An authoritative algorithm most appropriate for the prediction of pulmonary embolism in patients with AECOPD

Chronic obstructive pulmonary disease (COPD)is a leading cause of morbidity and mortality worldwide that induces a substantially and increasingly economic and social burden [1, 2]. The acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is an acute worsening of respiratory symptoms that results in additional therapy [3, 4].

Acute pulmonary embolism (PE) is the third most frequent acute cardiovascular syndrome behind myocardial infarction and stroke globally. Among patients who die of PE, 34% died suddenly before therapy could be initiated or take effect, 59% was diagnosed after death and only 7% were correctly diagnosed with PE before death [5]. PE that has an explicit indication for anticoagulant treatment is frequently encountered in AECOPD [6]. If it happens, PE is significantly associated with increased mortality and length of hospital stay in patients with AECOPD [7‐9]. Consequently, PE should be ruled out when a patient with COPD has an acute exacerbation [10].

The hitherto globally recognized algorithms for the likelihood prediction of PE mainly comprise the Standard algorithm [11], the Age-adjusted algorithm [12], the YEARS algorithm [13], the PERC algorithm [14], and the PEGeD algorithm [15]. By the Standard algorithm, pulmonary embolism is ruled out by a D-dimer level of less than 500 nanogram(ng)/milliliter(ml) and a low clinical pretest probability (C-PTP) [11]. C-PTP is most frequently assessed via the Wells score [11] or the Geneva score [5]. By the Age-adjusted D-dimer algorithm, pulmonary embolism is ruled out in patients who are 50 years of age or younger and have a low or moderate C-PTP as well as a D-dimer level of less than 500 ng/ml, or who are older than 50 years of age with a D-dimer level that is less than 10 times the patient’s age [12]. By the YEARS algorithm, pulmonary embolism is ruled out in patients with none of the clinical signs of DVT, hemoptysis, or the most likely probability of pulmonary embolism and a D-dimer level of less than 1000 ng/ml and in those with one or more of the aforementioned three criteria and a D-dimer level of less than 500 ng/ml [13]. By the PERC algorithm, PE is ruled out in patients who meet all of the following criteria: age < 50 years, SaO2 > 94%, pulse < 100 beats per minute, no haemoptysis, no recent trauma or surgery, no history of VTE, no unilateral leg swelling, and no oral hormone use [14]. By the PEGeD algorithm, PE is ruled out in patients with a low C-PTP and a D-dimer level of less than 1000 ng/ml and in patients with a moderate C-PTP as well as a D-dimer level of less than 500 ng/ml [15]. For each algorithm, if PE cannot be ruled out while its likelihood being predicted by using the algorithm, the further chest imaging investigations for the confirmation of the presence or absence of PE are warranted. The summary of the variables being involved in each algorithm is in Table 1.

In a previous study, the PEGeD algorithm and the YEARS algorithm were both regard as the safest strategies for PE prediction at the cost of minimum number of chest imaging performance in general population [15]. Nevertheless, for patients with AECOPD, which one of those algorithms has the best diagnostic accuracy to predict the likelihood of a PE? In other words, which algorithm can safely exclude PE for patients with AECOPD via the minimum number of chest imaging performance? Those questions had remained unanswered before the present study. Thus the current study was performed under such circumstances.

The pretest prediction of pulmonary embolism is more critically vital for patients with AECOPD who have a higher risk of PE than the general population. The prevalence of PE in patients with AECOPD was approximately 16.1% [6], being close to the 18.1% (210/1158) in the current study, is well above that in general population which is approximately 0.1% [17]. It could be devastating for patients with AECOPD if their life-threatening pulmonary embolism are missed. However, having all patients with AECOPD tested for CTPA and/or V/Q scan will obviously waste plenty of medical resources and increase the suffering and adverse effects for the patients. As a result, to single out a prediction method which can highly accurately predict the likelihood of PE among the hitherto authoritative algorithms for the PE prediction is imperative for patients with AECOPD.

The variables involved in all five algorithms mainly comprise: DVT signs, PE likely, heart rate (HR), recent immobilization or surgery, history of VTE, hemoptysis, cancer, one D-dimer cutoff value(500 ng/ml), two D-dimer cutoff values(500 ng/ml, 1000 ng/ml or the age-adjusted), age, SaO2, and oral hormone use [11‐15]. Among those variables, DVT signs and hemoptysis are both adopted in all five algorithms, whereas one D-dimer cutoff value is only adopted in the Standard algorithm, meanwhile, SaO2 and oral hormone use are both adopted in the PERC algorithm only. The PE likely, HR, recent immobilization or surgery, and history of VTE are all adopted in four algorithms, respectively. The two D-dimer cutoff values and cancer are both adopted in three algorithms, respectively. The number of variables contained in the Standard algorithm, the Age-adjusted algorithm, the YEARS algorithm, the PERC algorithm, and the PEGeD algorithm were 8, 9, 4, 8, and 8, respectively. (Table 1).

First of all, the results of the current study demonstrated that the Standard algorithm was not appropriate for the prediction of PE in patients with AECOPD, based on its poor PPV (18.1%) and the excessive number of chest imaging examinations compared with other algorithms, despite its most variables were unavailable due to the study design per se. The Standard algorithm which is widely used in clinical practice is a more cautious criterion for the PE screening, in comparison with the Age-adjusted algorithm, the YEARS algorithm, and the PEGeD algorithm. The probability of PE can only be excluded if both of low C-PTP and D-dimer level less than 500 ng/ml are met in the Standard algorithm [11]. However, due to the frequently higher D-dimer level [18] and more immobilization [19] in patients with AECOPD than those in general population, the Standard algorithm often leads to an increased false positive rate of PE prediction as well as an excessive imaging investigations instead of missed diagnoses.

Secondly, the same goes with the PERC algorithm. Being usually above 50 years old or tachycardic or hypoxic or oral glucocorticoid user are the common characteristics of patients with AECOPD, regardless of the suspicious degree of PE, whereas PE can be safely excluded only when all criteria of the PERC including age < 50 years, pulse < 100 beats per minute, SaO2 > 94%, and no oral hormone use are met by the PERC rule [14]. In this way the PERC rule makes almost every patient with AECOPD to be suspected to have a PE, which is obviously unreasonable. As a result, this algorithm also leads to a startling high false positive rate and excessive chest imaging investigations, although its sensitivity and number of missed diagnoses were 98.6% and 3, respectively.

Thirdly, this study suggested that the diagnostic accuracy of the Age-adjusted algorithm was slightly inferior to that of the YEARS algorithm, while their diagnostic accuracy were both inferior to that of the PEGeD algorithm. We think it may be because the Age-adjusted algorithm still applies the classic standard of D-dimer level to patients with AECOPD aged less than 50 years [12],which may lead to an increased false positive rate of PE prediction in that patient group. For the YEARS algorithm, in view of its C-PTP determination only measures three criteria in the Wells score, whereas omits important risk factors such as recent immobilization, recent surgery and cancer [13], accordingly the false negative rate may be elevated.

As a comparison, the PEGeD algorithm performed better in the Youden index, diagnostic accuracy, number of necessary imaging examinations, and number of missed diagnoses than all other algorithms, despite its sensitivity was lower than that of the PERC algorithm(difference, 10.0%) and its number of missed diagnoses was slightly more than that of the PERC(difference, 1.8%). We deem that maybe because it adopts the criteria consisted of the low C-PTP combined with the high D-dimer level (1000 ng/ml), the moderate C-PTP combined with the standard D-dimer level (500 ng/ml), and the high C-PTP without the reference to D-dimer [15], achieving the perfect complementarity between the C-PTP and D-dimer level, thus greatly improving the sensitivity and specificity and then the diagnostic accuracy of the prediction of PE likelihood in patients with AECOPD.

The implication of this study lies in that, for the first time, the most appropriate algorithm for the PE likelihood prediction in patients with AECOPD to date was discovered. The clinical and socioeconomic value of the PEGeD algorithm consists in that it can maximumly accurately identify the likelihood of PE in patients with AECOPD, so as to maximumly minimize the frequency of CTPA and V/Q scan and the missed diagnoses concurrently, thereby reducing the potentially physical injury resulted from imaging investigations, and avoiding unnecessary health care costs as well as the waste of medical resources, on the basis of ensuring medical safety.

This study has some limitations despite its values. First of all, it was a retrospective study. A similar prospective study is warranted in the future. Secondly, a few patients who were contraindicated to CTPA underwent V/Q scan only in the present study. Although it is recommended to reject the diagnosis of PE if the V/Q scan is normal, and to accept that the diagnosis of PE if the V/Q scan yields high probability for PE [5], the V/Q scan alone is slightly less convincing, compared with the combination of CTPA and V/Q scan in the diagnosis of pulmonary embolism.

The results of the current study indicate that the PEGeD algorithm is the most appropriate authoritative strategy, compared with the Standard algorithm, the Age-adjusted algorithm, the YEARS algorithm and the PERC algorithm, for the likelihood prediction of pulmonary embolism in patients with AECOPD to date. The findings are expected to shed some light on the clinical practice in this field and be validated by prospective studies in the future.