1. Introduction

Pulmonary embolism (PE) represents a major medical emergency with potentially serious consequences if not diagnosed and treated promptly [1]. This disorder, often caused by a blood clot that blocks one or more arteries in the lungs, is a crucial issue for health systems around the world, particularly in countries with limited resources [2]. In Burkina Faso, as in many other developing countries, the effective management of cardiovascular diseases and especially PE in particular is complicated by challenges such as lack of resources, limited access to advanced diagnostic technologies, and sometimes late recognition of symptoms [3]. In this context, clinical probability scores are of capital importance in the diagnostic process. The Wells score is a clinical tool developed to help assess the risk of pulmonary embolism in patients with compatible symptoms [4]. It uses a series of clinical criteria and risk factors to stratify patients into different risk categories [4]. Its validity and effectiveness as a screening tool have been widely documented in clinical contexts in developed countries, but its application and performance have not been validated in black African subjects [5]. This study aims at evaluating the performance of the Wells score in the diagnosis of pulmonary embolism at Bogodogo University Hospital Center.

2. Materials and Methods

2.1. Type, period and setting of the study

This was a descriptive and analytical cross-sectional study with diagnostic purposes, carried out from January 1, 2021 to July 31, 2023 at the Bogodogo University Hospital Center in Ouagadougou, Burkina Faso. The cardiology services and the pulmonology unit served as a framework.

2.2. Inclusion criteria

Patients hospitalized for clinical suspicion of PE and in whom a chest CT angiogram was performed were included in the study.

2.3. Study variables and operational definitions

The study variables collected upon patient admission were socio-demographic, clinico-biological and scannographic data. The Wells score was calculated from demographic data (age, sex), cardiovascular history, cardinal symptoms and signs of PE and the existence or not of an alternative diagnosis [4]. Patients were classified according to the clinical probability level of the Wells score as follows: low (score < 1 points); intermediate (score 2-6 points) and high (score > 7 points) [4]. For the purposes of calculating sensitivity and specificity, the score was dichotomized into diagnosis of probable PE for a score > 4 or unlikely for a score of ≤ 4. The diagnosis of certainty/exclusion of PE was based on the presence or absence of direct and/or indirect signs of PE on CT angiography of the pulmonary arteries.

2.4. Statistical methods and data analysis

Patients were subdivided into two groups: confirmed PE versus excluded PE. The data were entered into Excel and analyzed using R software. The Chi² and Fisher Exact tests were used in univariate analysis to determine the categorical variables associated with the occurrence of PE. For quantitative variables, a ‘Student’ mean comparison test was used after checking the normality of the distributions. In the absence of normality of the variable, the Kruskal-Wallis test was used for a comparison of medians. The normality of the distribution of continuous variables was tested by graphical methods. A value of p < 0.05 defined the significance threshold for the association between the independent variables and the occurrence of pulmonary embolism. All variables whose univariate p-value was <0.2 were included in a multivariate logistic regression model to determine independent predictors of PE.

Considering CT angiography as the “gold standard” and the Wells score as a diagnostic test, true positives (TP) are patients who have a probable clinical score and a positive CT angiogram. False positives (FP) are patients who have a probable clinical score and a negative CT angiogram. False negatives (FN) are patients who have an improbable clinical score and a positive CT angiogram. True negatives (TN) are patients who have an improbable clinical score and a negative CT angiogram. Sensitivity (Se) and specificity (Sp) were calculated respectively from the following formulas: Se = TP/TP+FN; Sp=TN / TN+FP.

The ‘ROC’ curve (Receiver Operating Characteristics) is the graphical representation of all sensitivity-specificity pairs corresponding to all possible thresholds. It is constructed from the maximum possible score thresholds. It shows on the ordinate the sensitivity of the test, that is to say the proportion of true positives, and on the abscissa the specificity of the test, that is to say the proportion of true negatives.

The accuracy of the test will be defined by the area under the ROC curve (AUC) according to Sweets JA as follows [6]:

• ♣ If AUC= 0.5 then the diagnostic score will be of zero benefit.
• ♣ If AUC is between 0.5 and 0.70 then the diagnostic score will be of little information.
• ♣ If AUC is between 0.70 and 0.90 then the diagnostic score will be moderately informative.
• ♣ If AUC is between 0.90 and 1 then the diagnostic score will be very informative and perfect.

2.5. Ethical considerations

The study obtained approval from the institutional ethics committee of the hospital and anonymity was respected during data collection and processing.

3. Results

During the study period, 1228 patients were hospitalized in the two departments, including 999 patients in cardiology and 232 patients in the infectious and tropical diseases department. There were 472 patients hospitalized for suspected pulmonary embolism. This diagnosis was confirmed by chest CT angiography in 267 cases. The hospital prevalence of PE was 21.74%.

3.1. General characteristics of the study population

Male patients represented 47.88% (n=226), i.e. a sex ratio of 0.93. The average age of the patients was 54.33 years ± 16.82 with extremes of 19 years and 97 years. The modal class corresponded to the age group of [60-70] years with a number of 110 patients or 23.30% of patients (Figure 1).

Figure 1: Distribution by age group of 472 patients suspected of PE.

The main functional signs were dyspnea and chest pain, respectively in 84.32% (n=398) and 80.08M (n=378) (Reference Table 1). The average blood pressure is 125 mmHg; 95% confidence interval [80-231] for systolic and 79 mmHg; 95% confidence interval [47-133] for diastolic. The mean heart rate is 96 beats per min, 95% confidence interval [22-155] and spontaneous oxygen saturation is 96%, 95% confidence interval [44-100]. The proportions of physical signs found range from 1.69% (n=8) for a positive Homans sign at 44.49% for tachycardia (Reference Table 2).

Table 1: Simplified and original Wells score.

3.2. Univariate and multivariate analysis of clinical characteristics associated with the discovery of pulmonary embolism

In univariate analysis, the thromboembolic risk factors, functional and physical signs associated with the discovery of a PE are reported in Table 2. In the multivariate analysis by logistic regression, the variables significantly associated with the discovery of embolism lung on CT angiography were a history of venous thromboembolic diseases (p=0.0171), recent surgery (p=0.0005), dyspnea (p=0.0166), chest pain (p=0.0053), syncope (p=0.0010) (reference Table 3).

Table 2: Univariate analysis of thromboembolic risk factors, functional and physical signs of confirmed versus invalidated PE patients

3.3. Wells Score Performance Assessment Metrics

Contingency Table 4 presents the results of chest CT angiography according to the dichotomized Wells score (0-4 versus >4). The sensitivity and specificity of the Wells score at 2 probability levels were 78.65% and 92.19%, respectively. As for the positive predictive value (PPV) and the negative predictive value, they were 92.92% and 76.82%. Figure 2 shows the ROC curve of the Wells score in our study. The area under the ROC curve was estimated at 0.85, confidence interval at 95% [0.81-0.89].

Figure 2: ROC Curve of wells score in the diagnostic of PE.

Table 3: Multivariate analysis of clinical variables associated with the discovery of PE on the CT-scan.

Table 4: Contingency table of CT-scan results according to wells score.

4. Discussion

The objective of our study was to evaluate the diagnostic value of the Wells score in pulmonary embolism applied to a sub-Saharan African population. The prevalence of pulmonary embolism was.

The average age of the patients was 54.3 years ± 16.82 with extremes of 19 years and 97 years. The age group of [60-70] years represented the modal class with a number of 110 patients or 23.30%. Similar results were found by Dénakpo et al. [7] in Benin where the average age of patients was 52.31 years, and the age group of 60 years and over, was the largest at around 38.2%, and by Spencer et al. [8] in whom the majority of patients (55%) had an age greater than or equal to 65 years. Our results are nearly similar to those found by Pessinaba et al. [9] in Togo and Daher et al. [10] in Morocco concerning the average age with a divergence compared to the modal classes which can be explained by a means of selection previously quoted.

The female gender was predominant with 52.12% in our study. Most authors agree on the greater predisposition of women to develop VTE [11-13]. This female predominance could be explained by the existence of thromboembolic risk factors specific to this gender, notably pregnancy, peripartum, and cesarean section.

Thromboembolic risk factors in this study include not only pre-existing medical conditions, such as cancer or cardiovascular disease, but also local specific obesity (P<0.01), history of venous thromboembolism (P<0.01). 01), recent surgery (P=0.04) were significantly associated with CT discovery of pulmonary embolism. Samadoulougou et al. found that immobilization (p=0.05) and peripartum (p<0.02) were significantly associated with the CT discovery of pulmonary embolism [12]. Prolonged immobilization, often linked to limited movement or difficult living conditions, constitutes a significant risk factor for thromboembolism [4]. The prevalence of these factors in the local population may influence the validity of the Wells score.

In our study, the reasons for consultation were dominated by dyspnea (84.32%) and chest pain (80.0%). These clinical signs of PE, although the most frequent, remain less specific. They constituted the main reasons for consultations in the studies by Camara Y and Ndongo [14,15]. Tachycardia was the most frequent physical sign (44.49%) in our study followed by edema in the lower limbs as in the study by Chalal et al. [16]. Homans sign (P<0.01) and heart failure (P<0.01) were the clinical characteristics significantly associated with the CT finding of pulmonary embolism. However, these signs, which are also elements of the Wells score, may be less reliable in the event of a high prevalence of leg ulcers, often confused with symptoms of PE.

The prevalence of PE was higher in the “probable” category than in the “unlikely” category (92.92% vs 23.17%) in our study. Penazola and al. in a comparative study of the Wells score with the revised Geneva score to evaluate the pretest probability of pulmonary embolism noted (44.35% vs. 8.5%) [17].

The sensitivity of the Wells score was 78.65%. This sensitivity is lower than that of Gökhan et al. [18] (87.8%) who conducted a prospective study and superior to that of Posadas-Martinez et al. with a sensitivity of 65% [19]. This difference could be due to the not completely objective nature of the Wells score and the retrospective nature of our study. The sensitivity and specificity of a test are not constant, but variable depending on the characteristics of the population to which the test is applied [20].

In our study, the area under the curve was similar to that found by Penazola et al. [17] which is 0.85 (CI: 0.81 to 0.89) and close to that found by Yuxia et al. 0.83 [21]. Nevertheless, Klok et al. in 2008 in the Netherlands had 0.79(95% CI 0.72–0.87) noted 0.76, all lower than ours [22]. Despite the fluctuations in the area under the curve observed depending on the studies, this does not alter the informative value of the score which has an average informative value. In view of the results obtained, the diagnostic value of the Wells score in discriminating sick and non-sick is therefore acceptable. This test could therefore be used in hospitalization in all patients presenting signs suggesting a pulmonary embolism.

5. Conclusion

Pulmonary embolism, previously rare in sub-Saharan Africa, has become more and more frequent, putting the practitioner to the test in his daily practice. In our context, the geographic unavailability and high cost of pulmonary artery CT angiography constitute our main concern. The use of clinical probability scores constitutes the best alternative for an efficient diagnostic approach. The Wells score can be used as a diagnostic tool in hospitalized patients in a context of precarious technical support with some reservations. An adaptation of the score based on local data from prospective and multicenter studies could improve diagnostic accuracy in pulmonary embolism and, consequently, clinical results.

References

1. Philippot Q, Roche A, Goyard C, et al. Prise en charge de l’embolie pulmonaire grave en réanimation. Médecine Intensive Réanimation 27 (2018): 443-51.
2. Houénassi M, Véhounkpe JS, Tchabi Y, et al. Evaluation de la prise en charge de l’embolie pulmonaire dans un pays à faible niveau socio- économique: cas du Bénin (2013).
3. Traoré S, Sanre TM, Jean-Baptiste Tougouma S, et al. Coût de la prise en charge des pathologies cardio-vasculaires au Burkina Faso: Santé Publique 36 (2024): 127-36.
4. Wells PS, Anderson DR, Rodger M. Evaluation of D-Dimer in the diagnosis of suspected deep-vein thrombosis. ACC Current Journal Review 1 (2004): 15.
5. Ambid-Lacombe C, Cambou JP, Bataille V, et al. Excellentes performances du score de Wells et du score de Wells modifié dans le diagnostic de thrombose veineuse profonde proximale ou distale chez des patients hospitalisés ou ambulatoires au CHU de Toulouse: étude TVP-PREDICT. Journal des Maladies Vasculaires 34 (2009): 211-7.
6. Swets JA. Measuring the accuracy of diagnostic systems Science 240 (1988): 1285-93.
7. Dénakpo JL, Zoumènou E, Kérékou A, et al. Fréquence et facteurs de risque de la maladie thromboembolique veineuse chez la femme en milieu hospitalier à Cotonou, Benin. Clinics Mother Child H 9 (2012): 1-5.
8. Spencer FA, Gore JM, Lessard D, et al. Venous thromboembolism in the elderly. ThrombHaemost 100 (2008): 780-788.
9. Pessinaba S, Atti YDM, Baragou S, et al. L’embolie pulmonaire au centre hospitalier universitaire Campus de Lomé (Togo): étude rétrospective à propos de 51 cas. Pan Afr Med J 27 (2017): 129.
10. Daher B, Marc K, Bourkadi J. Valeur diagnostique du score de Genève révisé dans l’embolie pulmonaire. À propos de 91 cas (2018).
11. Diall IB, Coulibaly S, Minta I, et al. Etiologie, clinique et évolution de l’embolie pulmonaire a propos de 30 cas; MALI MEDICAL (2011).
12. Samadoulougou A, Millogo G, Seghda T, et al. Performance des scores cliniques de Génève révisé et de Wells dans le diagnostic de l’embolie pulmonaire au Centre hospitalier Universitaire Yalgado Ouédraogo (Burkina Faso) 67 (2015).
13. Seghda TAA, Yaméogo NV, Millogo GRC, et al. Prise en charge et pronostic des embolies pulmonaires associées à des thrombi intracavitaires droits: à propos d’une série prospective au centre Hospitalier Universitaire Yalgado Ouédraogo. Annales de Cardiologie et d’Angéiologie 68 (2019): 65-70.
14. Camara B, Diagne NR, Faye PM, et al. Critères de gravité et facteurs pronostiques du paludisme chez l’enfant à Dakar Médecine et maladies infectieuses 41 (2011): 63-7.
15. Sylvie NA, Amalia O, et al. Aspects Cliniques, Paracliniques, Étiologiques et Évolutifs de la Maladie Thromboembolique Veineuse dans Deux Services de Médicine Interne de Yaoundé. Health Sciences and Disease. 10 févr 23 (2022).
16. Chalal N, Demmouche A. Venous thromboembolic disease in the region of Sidi Bel Abbes, Algeria: frequency and risk factors]. Pan Afr Med J 16 (2013): 45.
17. Penaloza A, Melot C, Motte S. Comparison of the Wells score with the simplified revised Geneva score for assessing pretest probability of pulmonary embolism. Thrombosis Research 127 (2011): 81-4.
18. Çorapli G, Topçu AF. Combination of the Simplified Modified Geneva and Wells Clinical Prediction Scoring promise a good performance in pulmonary embolism diagnosis. Journal of Surgery and Medicine 5 (2021): 828-31.
19. Posadas-Martínez ML, Vázquez FJ, Giunta DH, et al. Performance of the Wells score in patients with suspected pulmonary embolism during hospitalization: A delayed-type cross sectional study in a community hospital. Thrombosis Research 133 (2014): 177-81.
20. Whiting P, Anne WS, Rutjesn A, et al. Sources of Variation and Bias in Studies of Diagnostic Accuracy. Ann Intern Med 140 (2004): 189-202.
21. Ma Y, Huang J, Wang Y, et al. Comparison of the Wells score with the revised Geneva score for assessing pretest probability of pulmonary embolism in hospitalized elderly patients. European Journal of Internal Medicine 36 (2016): e18-9.
22. Klok FA, Kruisman E, Spaan J, et al. Comparison of the revised Geneva score with the Wells rule for assessing clinical probability of pulmonary embolism. Journal of Thrombosis and Haemostasis 6 (2008): 40-4.