Differential value of risk factors and clinical signs for diagnosing pulmonary embolism according to age

Authors


Grégoire Le Gal, EA 3878 (GETBO), Department of Internal Medicine and Chest Diseases, Centre Hospitalier Universitaire de la Cavale Blanche, Boulevard Tanguy Prigent, 29609 Brest Cedex, France.
Tel.: +33 2 983 47336; fax: +33 2 983 47944; e-mail: gregoire.legal@chu-brest.fr

Abstract

Summary. Introduction: The diagnostic value of clinical presentation of pulmonary embolism (PE) is uncertain in the elderly, who often have concomitant cardiopulmonary diseases that may mimic PE. The aim of our study was to assess the differential value of risk factors, symptoms and clinical signs of venous thromboembolism, results of electrocardiogram and chest X-ray for the diagnosis of PE in suspected patients according to age. Methods: We analyzed data from two outcome studies which enrolled 1721 consecutive patients presenting in the emergency department with clinically suspected PE defined as acute onset of new or worsening shortness of breath or chest pain without any other obvious etiology. All patients underwent a sequential diagnostic work-up and a 3-month follow-up. Results: The proportion of confirmed PE was 24.2% (416 of 1721). Strength of the association with PE did not differ according to age group for history of venous thromboembolism (VTE), recent surgery, tachypnea at admission or right ventricular strain on electrocardiogram. Active malignancy, hemoptysis, tachycardia, hemidiaphragmatic elevation and pleural effusion at chest X-ray were no more associated with PE in the patients aged of 75 years or more. Finally, symptoms and signs of deep venous thrombosis, and an alternative diagnosis less probable than PE were associated with PE in all age groups, but the strength of this association decreased significantly with advancing age. Conclusion: Some risk factors, symptoms and signs of VTE are less strongly or even not at all associated with PE in the elderly. Physicians should take this into account when attending elderly patients suspected of PE and when assessing their clinical probability of PE.

Introduction

In spite of recent emergence of non-invasive tools [1], diagnosing pulmonary embolism (PE) remains a challenge. Current strategies rest on clinical probability assessment, plasma D-dimer measurement, helical computerized tomography (CT), ventilation-perfusion lung scan, and if necessary pulmonary angiography. However, PE is still an under-diagnosed disease and a common cause of death [2,3]. This may be particularly important in the elderly because (i) the incidence of venous thromboembolism (VTE) raises dramatically after 70 years [4]; (ii) PE is associated with a greater short-term mortality in the elderly [3]; (iii) the characteristics of some diagnostic tests for PE are different in this age group [5]. Indeed, although the sensitivity of D-dimer testing is 100% in all age groups, its specificity and hence its diagnostic yield decreases markedly with age. In a study of 1029 consecutive patients presenting to the emergency department with clinically suspected PE, the proportion of patients in which this diagnosis could be ruled out by a negative D-dimer test decreased from 67% in those under 40 years of age to 10% in those aged 80 or more [6]. The proportion of diagnostic lung scans (normal or high probability) also decreased from 68% to 42% with increasing age. Conversely, the proportion of patients in which the diagnosis of PE could be made by detecting a proximal deep vein thrombosis on compression ultrasonography (CUS) was higher in older patients: 61% in patients aged 70 years or more vs. 44% in younger patients. Age did not affect the diagnostic yield of helical computed tomography [7].

On the contrary, clinical symptoms and signs of PE have a poor sensitivity and specificity, and are not sensitive or specific enough to allow the physician to discard or to accept the diagnosis of PE without further testing. Nevertheless, clinical presentation is useful to evaluate the clinical probability of PE, whether implicitly or by prediction rules. Clinical probability assessment has become a mandatory step in the evaluation of a patient suspected of PE, allowing the identification of patients at lower risk of the disease who require a less extensive and/or invasive diagnostic workup. For example, PE can be ruled out by a negative D-dimer test only in patients with a non-high clinical probability of PE [8]. Clinical probability assessment is also important to interpret correctly the results of ventilation-perfusion lung scan and helical computed tomography [9]. Clinical presentation might be less discriminative in the elderly, because aged individuals frequently suffer from concomitant cardiopulmonary conditions that share symptoms and signs with PE. Moreover, the case-fatality rate of PE and the risk of bleeding on anticoagulant therapy are higher in older patients, underlining the importance of accurate diagnosis. It is therefore important to know the diagnostic value of risk factors, symptoms, and clinical signs of VTE for the diagnostic of PE in the elderly. Hence, the aim of our study was to assess the differential value of risk factors, symptoms, and clinical signs in patients with suspected PE according to age.

Material and methods

Patients

We analyzed the combined data from two prospective multicenter cohort studies, that included 1721 consecutive patients admitted to the emergency ward with suspected PE [1,10]. These outcome studies were designed to evaluate diagnostic strategies for PE, combining clinical probability assessment, plasma D-dimer measurement, lower limb venous CUS, and CT.

Briefly, all consecutive patients admitted to the emergency department of three general and teaching hospitals were included if they had a clinical suspicion of PE defined as acute onset of new or worsening shortness of breath or chest pain without any other obvious etiology. Written informed consent was obtained from all patients. The first study was conducted at the Geneva University Hospital, the Centre Hospitalier Universitaire Vaudois, Lausanne, both in Switzerland, and the Angers University Hospital in Angers, France, between October 2000 and June 2002. Of 1280 eligible patients, 258 were excluded because of the following predefined criteria: ongoing anticoagulant treatment for reasons other than VTE (n = 43); contraindication to CT scan: known allergy to iodine contrast agents or at risk of allergic reaction (n = 36), or creatinine clearance below 30 mL min−1 as calculated by the Cockroft formula (n = 53); informed consent impossible because of cognitive impairment (n = 24); patient refusal (n = 57); suspected massive PE with shock (n = 10); pregnancy (n = 9); estimated survival less than 3 months (n = 8); follow-up not possible (n = 11); and other reasons (n = 7). Another 67 patients were excluded because of protocol violations: diagnostic test not performed (D-dimer measurement: n = 1; ultrasonography: n = 21; helical CT: n = 17; pulmonary angiography: n = 17); and final diagnosis established by criteria different from those required by the study (n = 11). Thus, the final study population comprised 965 patients. The second was conducted at the Geneva University Hospital, Switzerland, the Angers University Hospital, and the Hôpital Européen Georges Pompidou, Paris, France, between September 2002 and October 2003. Of 1014 eligible patients, 185 were excluded because of ongoing anticoagulant treatment for reasons other than VTE (n = 38); a contraindication to CT scan: known allergy to iodine contrast agents or a risk of allergic reaction (n = 7), or creatinine clearance below 30 mL min−1 as calculated by the Cockcroft formula (n = 38), informed consent impossible because of cognitive impairment (n = 15); patient refusal (n = 41); suspected massive PE with shock (n = 3); pregnancy (n = 7); estimated survival <3 months (n = 9); follow-up not possible (n = 10); and other reasons (n = 17). Another 73 patients were excluded because of the following violations in study protocol: clinical probability was not assessed (n = 14); diagnostic tests were not performed, including D-dimer measurement (n = 4), ultrasonography (n = 15), CT (n = 8), and angiography (n = 7); tests were inadequately performed despite a negative D-dimer test (n = 2); a decision was made by the physicians in charge to prescribe anticoagulants, despite the absence of PE according to the study criteria (n = 5); or the diagnosis was not confirmed according to study criteria in patients whose CT scan was inconclusive for technical reasons (n = 17) or because of the presence of isolated subsegmental PE (n = 1). Thus, the final study population comprised 756 patients. Both studies were approved by the Ethics Committee of each participating institution. All participating hospitals are general public hospitals for the surrounding population and serve also as referral institutions for a larger population; however, the vast majority of included patients presented to the emergency ward primarily or via their GP.

Diagnostic work-up

All patients underwent a sequential diagnostic work-up. At the time of admission in the emergency department, the physicians in charge performed a clinical evaluation of the patient before any specific tests for suspected PE were performed. They filled out a standardized data form for each patient, recording demographic characteristics, risk factors, clinical signs and symptoms of VTE, results of electrocardiogram (EKG) and chest X-ray, presence of an alternative diagnosis to PE and its likelihood compared with that of PE. Based on that information, they assigned each patient into a clinical probability category using the Geneva prediction rule (Table 1) [11]. In case of disagreement between the clinical probability given by the Geneva prediction rule and their own clinical judgment, physicians were allowed to override score assessment by implicit judgment, which occurred in about 20% of patients [12]. Sequential tests were then performed, including plasma D-dimer measurement by an enzyme-linked immunosorbent assay (rapid ELISA assay, Vidas DD, BioMérieux, Marcy l'Etoile, France), except in high clinical probability patients for the second study. PE was ruled out by: (i) a D-dimer level below the cut-off value of 500 μg L−1; (ii) a negative lower limb venous CUS and a negative CT in patients with a low or intermediate clinical probability of PE; or (iii) in patients with a high clinical probability, a normal ventilation-perfusion lung scan or a normal pulmonary angiogram were furthermore required. PE was established by: (i) finding a proximal deep venous thrombosis (DVT) on CUS; (ii) a positive CT; or (iii) a high-probability ventilation-perfusion lung scan or a positive pulmonary angiogram. All patients had a 3 month follow-up. The risk of venous thromboembolic events during follow-up in patients fulfilling the criteria for absence of PE was below 2% in both studies.

Table 1.  The Geneva prediction rule
VariableScore
  1. PaCO2, partial pressure of carbon dioxide, arterial; PaO2, partial pressure of oxygen, arterial. From Wicki et al. [11].

Previous pulmonary embolism or deep vein thrombosis+2
Heart rate >100 beats per minute+1
Recent surgery+3
Age (years)
 60–79+1
 ≥80+2
PaCO2
 <4.8 kPa (36 mm Hg)+2
 4.8–5.19 kPa (36–38.9 mm Hg)+1
PaO2
 <6.5 kPa (48.7 mm Hg)+4
 6.5–7.99 kPa (48.7–59.9 mm Hg)+3
 8–9.49 kPa (60–71.2 mm Hg)+2
 9.5–10.99 kPa (71.3–82.4 mm Hg)+1
Chest radiograph
 Platelike atelectasis+1
 Elevated hemidiaphragm+1
Clinical probability
 Low0 to 4 points
 Intermediate5 to 8 points
 High9 points or more

Data analysis

We arbitrarily divided our sample into three age groups: <50 years (n = 520), 50–74 years (n = 693), and 75 years and more (n = 508). We studied all the clinical variables in our database known to be potentially associated with PE (demographic characteristics, known risk factors for VTE, comorbidities, symptoms and clinical signs). Their incidence in patients with PE (i.e. true-positive rate or sensitivity for PE) and in patients without PE (i.e. false-positive rate or one minus specificity) were compared by a chi-squared test in each age group. Interaction between age group and each criterion for the risk of PE was assessed by a Breslow-Day test [13] that tests the homogeneity of the odds-ratios associated with one criterion for the diagnosis of PE across age groups. We also estimated the positive likelihood ratio (PLR) of each risk factor and clinical sign for the diagnosis of PE and its 95% confidence interval in each age category. It is computed as sensitivity divided by 1-specificity. The posttest probability of PE can be computed for any clinical finding by Bayes’ theorem (posttest odds = pretest odds × positive likelihood ratio). Therefore, a PLR of one does not modify pretest probability. Conversely, the higher the PLR, the more likely PE becomes in presence of a positive finding (presence of a risk factor or clinical sign).

Results

Overall, the proportion of patients eventually diagnosed with PE in our study population was 24.2% (416 of 1721). Mean age was 60.5 (standard deviation 19.1), and 41.0% were male. The proportion of confirmed PE among suspected patients increased with age: 81 of the 520 patients (15.6%) aged <50 years, 168 of 693 (24.2%) aged 50 to 74 years and 167 of 508 (32.9%) aged 75 years or more had PE.

Incidence of risk factors for VTE (Table 2), symptoms and clinical signs (Table 3) and chest X-ray and electrocardiographic abnormalities (Table 4) were estimated in patients with and without PE according to age groups. Note that for each criterion, its presence in patients with PE corresponds to its sensitivity, and that in patients without PE corresponds to 1-specificity. Positive likelihood ratios (and their 95% confidence intervals) for the diagnosis of PE for each of these criteria are presented in Table 5.

Table 2.  Incidence of risk factors for VTE in patients with and without PE according to age group
CharacteristicsAll patients<50 years50–74 years≥75 yearsP-value*
No PEPEP-valueNo PEPEP-valueNo PEPEP-valueNo PEPEP-value
  1. *For homogeneity of odds-ratio across age groups.

  2. PE, pulmonary embolism; VTE, venous thromboembolism; COPD, chronic obstructive pulmonary disease.

No. of patients (%)1305 (75.8)416 (24.2) 439 (84.4)81 (15.6) 525 (75.8)168 (24.2) 341 (67.1)167 (32.9)  
Male gender  40.0 44.00.15 34.444.40.08 46.1 51.20.25 37.8 36.50.780.30
Personal history of VTE14.129.9<0.019.818.50.0212.824.4<0.0121.841.0<0.010.87
Familial history of VTE10.317.3<0.0114.019.80.189.919.1<0.016.014.4<0.010.45
Surgery4.58.4<0.015.712.40.034.68.30.062.96.60.050.90
Active malignancy8.513.00.013.66.20.2910.920.2<0.0110.99.00.500.05
COPD11.07.20.032.54.90.2314.97.10.0115.88.40.020.05
Congestive heart failure9.47.90.380.91.20.789.04.20.0420.815.00.110.53
Stroke with paralysis1.92.60.371.41.20.921.33.00.163.53.00.760.43
Varicose veins18.725.9<0.018.111.10.3620.421.40.7630.137.60.090.62
Table 3.  Incidence of symptoms and clinical signs of VTE in patients with and without PE according to age group
CharacteristicsAll patients<50 years50–74 years≥75 yearsP-value*
No PEPEP-valueNo PEPEP-valueNo PEPEP-valueNo PEPEP-value
  1. *For homogeneity of odds-ratio across age groups.

  2. PE, pulmonary embolism; VTE, venous thromboembolism; COPD, chronic obstructive pulmonary disease.

No. of patients (%)1305 (75.8)416 (24.2) 439 (84.4)81 (15.6) 525 (75.8)168 (24.2) 341 (67.1)167 (32.9)  
Chest pain  70.4 57.5<0.01 84.377.80.15 72.0 58.9<0.01 50.2 46.10.390.28
Dyspnea64.579.8<0.01 55.171.60.0164.081.6<0.0177.481.90.240.15
Lower limb pain14.335.3<0.01 12.135.8<0.0113.741.1<0.0118.229.3<0.010.01
Hemoptysis4.16.30.07  3.612.4<0.014.64.80.924.14.80.720.05
Syncope14.416.60.26 13.09.90.4414.317.90.2616.218.60.510.45
Lower limb pain at palpation and/or edema9.330.1<0.01  4.624.7<0.0110.334.7<0.0113.828.1<0.010.02
Neck vein distension7.313.3<0.01  1.42.50.466.110.80.0416.721.00.240.65
Tachypnea >20/min46.864.7<0.01 35.554.3<0.0145.361.3<0.0163.673.10.030.56
Tachycardia>100 bpm23.337.0<0.01 21.648.2<0.0123.435.7<0.0125.232.90.070.03
Table 4.  Incidence of radiographic and electrocardiographic signs of PE in patients with and without PE according to age group
CharacteristicsAll patients<50 years50–74 years≥75 yearsP-value*
No PEPEP-valueNo PEPEP-valueNo PEPEP-valueNo PEPEP-value
  1. *For homogeneity of odds-ratio across age groups.

  2. PE, pulmonary embolism; VTE, venous thromboembolism; COPD, chronic obstructive pulmonary disease.

No. of patients (%)1305 (75.8)416 (24.2) 439 (84.4)81 (15.6) 525 (75.8)168 (24.2) 341 (67.1)167 (32.9)  
Pleural effusion  14.1 19.50.01  8.516.10.04 13.6 25.00.01 22.115.80.10<0.01
Atelectasis5.915.4<0.013.613.6<0.01  6.0 17.6<0.01 8.814.00.070.11
Hemidiaphragmatic elevation10.421.2<0.017.829.6<0.0110.2 20.0<0.0113.818.20.20<0.01
Right ventricular strain (EKG)16.230.5<0.0113.626.00.0116.0 29.1<0.0119.734.1<0.010.99
Alternative diagnosis more likely than PE78.927.6<0.0179.233.8<0.0181.6 22.1<0.0174.430.2<0.010.01
Table 5.  Positive likelihood ratios of general characteristics, risk factors, symptoms and clinical signs of VTE, radiographic and electrocardiographic signs of PE, across age groups
 <50 years50–74 years>75 years
  1. *Negative likelihood ratio.

General characteristics, risk factors
 Male gender1.3 (1.0–1.7)1.1 (0.9–1.3)1.0 (0.8–1.3)
 Personal history of VTE1.9 (1.1–3.2)1.9 (1.4–2.7)1.9 (1.4–2.5)
 Familial history of VTE1.4 (0.9–2.3)1.9 (1.3–2.9)2.4 (1.4–4.2)
 Surgery2.2 (1.1–4.3)1.8 (1.0–3.4)2.2 (1.0–5.2)
 Active malignancy1.7 (0.6–4.5)1.9 (1.3–2.7)0.8 (0.5–1.5)
 COPD2.0 (0.6–6.0)0.5 (0.3–0.9)0.5 (0.3–0.9)
 Congestive heart failure1.4 (0.2–12.0)0.5 (0.2–1.0)0.7 (0.5–1.1)
 Stroke with paralysis0.9 (0.1–7.4)2.2 (0.7–6.9)0.9 (0.3–2.4)
 Varicose veins1.4 (0.7–2.8)1.1 (0.8–1.5)1.3 (1.0–1.6)
Symptoms and clinical signs
 Chest pain0.9 (0.8–1.0)0.8 (0.7–0.9)1.0 (0.8–1.2)
 Dyspnea1.3 (1.1–1.5)1.3 (1.2–1.4)1.1 (1.0–1.2)
 Lower limb pain3.0 (2.0–4.4)3.0 (2.3–4.0)1.6 (1.2–2.2)
 Hemoptysis3.4 (1.6–7.2)1.0 (0.5–2.3)1.2 (0.5–2.7)
 Syncope0.8 (0.4–1.5)1.3 (0.9–1.8)1.1 (0.8–1.7)
 Lower limb pain at palpation and/or edema5.4 (3.1–9.6)3.4 (2.4–4.7)2.0 (1.4–2.9)
 Neck vein distension1.8 (0.4–8.8)1.8 (1.0–3.1)1.3 (0.9–1.8)
 Tachypnea >20/min1.5 (1.2–1.9)1.4 (1.2–1.6)1.1 (1.0–1.3)
 Tachycardia>100 bpm2.2 (1.7–3.0)1.5 (1.2–2.0)1.3 (1.0–1.7)
Chest X-ray, EKG, likelihood of alternative diagnosis
 Pleural effusion1.9 (1.0–3.4)1.8 (1.3–2.6)0.7 (0.5–1.1)
 Atelectasis3.8 (1.8–8.0)3.0 (1.8–4.8)1.6 (1.0–2.6)
 Hemi-diaphragmatic elevation3.8 (2.4–6.1)2.0 (1.3–2.9)1.3 (0.9–2.0)
 Right ventricular strain (EKG)1.9 (1.2–2.9)1.8 (1.3–2.5)1.7 (1.3–2.3)
 Alternative diagnosis more likely than PE*3.2 (2.5–4.1)*4.2 (3.5–5.2)*2.7 (2.2–3.4)*

Potential risk factors and clinical signs tested could be schematically divided into four groups. First, no strongly significant association was found in any age group between PE and gender, history of stroke with paralysis, congestive heart failure, varicose veins, syncope and neck vein distension. Second, other variables were associated with PE across all age groups, and the strength of their association with PE did not differ significantly according to age group: personal and familial history of VTE, recent surgery, tachypnea at admission or right ventricular strain on EKG. For example a personal history of VTE was found in 18.5% of younger patients and in 41.0% of older patients with PE, but with an almost constant PLR of 1.9 across age groups (Table 5). More interestingly, significance and/or direction of the association between PE and a history of chronic obstructive pulmonary disease (COPD), active malignancy, presence of hemoptysis, tachycardia, hemidiaphragmatic elevation and pleural effusion at chest X-ray varied significantly with age. Although the test for interaction with age was not highly significant for these criteria (P = 0.05), hemoptysis was strongly and positively associated with PE in patients of <50 years (PLR 3.4, 95% CI 1.6–7.2) but not above this age (PLR 1.0); below 75 years, about twice as many patients with PE had active malignancy as did patients without PE, whereas the incidence of active malignancy was not different in patients with and without PE of 75 years or older; finally, COPD was not significantly associated with PE before the age of 50 years, whereas it was negatively associated with this diagnosis after this age. Presence of tachycardia, pleural effusion and hemidiaphragmatic elevation at chest X-ray were associated with PE up to the age of 75 but not in patients older than 75 years (P-values for interaction with age 0.03, <0.01 and <0.01, respectively). Finally, symptoms of DVT (lower limb pain reported by the patient), signs of DVT on clinical examination (provoked pain at lower limb palpation and/or edema) and existence of an alternative diagnosis judged more likely than PE were strongly and significantly associated with PE in all age groups, but the strength of this association decreased significantly with advancing age (P-values 0.01, 0.02 and 0.01, respectively). As a consequence, PLRs computed for these criteria for the diagnosis of PE decreased across age groups as shown in Table 5. As an example, if clinical signs of DVT are present, the odds of PE should be multiplied by 5.4 (95% CI 3.1–9.6) in patients aged 50 years or less, by 3.4 (95% CI 2.4–4.7) in patients aged 50–74 years, and only by 2.0 (95% CI 1.4–2.9) in patients aged 75 years or above.

Discussion

This study shows that the predictive value of several risk factors, symptoms, clinical, radiographic and electrocardiographic signs of PE varies significantly with age. The strength of the association decreased significantly with age for symptoms and signs of DVT and an alternative diagnosis more likely than PE. In contrast, several variables were no longer associated with PE in the elderly: active malignancy, hemoptysis, tachycardia and chest X-ray anomalies. Finally, a personal or family history of VTE, recent surgery, tachypnea at admission or right ventricular strain on EKG were predictive of PE in all age groups.

Patients included in our series presented with a clinical suspicion of PE, defined as acute onset of new or worsening shortness of breath or chest pain without any other obvious etiology. Thus, our results only apply to patients with a similar clinical pattern, and not to all patients presenting to emergency wards with dyspnea and/or chest pain. However, the 24% proportion of confirmed PE among suspected patients in this study is in agreement with present referral patterns in Europe. Moreover, the two studies that were merged in the present analysis took place in emergency departments of general and teaching hospitals, and all consecutive patients presenting to the emergency department fulfilling inclusion criteria were included. Thus, we believe that our sample is not selected and that our results may be generalized to all outpatients referred to an emergency department with clinically suspected PE. Another limitation of our analysis is that in the first study, finding a proximal DVT on CUS allowed to rule in the diagnosis of PE without further test. Thus, the diagnosis of PE was made on the basis of a positive CUS without thoracic examination in 86 of the 222 patients considered as having PE. It is possible that not all patients with proven DVT had concomitant PE, particularly in the elderly, and that their dyspnea and/or chest pain was in fact because of chronic cardio-respiratory conditions. This could have biased our results on the value of symptoms and clinical signs of PE. However, we believe that our inclusion criteria, requiring acute onset of new or worsening thoracic symptoms, avoided inclusion of a substantial proportion of patients with DVT but without PE. As a matter of fact, in the second study, all patients with either a positive D-dimer test and a non-high clinical probability of PE, or a high clinical probability of PE underwent both CUS and multi-detector CT scan: 96% of patients with DVT on CUS had PE on CT (73 of 76). Finally, because of the number of statistical tests performed, we cannot exclude that some of the significant associations found occurred by chance. However, this should not limit our main conclusions that are sustained by strongly significant tests.

Some authors focused on the clinical presentation of PE in the elderly [14], or compared clinical presentation of PE between younger and older patients [15]. In a non-consecutive, retrospective series of 196 patients with suspected PE, Donnamaria et al. [16] reported no differences in the frequency of risk factors and clinical signs of PE between younger and older subjects, except for pleuritic chest pain that was more frequent in younger patients. Conversely, cardiac shadow enlargement at chest X-ray and hypoxemia were more frequent in patients suspected of PE aged 60 years or above. However, data were not analyzed according to whether the diagnosis of PE was confirmed or ruled out. In a retrospective study on 192 patients with and without PE, risk factors and clinical signs of PE were included in two multivariate analyses in patients of 65 years old or less and in patients above this age. Cough was found to be associated with PE only in patients over 65 years, whereas a significant association between a previous trauma and PE existed only in younger patients [17]. However, the interaction between age and these clinical parameters was not formally tested.

In our series, clinical presentation of PE differed according to age, as previously reported, but age also influenced the diagnostic value of those signs for PE. Why did some risk factors, clinical, radiographic and electrographic signs of PE perform less well in discriminating PE in older subjects? Higher incidence of comorbidities in the elderly might be the main explication for our results. Indeed, chronic heart failure, lung or lower respiratory tract infectious diseases are more prevalent among older patients with dyspnea and/or chest pain and may account for clinical and radiological signs also evocative of PE. This might also explain the lack of predictive accuracy of hemoptysis, tachycardia, or symptoms and signs of DVT. Thus, even if those criteria are present in elderly patients with dyspnea or chest pain, physicians should consider alternative diagnoses as carefully as that of PE. The proportion of confirmed cases of PE among suspected patients was higher in the oldest group (one of three patients in the oldest group compared with one of seven in the youngest group). This might suggest that physicians were more specific in diagnosing PE clinically in the elderly than in younger people. However, this does not seem very likely because of the higher incidence of conditions potentially masquerading PE in the elderly. Alternatively, it could only reflect the higher incidence of PE in the elderly as age is an important and independent risk factor for VTE. The only variables predictive of PE across all age groups were major risk factors for VTE (surgery, history of VTE), and right ventricular strain on EKG. Hence, PE should be more strongly considered when such findings are present in elderly patients with dyspnea or chest pain. Conversely, in our cohort a history of COPD made PE less likely in the elderly. Although COPD is classically thought to be a risk factor of PE, controversies exist about this association. Our data might argue against it, at least in the elderly.

We recently reported about the use of clinical prediction rules in the elderly [18]. Although both rules that we analyzed – the Geneva score with possible override [12] and the Wells’ score [19] – were able to identify three groups of patients with increasing incidence of PE in all age groups, the Wells’ score seemed to perform less well in the elderly. Indeed, the incidence of PE in patients classified by the Wells’ score as having a low clinical probability was higher in those aged 75 years or older than below 50 years of age (15% vs. 7%), and that observed in patients classified as having a high clinical probability was lower (56% vs. 100%). This was not true for the Geneva score, where patients aged 75 years or older had a incidence of 6% in the low clinical probability group, and of 92% in the high clinical probability group. Interestingly, five of the seven criteria of the Wells’ score have in our study a decreased capacity to discriminate PE in the elderly, including the likelihood of an alternative diagnosis and signs of DVT that have an important weight in this score. This is also true for some of the clinical criteria included in the Geneva score but might be counterbalanced by the points that are attributed in this score to the age – a major risk factor of VTE – and to the blood gases arterial analysis results, that noteworthy are not clinical items but require invasive arterial puncture (Table 1). This might explain the lower proportion of older patients classified as having low clinical probability by this score (37% of patients after 75 years vs. 72% before 50).

In summary, although a personal or family history of VTE, a recent surgery, a tachypnea at admission or a right ventricular strain on EKG are predictive of PE in all age groups, several known risk factors, clinical and radiographic symptoms and signs of VTE are significantly less strongly or even no longer associated with PE in elderly individuals. In patients aged 75 years or more, symptoms and signs of DVT, and an alternative diagnosis more likely than PE are less predictive of PE than in younger patients. Moreover, tachycardia, pleural effusion and hemidiaphragmatic elevation at chest X-ray are no longer predictive of PE in patients aged 75 years or more. Thus, even if those criteria are present in elderly patients with dyspnea or chest pain, physicians should consider alternative diagnoses as carefully as that of PE. Furthermore, this might alter assessment by physician of pretest clinical probability of PE – implicitly or by means of prediction rules – and thus lead to inappropriate diagnostic strategies and missed diagnoses in these patients. As a consequence, the development of specific clinical prediction rules, or the validation of existing rules specifically in elderly people should be encouraged.

Acknowledgements

The two studies that were used in this work were approved by the Ethics Committee of each participating institution. All patients gave written informed consent. These studies were supported by a grant from the Hirsch Fund of the University of Geneva, a grant (32-61773.00) from the Swiss National Research Foundation; grants 97/4-T10 and 00/4-T9 from the Royal College of Physicians and Surgeons, Canada; a grant from La Fondation Québécoise pour le Progrès de la Médecine Interne and Les Internistes et Rhumatologues Associés de l'Hôpital du Sacré-Cœur, Montreal, Canada; and grant 2001/021 from the Direction of Clinical Research of the Angers University Hospital. The funding sources did not have any role in designing the study, collecting, analysing or interpreting data, and were not associated either in writing the report or in the decision to submit the paper for publication.

Ancillary