Y | Y | U | U | Y | Y | U | U | Y | Y |
Note: Y: Yes, N: No, U: Unclear, -: This entry does not apply to this study, 1: Was a consecutive or random sample of patients enrolled? 2: Was a case-control design avoided? 3: Did the study avoid inappropriate exclusions? 4: Were the index test results interpreted without knowledge of the results of the reference standard? 5: If a threshold was used, was it pre-specified? 6: Is the reference standard likely to correctly classify the target condition? 7: Were the reference standard results interpreted without knowledge of the results of the index test? 8: Was there an appropriate interval between index test(s) and reference standard? 9: Did all patients receive a reference standard? 10: Were all patients included in the analysis?[43]
3.3. Threshold effect
Meta-DiSc 1.4 was used to analyze the threshold effect of the data. The results showed that the Spearman correlation coefficient for Wells rules was > 0.7, P < .05 and the ROC curve had a “shoulder-arm” shape (Fig. 3), indicating a threshold effect. There was no threshold effect on D-dimer, Geneva rules, and lung imaging (Table 3 and Fig. 3).
Table 3.
Threshold effect analysis.
| Diagnostic indicators | Spearman correlation coefficient | P |
|---|---|---|
| D-dimer | 0.355 | .285 |
| Geneva rules | 0.573 | .066 |
| Wells rules | 0.782 | .004 |
| Lung imaging | −0.446 | .147 |
3.4. Combined effect value
The results of the meta-analysis are shown in Table 4 and from Supplementary Figure 1 to Supplementary Figure 9, http://links.lww.com/MD/J301. D-dimer: sensitivity (SEN) = 0.92 (95% CI: 0.84–0.97), specificity (SPE) = 0.60 (95% CI: 0.54–0.67), positive likelihood ratio (+LR) = 2.34 (95% CI: 1.98–2.77), negative likelihood ratio (−LR) = 0.13 (95% CI: 0.06–0.27), diagnostic ratio (DOR) = 19 (95% CI: 8–43). Geneva rules: SEN = 0.78 (95% CI: 0.67–0.86), SPE = 0.68 (95% CI: 0.47–0.83), +LR = 2.41 (95% CI: 1.43–4.08), −LR = 0.33 (95% CI: 0.24–0.46), DOR = 7 (95% CI: 4–14). Wells rules: SEN = 0.77 (95% CI: 0.70–0.83), SPE = 0.67 (95% CI: 0.56–0.76), +LR = 2.30 (95% CI: 1.83–2.90), −LR = 0.35 (95% CI: 0.30–0.40), DOR = 7 (95% CI: 5–8). Lung imaging: SEN = 0.95 (95% CI: 0.91–0.97), SPE = 0.89 (95% CI: 0.82–0.93), +LR = 8.56 (95% CI: 5.04–14.54), −LR = 0.06 (95% CI: 0.03–0.11), DOR = 144 (95% CI: 54–385).
Table 4.
Meta-analysis of combined effect values for diagnostic indicators (95% CI).
| Diagnostic indicators | SEN | SPE | +LR | −LR | DOR |
|---|---|---|---|---|---|
| D-dimer | 0.92 (0.84, 0.97) | 0.60 (0.54, 0.67) | 2.34 (1.98, 2.77) | 0.13 (0.06, 0.27) | 19 (8, 43) |
| Geneva rules | 0.78 (0.67, 0.86) | 0.68 (0.47, 0.83) | 2.41 (1.43, 4.08) | 0.33 (0.24, 0.46) | 7 (4, 14) |
| Wells rules | 0.77 (0.70, 0.83) | 0.67 (0.56, 0.76) | 2.30 (1.83, 2.90) | 0.35 (0.30, 0.40) | 7 (5, 8) |
| Lung imaging | 0.95 (0.91, 0.97) | 0.89 (0.82, 0.93) | 8.56 (5.04, 14.54) | 0.06 (0.03, 0.11) | 144 (54, 385) |
3.5. SROC curve
The SROC curve of the D-dimer is shown in Supplementary Figure 10, http://links.lww.com/MD/J302, AUC = 0.74 (95% CI: 0.70–0.78). The SROC curve of Geneva rules is shown in Supplementary Figure 11, http://links.lww.com/MD/J303, AUC = 0.80 (95% CI: 0.77–0.84), the SROC curve of Wells rules is shown in Supplementary Figure 12, http://links.lww.com/MD/J304, AUC = 0.79 (95% CI: 0.75–0.82), the SROC curve of lung imaging is shown in Supplementary Figure 13, http://links.lww.com/MD/J305, AUC = 0.97 (95% CI: 0.95–0.98).
3.6. Heterogeneity test and subgroup analysis
Subgroup analysis was performed on the data according to publication time, bias level, country, study type, threshold, disease type and gold standard, as shown from Table 5–8, and from Supplementary Figure 14 to Supplementary Figure 15, http://links.lww.com/MD/J306. The results showed that during the detection of D-dimer, the subgroup heterogeneity of different thresholds was significantly reduced, and the OR value of the age-adjusted threshold was significantly increased. The literature on Wells rules and lung imaging diagnosis of PE was published according to the subgroup heterogeneity of time was reduced, suggesting that there are some differences in the methodology of Wells rules in different regions.
Table 5.
Subgroup analysis of D-dimer.
| Characteristics | Number of literatures | OR (95% CI) | P | I2 (%) |
|---|---|---|---|---|
| Overall | 11 | 16.90 (8.57, 33.34) | <.001 | 74.7 |
| Bias | ||||
| High | 5 | 17.57 (4.85, 63.73) | <.001 | 80.5 |
| Not high | 6 | 18.49 (9.38, 36.45) | .040 | 57.0 |
| Country | ||||
| China | 9 | 12.88 (6.72, 24.68) | <.001 | 72.8 |
| Overseas | 2 | 160.39 (32.04, 802.94) | .782 | 0.0 |
| Threshold | ||||
| Traditional threshold | 7 | 9.37 (4.62, 18.99) | .014 | 62.3 |
| Age-corrected threshold | 4 | 42.30 (15.25, 117.38) | .038 | 64.3 |
| Disease | ||||
| PE | 9 | 13.90 (6.53, 29.57) | .001 | 70.8 |
| ATPE | 2 | 34.93 (6.74, 181.12) | .019 | 81.9 |
| Gold standard | ||||
| Clinical | 5 | 9.84 (5.48, 17.67) | .251 | 25.6 |
| CTPA | 6 | 32.59 (9.77, 108.79) | <.001 | 84.7 |
CI = confidence interval, CTPA = computed tomography pulmonary angiography, I2 = I-square test.
Table 8.
Subgroup analysis of lung imaging.
| Characteristics | Number of literatures | OR (95% CI) | P | I2 (%) |
|---|---|---|---|---|
| Overall | 12 | 100.60 (45.60, 221.96) | .002 | 62.0 |
| Year | ||||
| Before 2004 | 3 | 583.99 (70.00, 4871.97) | .222 | 33.5 |
| After 2005 | 9 | 73.10 (33.66, 158.79) | .011 | 59.8 |
| Bias | ||||
| High | 8 | 114.33 (39.92, 327.50) | .004 | 66.9 |
| Not high | 4 | 86.68 (22.12, 339.62) | .053 | 60.9 |
| Country | ||||
| China | 9 | 69.44 (31.92, 151.06) | .015 | 57.9 |
| Overseas | 3 | 527.26 (81.49, 3411.41) | .216 | 34.7 |
| Disease | ||||
| PE | 11 | 120.66 (52.42, 277.71) | .005 | 60.7 |
| ATPE | 1 | 20.95 (5.43, 80.91) | - | - |
| Gold standard | ||||
| Pulmonary arteriography | 1 | 139.40 (6.03, 3220.28) | - | - |
| Clinical | 10 | 112.97 (43.17, 295.59) | .001 | 68.5 |
| CTPA | 1 | 57.33 (19.66, 167.23) | - | - |
CI = confidence interval, CTPA = computed tomography pulmonary angiography, I2 = I-square test.
3.7. Publication bias and sensitivity analysis
The results of publication bias analysis are shown in Supplementary Figures 16 to Supplementary Figures 19, http://links.lww.com/MD/J307. Deeks’ test found that the P values of D-dimer, Geneva rules and lung imaging were .16, .85, and .96 respectively. As P > .1, there was no publication bias. The P value of Wells rules was .07 < .1, so there was publication bias. Sensitivity analysis with sequential exclusion of literature revealed no significant change in any of the combined effect values and no significant decrease in heterogeneity, suggesting good stability for the meta-analysis in this study.
4. Discussion
In this study, a meta-analysis was performed on 30 papers related to the diagnosis of PE, including 11 about D-dimer, 11 papers about Geneva rules, 11 papers about Wells rules, and 12 papers about lung imaging. The best performance was achieved by lung imaging, with the best sensitivity, specificity, +LR and −LR of all 4.
In the literature included in this study, 2 diagnostic methods were used for the diagnosis of PE by D-dimer: one is the traditional threshold chosen to judge patients, that is D-dimer, where >500 μg/L was judged positive and <500 μg/L was judged negative; the other is age-corrected D-dimer, which was referenced in the 2019 European Society of Cardiology Guidelines for the diagnosis and management of acute pulmonary embolism,[44] for patients aged >50 years, where the threshold is no longer taken to be 500 μg/L, but age (in years) × 10 μg/L. A modified version of the Geneva rules is used to diagnose PE, with entries and scoring tables as shown in Supplementary Table 2, http://links.lww.com/MD/J308, with a total score of 0 to 2 meaning low probability, 3 to 6 meaning moderate probability and ≥7 meaning high probability, where patients with low probability are classified as negative and those with moderate and high probability are classified as positive.[45] The entries and scores for Wells rules for the diagnosis of PE are shown in Supplementary Table 3, http://links.lww.com/MD/J309, with a total score of 0 to 3 being low probability, 4 to 10 being moderate probability and ≥11 being high probability, where again patients with low probability are classified as negative and those with moderate and high probability are classified as positive.[46] The results of the lung imaging test were based on the diagnostic criteria of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED II), and patients with low probability and very low probability were classified as negative, while those with moderate probability and high probability were classified as positive.[47] Lung scintigraphy has stood the test of time as a reliable and validated examination for the determination of PE.[48] In the present study, most of the study subjects conformed to the standard study population and the same diagnostic thresholds were adopted to diagnose PE, effectively avoiding bias to a certain extent. Meta-analysis revealed that only Wells rules had a threshold effect for the diagnosis of PE, but there was significant heterogeneity in all 4 tests, so it was inferred that there were clinical, statistical and methodological differences between the D-dimer, Geneva rules and lung imaging for the diagnosis of PE. When the effect values were combined, lung imaging was found to have the best sensitivity and specificity of all 4 (SEN = 0.95, SPE = 0.89) and a larger diagnostic ratio, implying that it was more capable of diagnosing and excluding patients than the other 3. Although the sensitivity of the D-dimer was also higher (SEN = 0.92), its specificity did not dominate (SPE = 0.60), implying that although its ability to diagnose patients was stronger, its ability to exclude patients was weaker, whereas the sensitivity and specificity of Geneva rules and Wells rules, although not dominant, both had an AUC > 0.7, indicating better trueness and its use in They can still be chosen as an indicator reference in clinical practice. The subgroup analysis showed that the heterogeneity of D-dimer was significantly reduced when analyzed in different subgroups according to the threshold, indicating that the threshold is one of the sources of heterogeneity, and the subgroup analysis showed that the age-corrected threshold has more diagnostic advantages than the traditional threshold; the heterogeneity of lung imaging was significantly reduced when analyzed in different subgroups according to the country of origin, indicating that the different patient conditions or examination methods, machines and standards in different countries may be one of the sources of heterogeneity. This suggests that differences in patient status or examination methods, machines and standards between countries may be a source of heterogeneity.
As all 4 have good trueness and good sensitivity specificity, it is expected that the 4 tests can be integrated to construct a model for the joint diagnosis of PE, which may improve the diagnosis of early PE as a way to improve patient prognosis. For Geneva rules and Wells rules, a comparison of the 2 scales revealed that several variables overlapped, such as previous history of DVT or PE, history of surgery within 1 month, tumor, hemoptysis, etc. A new scale could be constructed to improve the sensitivity and specificity by integrating the different entries of the 2 scales, such as unilateral lower limb pain, age >65 years, clinical symptoms of DVT, etc. The newly constructed scale can be combined with D-dimer (using age-corrected thresholds) and lung imaging to diagnose PE and build a multifactorial decision model for the early diagnosis of PE, but due to the insufficient number of existing studies, it is recommended that a large number of clinical trials can be conducted to supplement the clinical evidence.
At the same time, there are certain limitations in this study, such as the fact that, with regard to Geneva rules and Wells rules, although the criteria for determining the total score were consistent in most studies, there were differences between studies with regard to the scores of specific items, and some potential further causes which results in high Wells and Geneva score didn’t exclude, which may be a source of heterogeneity, and the fact that the review of the PE examination relied on the subjective judgement of the physician, which was not described in some of the literature in relation to its blinding, Some studies classified low probability patients as positive or moderate probability patients as undiagnosed, which has an impact on the results of this study. Moreover, not all studies considered and excluded APE related to COVID-19 which is 1 reason for the increased incidence of APE in recent years. And the increase of D-dimers due to the comorbidities of the patients was also not excluded. There is 1 study comes from USA, and 7 studies are from Europe. Approximately, over 60% of the crowd are Chinese. This situation maybe will weaken its universal applicability.
5. Conclusion
In summary, D-dimer, Geneva rules, Wells rules and lung imaging are all of good diagnostic value, with lung imaging being the best of the 4, however considering that it is less convenient than the other 3 in terms of clinical use. Various factors should be considered when constructing the decision model, and clinicians need to be flexible in choosing the appropriate diagnostic method for their patients. It is expected that clinical trials will be conducted to provide data for the construction of this multifactorial decision model.
Table 6.
Subgroup analysis of Geneva rules.
| Characteristics | Number of literatures | OR (95% CI) | P | I2 (%) |
|---|---|---|---|---|
| Overall | 11 | 7.02 (4.19, 11.75) | <.001 | 88.1 |
| Year | ||||
| Before 2015 | 6 | 6.84 (2.92, 16.00) | <.001 | 90.2 |
| After 2016 | 5 | 7.47 (3.60, 15.51) | <.001 | 87.6 |
| Bias | ||||
| Low | 3 | 3.51 (2.40, 5.12) | .126 | 51.8 |
| Medium | 6 | 7.76 (3.59, 16.73) | <.001 | 89.0 |
| High | 2 | 26.90 (1.29, 560.66) | <.001 | 93.2 |
| Country | ||||
| China | 9 | 5.90 (3.74, 9.32) | <.001 | 84.0 |
| Overseas | 2 | 15.98 (0.28, 917.39) | <.001 | 96.5 |
| Disease | ||||
| PE | 6 | 6.09 (2.85, 13.01) | <.001 | 89.8 |
| ATPE | 5 | 8.52 (3.92, 18.53) | <.001 | 86.5 |
| Gold standard | ||||
| Clinical | 2 | 26.90 (1.29, 560.66) | <.001 | 93.2 |
| CTPA | 9 | 5.40 (3.40, 8.56) | <.001 | 84.6 |
CI = confidence interval, CTPA = computed tomography pulmonary angiography, I2 = I-square test.
Table 7.
Subgroup analysis of Wells rules.
| Characteristics | Number of literatures | OR (95% CI) | P | I2 (%) |
|---|---|---|---|---|
| Overall | 11 | 6.59 (5.60, 7.74) | .147 | 31.6 |
| Year | ||||
| Before 2015 | 7 | 7.24 (6.03, 8.70) | .307 | 16.1 |
| After 2016 | 4 | 5.11 (3.66, 7.11) | .241 | 28.5 |
| Bias | ||||
| Low | 3 | 6.27 (4.76, 8.25) | .003 | 82.4 |
| Medium | 5 | 6.90 (5.56, 8.57) | .746 | 0.0 |
| High | 3 | 6.24 (3.78, 10.30) | .511 | 0.0 |
| Country | ||||
| China | 8 | 6.31 (5.21, 7.65) | .358 | 9.3 |
| Overseas | 3 | 7.42 (5.50, 10.02) | .045 | 67.8 |
| Study type | ||||
| Cohort study | 10 | 6.69 (5.67, 7.88) | .144 | 33.0 |
| Case-control | 1 | 4.05 (1.61, 10.16) | - | 0.0 |
| Disease | ||||
| PE | 8 | 6.19 (5.12, 7.49) | .069 | 46.6 |
| ATPE | 3 | 7.78 (5.73, 10.56) | .948 | 0.0 |
| Gold standard | ||||
| Clinical | 3 | 6.75 (5.00, 9.10) | .877 | 0.0 |
| CTPA | 8 | 6.53 (5.39, 7.91) | .045 | 51.2 |
CI = confidence interval, CTPA = computed tomography pulmonary angiography, I2 = I-square test.
Acknowledgments
Thanks to all the participants and clinical researchers involved in the publications cited in this review. Thanks to all the peer reviewers who contributed to the continuous improvement of this article.
Author contributions
Conceptualization: Xiujuan Xu, Conghua JI.
Data curation: Minjun Ma, Yu Li.
Funding acquisition: Xiujuan Xu, Conghua Ji.
Investigation: Minjun Ma, Yu Li.
Methodology: Yu Li, Xiujuan Xu, Conghua Ji.
Project administration: Xiujuan Xu, Conghua Ji.
Resources: Yu Li.
Software: Minjun Ma, Yu Li.
Validation: Yu Li.
Writing – original draft: Minjun Ma.
Writing – review & editing: Xiujuan Xu, Conghua Ji.
Supplementary Material
Abbreviations:
- APE
- acute pulmonary embolism
- CI
- confidence interval
- CTPA
- computed tomography pulmonary angiography
- PE
- pulmonary embolism
This work was supported by the Department of Science and Technology of Zhejiang Province (No. 2023C25012), and the Health Commission of Zhejiang Province (No. 2019KY348 and 2021KY606), China.
All data generated or analyzed during this study are included in this published article [and its supplementary information files].
Supplemental Digital Content is available for this article.
The authors have no conflicts of interest to disclose.
How to cite this article: Ma M, Li Y, Xu X, Ji C. Early diagnosis for pulmonary embolism: A systematic review and meta-analysis. Medicine 2023;102:28(e34352).
Contributor Information
Minjun Ma, Email: minjun.ma@alu.zcmu.edu.cn.
Yu Li, Email: zhejiangrenaixuexi@163.com.
Conghua Ji, Email: jchi2005@126.com.
References
- [1].Konstantinides SV, Torbicki A, Agnelli G, et al. 2014 ESC Guidelines on the diagnosis and management of acute pulmonary embolism The Task Force for the Diagnosis and Management of Acute Pulmonary Embolism of the European Society of Cardiology (ESC). Eur Heart J. 2014;35:3033–69, 3069a. [DOI] [PubMed] [Google Scholar]
- [2].Konstantinides SV, Barco S, Lankeit M, et al. Management of pulmonary embolism an update [J]. J Am Coll Cardiol. 2016;67:976–90. [DOI] [PubMed] [Google Scholar]
- [3].Fang J. Epidemiological Study of Pulmonary Thromboembolism [D]; Zhejiang University, 2016. [Google Scholar]
- [4].Li X. Epidemiology and Risk Factor Analysis of Pulmonary Embolism [D]; Jilin University, 2016. [Google Scholar]
- [5].Xiao H. Clinical Characteristics and Prognostic Factors of Non-High-Risk Acute Pulmonary Embolism in Different Age Groups [D]; Southwest Medical University, 2020. [Google Scholar]
- [6].Castelli R, Bergamaschini L, Sailis P, et al. The impact of an aging population on the diagnosis of pulmonary embolism: comparison of young and elderly patients. Clin Appl Thromb Hemost. 2009;15:65–72. [DOI] [PubMed] [Google Scholar]
- [7].Righini M, Goehring C, Bounameaux H, et al. Effects of age on the performance of common diagnostic tests for pulmonary embolism. Am J Med. 2000;109:357–61. [DOI] [PubMed] [Google Scholar]
- [8].Schultz MA, Doty M. Why the history and physical examination still matter. Jaapa-J Am Acad Phys Assistants. 2016;29:41–5. [DOI] [PubMed] [Google Scholar]
- [9].Yanhui L, Zhao Y, Liu G, et al. Analysis of the current status of pulmonary embolism diagnosis. Chinese J Circ. 2001;01:45–7. [Google Scholar]
- [10].Teleb M, Porres-Aguilar M, Anaya-Ayala JE, et al. Potential role of systemic thrombolysis in acute submassive intermediate risk pulmonary embolism: review and future perspectives. Ther Adv Cardiovasc Dis. 2016;10:103–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Murthy C, Davis R, N KCF, et al. The impact of an electronic clinical decision support for pulmonary embolism imaging on the efficiency of computed tomography pulmonary angiography utilisation in a resource-limited setting. S Afr Med J. 2015;106:62–4. [DOI] [PubMed] [Google Scholar]
- [12].Walker RH. The value of lung scanning in the diagnosis of pulmonary embolism in the elderly. Geriatrics. 1969;24:81–94. [PubMed] [Google Scholar]
- [13].Wang JC, Mi HZ, Wang SQ, et al. Comparative study of pulmonary perfusion/ventilation imaging and pulmonary arteriography in the diagnosis of pulmonary embolism. Chinese J Nucl Med. 2001;04:24–6. [Google Scholar]
- [14].Bajc M, Albrechtsson U, Olsson C-G, et al. Comparison of ventilation/perfusion scintigraphy and helical CT for diagnosis of pulmonary embolism; strategy using clinical data and ancillary findings. Clin Physiol Funct Imaging. 2002;22:392–7. [DOI] [PubMed] [Google Scholar]
- [15].Reinartz P, Wildberger JE, Schaefer W, et al. Tomographic imaging in the diagnosis of pulmonary embolism: a comparison between V/Q lung scintigraphy in SPECT technique and multislice spiral CT. J Nucl Med. 2004;45:1501–8. [PubMed] [Google Scholar]
- [16].Perrier A, Roy P-M, Sanchez O, et al. Multidetector-row computed tomography in suspected pulmonary embolism. N Engl J Med. 2005;352:1760–8. [DOI] [PubMed] [Google Scholar]
- [17].Juan L, Liu B, Zhao F, et al. Value of pulmonary ventilation/perfusion imaging and HCTPA in the diagnosis of pulmonary artery embolism. Chinese J Nucl Med. 2005;02:105–7. [Google Scholar]
- [18].Wang X, Huang L, Qin S, et al. Diagnostic value of pulmonary ventilation/perfusion imaging combined with plasma D-dimer analysis in pulmonary embolism. Chinese J Nucl Med. 2005;01:49–51. [Google Scholar]
- [19].Zhao H, Li S. Diagnosis of pulmonary embolism by V/Q and plasma D-dimer analysis. Shanxi J Med. 2005;11:71–2. [Google Scholar]
- [20].Righini M, Le Gal G, Perrier A, et al. Clinical probability assessment of pulmonary embolism by the Wells’ score: is the easiest the best?. J Thromb Haemost. 2006;4:702–4. [DOI] [PubMed] [Google Scholar]
- [21].Tang HP, Han W, Li SB, et al. Application of clinical scales in the diagnosis of acute pulmonary thromboembolism. Chinese J Respir Critical Care. 2007;05:364–8. [Google Scholar]
- [22].Wang F, Wang ZZ, Zhang LEL, et al. A comparative study of one-day pulmonary ventilation/perfusion imaging and multi-row spiral CT for the diagnosis of pulmonary embolism. J Soochow Univ (Medical Edition). 2007;06:892–4 + 915. [Google Scholar]
- [23].Gutte H, Mortensen J, Jensen CV, et al. Detection of pulmonary embolism with combined ventilation-perfusion SPECT and Low-Dose CT: head-to-head comparison with multidetector CT angiography [J]. J Nucl Med. 2009;50:1987–92. [DOI] [PubMed] [Google Scholar]
- [24].Wang J, Zhu L, Liu M, et al. Comparative analysis of the predictive value of three acute pulmonary embolism scores. Chinese J Pract Intern Med. 2009;29:322–4. [Google Scholar]
- [25].Hemnes AR, Newman AL, Rosenbaum B, et al. Bedside end-tidal CO2 tension as a screening tool to exclude pulmonary embolism. Eur Respir J. 2010;35:735–41. [DOI] [PubMed] [Google Scholar]
- [26].Xiong GJ, Qi QIAN. Diagnostic value of clinical score and D-dimer test for acute pulmonary embolism. Shandong Med. 2011;51:47–8. [Google Scholar]
- [27].Bai Y, Cheng J, Yuanqing H, et al. Comparative study of pulmonary ventilation and perfusion imaging versus CT pulmonary arteriography in the diagnosis of pulmonary embolism. J Pract Clin Med. 2013;17:147–8. [Google Scholar]
- [28].Su F, Fan Q, Chen GQ. Diagnostic value of pulmonary perfusion imaging combined with 64-row spiral CT scan for acute pulmonary embolism. Fujian J Med. 2013;35:116–8. [Google Scholar]
- [29].Zhao M, Xuan C, Xia RQ, et al. Diagnostic value of plasma D-dimer in patients with suspected pulmonary embolism. Chinese J Clin Physicians. (electronic version). 2013;7:10488–91. [Google Scholar]
- [30].Di Marca S, Cilia C, Campagna A, et al. Comparison of wells and revised geneva rule to assess pretest probability of pulmonary embolism in high-risk hospitalized elderly adults. J Am Geriatr Soc. 2015;63:1091–7. [DOI] [PubMed] [Google Scholar]
- [31].Li WL, Wang ZA, Tian XC, et al. Exploring the diagnostic value of Wells score, Geneva score and modified Geneva score applied to patients with pulmonary thromboembolism in China. Chinese J Modern Med. 2015;25:51–5. [Google Scholar]
- [32].Zhou X, Zhu Y, Zhao Y, et al. Diagnostic value of pulmonary ventilation/perfusion imaging combined with D-dimer in pulmonary embolism. Marker Immunoassay Clin. 2015;22:887–9 + 6. [Google Scholar]
- [33].Lu X, Meng JJ, Xie SF, et al. Value of combined semi-quantitative analysis of plasma D-dimer for the diagnosis of pulmonary embolism by SPECT V/Q tomography. Isotope. 2016;29:204–8. [Google Scholar]
- [34].Cao DAL. Diagnostic value of simplified Wells score, D-dimer and ultrasound for pulmonary thromboembolism. Huaihai Med. 2019;37:554–6. [Google Scholar]
- [35].Jing W, Hongyu L, Guan Y. Comparison of the diagnostic value of Wells score and YEARS method for pulmonary embolism. J Jilin Univ (Medical Edition). 2019;45:88–93. [Google Scholar]
- [36].Danyang L, Mao Y, Xiaofang L. Diagnostic value of age-corrected D-dimer combined with simplified Geneva score for acute pulmonary embolism in the elderly. J Pract Cardiopulm Vasc Dis. 2019;27:58–62. [Google Scholar]
- [37].Mao Y, Zhu A, Chu A, et al. Value of Wells score combined with electrocardiography and cardiac enzyme profile in the diagnosis of pulmonary thromboembolism. J Clin Psychosomatic Dis. 2019;25:95–8. [Google Scholar]
- [38].De Pooter N, Brionne-Francois M, Smahi M, et al. Age-adjusted D-dimer cut-off levels to rule out venous thromboembolism in patients with non-high pre-test probability: clinical performance and cost-effectiveness analysis. J Thromb Haemost. 2021;19:1271–82. [DOI] [PubMed] [Google Scholar]
- [39].Han J, Shao X, Liu M, et al. Clinical application of diagnostic strategy based on YEARs score in acute pulmonary embolism. Chinese J Pract Intern Med. 2021;41:396–400. [Google Scholar]
- [40].Liu J, Sun L, Xiang H, et al. A comparative study of the diagnostic value of Wells score, Geneva score and YEARS method in patients hospitalized with suspected pulmonary embolism. J Interv Radiol. 2021;30:552–6. [Google Scholar]
- [41].Shao ZEB, Qunxing L, Zhao XH, et al. Clinical significance of modified Geneva score with Daniel ECG score and D-dimer in predicting acute pulmonary embolism. Chinese J Clin Health Care. 2021;24:217–20. [Google Scholar]
- [42].Zhang Y, Zhang J, Wang X, et al. Diagnostic value of age-corrected D-dimer combined with simplified Geneva score for acute pulmonary embolism in the elderly. Western Med. 2021;33:1797–801. [Google Scholar]
- [43].Yanji Q, Yang Z, Sun F, et al. Risk of bias assessment series: (VI) diagnostic tests. Chinese J Epidemiol. 2018;39:524–31. [Google Scholar]
- [44].Konstantinides SV, Meyer G, Becattini C, et al. 2019 ESC Guidelines for the dignosis and management of acute pulmonary embolism development Toed in colaboration with the European Respiratory Society (ERS). Eur Heart J. 2020;41:543–603. [DOI] [PubMed] [Google Scholar]
- [45].Le Gal G, Righini M, Roy P-M, et al. Prediction of pulmonary embolism in the emergency department: the revised Geneva score. Ann Intern Med. 2006;144:165–71. [DOI] [PubMed] [Google Scholar]
- [46].Wells PS, Anderson DR, Rodger M, et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer. Thromb Haemost. 2000;83:416–20. [PubMed] [Google Scholar]
- [47].Investigators P. Value of the ventilation/perfusion scan in acute pulmonary embolism. Results of the prospective investigation of pulmonary embolism diagnosis (PIOPED). JAMA. 1990;263:2753–9. [DOI] [PubMed] [Google Scholar]
- [48].Le Roux PY, Le Pennec R, Salaun PY, et al. Scintigraphic diagnosis of acute pulmonary embolism: from basics to best practices. Semin Nucl Med. 2023:S0001-2998(23)00030-2. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.