methods in patients with pulmonary thromboembolism
Gülfer OKUMUŞ1, Rabia ENGİN ÜNVER1, Esen KIYAN1, Levent TABAK1, Halim İŞSEVER2, Orhan ARSEVEN1
1 İstanbul Üniversitesi İstanbul Tıp Fakültesi, Göğüs Hastalıkları Anabilim Dalı,
2İstanbul Üniversitesi İstanbul Tıp Fakültesi, Halk Sağlığı Anabilim Dalı, İstanbul.
ÖZET
Pulmoner tromboemboli tanısı alan olgularda üç farklı klinik skorlama yönteminin karşılaştırılması
Pulmoner tromboemboli (PTE)’de en sık görülen semptomların nonspesifik olması nedeniyle klinik tanı güvenilir değildir. Kli- nik olasılık değerlendirmesi için günümüzde değişik skorlama yöntemleri mevcuttur. Bu çalışmanın amacı; PTE’si kanıtlan- mış olgularda Wells, Wicki ve modifiye Hyer skorlamalarına göre elde edilen klinik olasılıkları karşılaştırarak aralarındaki ko- relasyonu araştırmaktır. Bu amaçla kliniğimizde PTE tanısı alan olguların klinik olasılıkları Wells, Wicki ve modifiye Hyer yön- temleri ile ayrı ayrı belirlendi. Üç yöntemin uyum katsayıları; cramer, kontenjan katsayısı (CC), Kendal’ın tau-b ve kappa test- leri ile, sıra katsayıları ise Spearman’ın rho testi ile hesaplandı. Çalışmaya alınan 248 olgunun 119’ u erkek, 129’u kadın ve ortalama yaşları 57 ± 16.7 idi. Modifiye Hyer yöntemine göre olguların %61.3’ü yüksek klinik olasılık olarak tanımlanırken, Wicki’ye göre %56, Wells’e göre %50 olgu orta klinik olasılıklı olarak tanımlanmıştı. Çalışmamızda klinik olasılığı skorlayan üç yöntem karşılaştırıldı. Bu yöntemler arasındaki korelasyon; kappa uyum katsayısına göre zayıf, diğer katsayılara göre is- tatistiksel olarak orta güçte bulundu. Gruplar arasında en güçlü korelasyon modifiye Hyer yöntemi ile Wells yöntemi arasın- da bulunmuştur.
Anahtar Kelimeler: Klinik olasılık, pulmoner tromboemboli, skorlama.
Yazışma Adresi (Address for Correspondence):
Dr. Gülfer OKUMUŞ, İstanbul Üniversitesi İstanbul Tıp Fakültesi, Göğüs Hastalıkları Anabilim Dalı, 34390 Çapa/Fatih, İSTANBUL - TURKEY
e-mail: gulferokumus@yahoo.com
Patients presenting with suspected pulmonary thromboembolism (PTE) are frequent in both in- patient and outpatient settings. Although risk factors for the development of PTE and presen- ting signs and symptoms are well known, none of these are specific enough to make a definite diagnosis of the disease. Clinical presentation varies largely because it depends on severity, lo- calization of thrombi, age and underlying cardi- opulmonary function (1,2).
The pre-test probability (PTP) of disease repre- sents a formal assessment of the likelihood of disease before a confirmatory test or investiga- tion is performed. Assigning a PTP of disease when diagnosing PTE can be used in combinati- on with objective diagnostic imaging including ventilation/perfusion (V/Q) lung scans, deep ve- nous ultrasonography and spiral computerized tomography (CT) angiography.
Clinical prediction rules are used to aid clinici- ans in assigning a PTP to patients with clinical signs and symptoms of PTE. A few clinical sco-
res for PTE have been developed which utilize a combination of information from history, physi- cal examination and laboratory tests to catego- rize patients into low, intermediate and high cli- nical probability of PTE (3-5). One of these cli- nical scores is an algorithm published by Hyers in 1995 (Table 1)(3). This score is based on risk factors and radiographic or gas exchange ab- normalities. In 1998, Wells et al. developed a cli- nical score derived from a multivariate analysis of risk factors and clinical signs associated with the diagnosis of PTE in a large database (4). La- ter, this score was subsequently simplified to a limited number of signs, symptoms and risk fac- tors to produce the Wells score (Canadian sco- re) (Table 2)(5). In 2001, Wicki et al. developed a similar clinical scoring model (Geneva score) but this score additionally included room air ar- terial blood gas measurement (Table 3) (6).
Wells and Wicki scores were compared with implicit judgment and demonstrated similar ac- curacy (7). However, these scores have some li- mitations and they need to be complemented SUMMARY
Comparison of three clinical scoring methods in patients with pulmonary thromboembolism
Gülfer OKUMUŞ1, Rabia ENGİN ÜNVER1, Esen KIYAN1, Levent TABAK1, Halim İŞSEVER2, Orhan ARSEVEN1
1 Department of Chest Diseases, Istanbul Faculty of Medicine, Istanbul University, Istanbul, Turkey,
2 Department of Public Health, Istanbul Faculty of Medicine, Istanbul University, Istanbul, Turkey.
The clinical diagnosis is not reliable in pulmonary thromboembolism (PTE) because the symptoms are mostly nonspecific.
Different clinical prediction rules for PTE have been described recently. These rules are used to aid clinicians in assigning a pre-test probability to patients with clinical signs and symptoms of disease. The aim of this study was to assess the clini- cal probability of PTE using three different models (Wells, Wicki and modified Hyers scores) and to find their power to de- termine PTE probability. Clinical probabilities of patients with PTE were determined with Wells, Wicki and modified Hyers scores. Cramer’s, contingency coefficient (CC), Spearman’s rho, Kendal’s tau-b and kappa tests were used for statistical analysis. The study included 248 patients (119 male, 129 female; mean age= 57 ± 16.7 years). Although 61.3% of the pati- ents were assigned high clinical probability with modified Hyers score, 56% and 50% of them were assigned moderate cli- nical probability with Wicki and Wells scores, respectively. The correlations of the clinical probabilities was low according to kappa correlation test and moderate according to the Cramer’s, CC, Spearman’s rho and Kendal’s tau-b tests among the- se three scores. In conclusion, modified Hyers score showed higher percentage of high probability compared to Wells and Wicki scoring systems and the best correlation was found between the modified Hyers and Wells scores.
Key Words: Pre-test probability, pulmonary thromboembolism, scoring.
with evaluation of the clinical findings, arterial blood gas analysis, radiologic evaluation and the clinician’s judgment of whether an alternative di- agnosis is more likely than a PTE diagnosis. Al- so it’s hard to remember scoring parameters of Wells and Wicki scores.
The aim of this study was to assess the clinical probability of PTE using 3 different scores (Wells score, Wicki score and modified Hyers score) in order to find their power to determine PTE pro- bability.
MATERIALS and METHODS
We have retrospectively investigated medical re- cords of 385 patients with PTE and 248 of them with objectively documented PTE were included in the study. Of 248 patients, 149 had high pro- bability V/Q scan and 99 had thrombus in spiral CT angiography. Of them, 110 had also acute deep ven thrombosis (DVT) documented by Doppler ultrasonography or venography. There were 119 males and 129 females, and the mean age was 57 ± 16.7 years, ranging from 16 to 91 years. For each patient, medical records were re- viewed for acquired risk factors of PTE (such as recent surgery, immobilization, neoplasia, etc).
Clinical probability of each patient was retros- pectively determined by the first author accor- ding to the Wells, Wicki and modified Hyers sco- res (Hyers score without arterial blood gas analy- sis and radiographic findings) (Table 4) (5,6).
Statistical Analysis:
Cramer’s, contingency coefficient (CC), Spear- men’s rho, Kendal’s tau-b and kappa test were used to compare the likelihood of each PTP sco- re to accurately predict the presence of PTE and Table 1. Criteria for the calculation of the Hyers score.
Low probability • Risk factor not present
• Dispnea, tachipnea or pleuritic pain may be present but explainable by another condition
• Radiographic or gas exchange abnormality may be present but explainable by another condition
Intermediate probability • Neither high or low clinical probability High probability • Risk factor present
• Otherwise unexplained dyspnea, tachipnea or pleuritic pain
• Otherwise unexplained radiographic or gas exchange abnormality
Table 2. Criteria for the calculation of the Wells score.
Criterion Points
Clinical signs of DVT 3
Alternative diagnosis less probable than PTE 3
Heart rate > 100 bpm 1.5
Immobilization or surgery < 4 weeks ago 1.5
Previous DVT or PE 1.5
Haemoptysis 1
Cancer 1
< 2 Low probability, 2-6 Intermediate probability,
> 6 High probability.
Table 3. Criteria for the calculation of Wicki score.
Criterion Points
History of DVT or PTE 2
Heart rate > 100/min 1
History of surgical operation 3 Age
60-79 1
≥ 80 2
PaCO2 (mmHg)
< 36 2
36-39 1
PaO2 (mmHg)
< 49 4
49-60 3
60.1-71.3 2
71.4-82.7 1
Atelectasis 1
Elevation of the diaphragma 1
0-4 Low probability, 5-8 Intermediate probability,
≥ 9 High probability.
95% confidence intervals (95% CI) were calcula- ted employing the normal approximation. p<
0.05 was accepted significant.
RESULTS
Of 248 patients with objectively documented PTE, 81% had acquired risk factors, including age (n= 201), recent surgery (n= 65), immobili- zation (n= 56), obesity (n= 51), neoplasia (n=
32), heart failure (n= 22), trauma (n= 17), preg- nancy (n= 10), travel (n= 9), use of oral contra- ceptive (n= 8) and others (n= 5; 2 nephrotic syndromes and 3 strokes) (Figure 1).
According to modified Hyers score, 61.3% of the patients had high probability, 29% moderate
probability, and 9.7% low probability. These pro- babilities were 17%, 56%, 27% for Wicki score and 28.2%, 50%, 21.8% for Wells score respecti- vely (Table 5). Among 151 patients with high probability according to modified Hyers, 40 pre- sented by high probability, 86 by moderate, and 25 by low according to Wicki. These values we- re 63 for high probability, 82 for moderate, and 6 for low probability in the case of Wells.
According to kappa correlation test, there was a weak correlation between 3 scoring models (Hyers vs Wicki 0.19; Hyers vs Wells 0.27; Wick vs Wells 0.3) (Table 6). The Cramer’s test, CC, Spearman’s rho test and Kendal’s tau-b test we-
Immobilization
Risk factors
201 65
56 51 32 22 17 10 9 8 5
0 50 100 150 200 250
Age Surgery Obesity Neoplasia Heart failure Trauma Pregnancy Travel Oral contr.
Other
Figure 1. Risk factors of patients with PTE.
Table 4. Criteria for the calculation modified Hyers score.
No risk factors, non-typical PTE findings Low probability
a. No risk factors, massive-submassive PTE findings Moderate probability b. One or more risk factors, non-typical PTE findings
Massive-submassive PTE findings or a few risk factors High probability Non-typical PTE findings;
Fever, confusion, wheezing, resistant left ventricular insufficiency, cardiac tachyaritmia, progressive right ventricular insufficiency, sub acute dyspnea.
re showed moderate correlation between scoring models. The best correlation was found between the modified Hyers and Wells scoring models (0.55, p< 0.01)
DISCUSSION
The evaluation of suspected PTE is difficult due to the low specificity of the presenting signs and symptoms. Physicians must obtain imaging stu- dies and interpret them with the underlying PTP.
A study showed that only 37% of the physicians recorded their PTP in the chart (7). This result points out that there is a need for rapid and ob- jective PTP.
Moores et al., Ulukavak Ciftci et al. and Boyacı et al. showed that Wells score is practical and easy to apply to all patients with suspected PTE (8-10). But Wells score includes the clinician’s judgment of whether an alternative diagnosis is more likely than a PTE diagnosis. This criterion carries a major weight in this score and can ob- viously not be standardized.
Iles et al. compared Wells, Wicki and empirical assessment in patients with suspected PTE.
They found that Wicki score is the most consis- tent method of determining PTP and is not affec- ted by clinical experience (11). However, Wicki score requires arterial blood gas values at room air, a variable that was not available in 15% of the patients (6). Therefore Le Gal et al. revised the Wicki score and they developed a new pre- diction score (12).
In addition to Wells and Wicki scores we also evaluated modified Hyers scoring system in this study. Although Hyers score is not a validated algorithm for estimating the clinical probability of PTE, it has been advised in guidelines for es- tablishing clinical probability. According to this score, a high probability patient (80-100%) wo- uld have a plausible risk factor, one or more of the common screening findings, and a radiog- raphic or gas exchange abnormality. A low pro- bability patient (1-19%) would have no risk fac- Table 6. The correlation between three clinical scores.
Scoring
Wicki
Spearman’s Cramer’s V CC Kendall’s tau b Kappa
95% CI 95% CI 95% CI 95% CI 95% CI
Hyers (modf) 0.38* 0.28* 0.37* 0.36* 0.19*
0.26-0.51 0.01-0.33 0.25-0.50 0.19-0.89 0.13-0.26
Wells
Spearman’s Cramer’s V CC Kendall’s tau b Kappa
95% CI 95% CI 95% CI 95% CI 95% CI
Hyers (modf) 0.55* 0.41* 0.50* 0.50* 0.27*
0.43-0.68 0.01-0.53 0.38-0.63 0.14-0.89 0.18-0.37
Wicki 0.37* 0.28* 0.37* 0.33* 0.30*
0.25-0.50 0.01-0.32 0.25-0.50 0.09-0.67 0.22-0.38
* p< 0.01
Table 5. Clinical probabilities of patients according to Hyers, Wicki and Wells scores.
Low probability Moderate probability High probability
Wicki 27% 56% 17%
Wells 21.8% 50% 28.2%
Modified Hyers 9.7% 29% 61.3%
tors, clinical symptoms, or findings that were explainable by another disease, and radiograp- hic or gas exchange abnormalities that were al- so explainable by another condition. Intermedi- ate category patients (20-79% probability) are those who do not meet criteria for either the high or low category. Hyers has advised this algo- rithm or some reasonable variant for PTP (3).
This algorithm is based on radiographic and ar- terial blood gas values.
Our country is a developing country. For PTE there is not enough diagnostic equipments (e.g.
CT, V/Q scan, Doppler ultrasonography) at every part of the country (except university hos- pitals and some private hospitals in big cities) (13). Furthermore, it is not possible to have ar- terial blood gas analysis and chest X-ray in most of health care units. Therefore, the algorithm published by Hyers was modified according to the conditions of our country. Modified Hyers score showed higher percentage of high probabi- lity compared to Wells and Wicki scoring sys- tems (p< 0.05). This modified algorithm is simp- le, easy to remember and easy to use without a lot of high technology. It also does not include ar- terial blood gases values and chest X-Ray.
The characteristic of this study is that only pati- ents with objectively documented PTE were eva- luated. On the other hand, the study has some limitations:
1. Medical files/records of the patients with PTE were investigated retrospectively by an author, 2. Only inpatients were evaluated,
3. There is no control group. Also modified Hyers score is not validated.
In conclusion; modified Hyers algorithm was fo- und to be more meaningful compared to other scoring systems and the best correlation was fo- und between the modified Hyers and Wells algo- rithms in our study. We have thought that a new prospective study should be done for validation of modified Hyers scores.
Conflict of Interest Statement
None of the authors have a conflict of interest to declare in relation to this work.
REFERENCES
1. Stein PD, Saltzman HA, Weg JG. Clinical characteristics of patients with acute pulmonary embolism. Am J Car- diol 1991; 68: 1723.
2. Stein PD, Terrin ML, Hales CA. Clinical, laboratory, roent- genographic and electrocardiographic findings in pati- ents with acute pulmonary embolism and no pre-exis- ting cardiac or pulmonary disease. Chest 1991; 100: 598.
3. Hyers TM. Diagnosis of pulmonary embolism. Thorax 1995; 50: 930-2.
4. Wells PS, Gingsberg JS, Anderson DR, et al. Use of a cli- nical model for safe management of patients with sus- pected pulmonary embolism. Ann Intern Med 1998; 129:
997-1005.
5. Wells PS, Gingsberg JS, Anderson DR, 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.
6. Wicki J, Pergener TV, Jumod AF, et al. Assessing clinical probability of pulmonary embolism in the emergency ward: A simple score. Arch Intern Med 2001; 161: 92-7.
7. Chagnon I, Bounameaux H, Aujesky D, et al. Compari- son of two clinical prediction rules and implicit assess- ment among patients with suspected pulmonary embo- lism. Am J Med 2002; 113: 269-75.
8. Moores LK, Collen JF, Woods KM, Shorr AF. Practical uti- lity of clinical prediction rules for suspected acute pul- monary embolism in a large academic institution.
Thromb Res 2004; 113: 1-6.
9. Ulukavak Ciftci T, Kokturk N, Demir N, et al. Comparison of three clinical prediction rules among patients with suspected pulmonary embolism. Tuberk Toraks 2005;
53: 252-8.
10. Boyacı H, Yıldız F, Başyiğit I, Pala A. The role of clinical scoring methods in the diagnosis of pulmonary embo- lism. Solunum Hastalıkları 2006; 17: 111-5.
11. Iles S, Hodges AM, Darley JR, et al. Clinical experience and pre-test probability scores in the diagnosis of pulmo- nary embolism. Q J Med 2003; 96: 211-5.
12. Le Gal G, Righini M, Roy PM, et al. Prediction of pulmo- nary embolism in the emergency department: The revi- sed geneva score. Ann Intern Med 2006; 144: 165-71.
13. Okumus G, Yalnız E, Musellim B, et al. The diagnostic possibilities of venous thromboembolism in our hospi- tals. Turkish Thoracic Society. 8thAnnual Congress, An- talya, 2005; 6(Supp 1): p. 206.
