PARIS score for evaluation of probability of SARS-CoV-2 infection in cancer patients.

Introduction

Among cancer patients, community respiratory viruses are the most common causes of respiratory infections [ 1 ] and adverse outcomes occur more frequently in comparison to immunocompetent patients [ 2 ]. In various Chinese studies, up to 3% of patients with COVID-19 had cancer, a figure much higher than the 0.29% overall cancer incidence in the same populations [ 3 ]. The mortality rate in these patients is also increased [ 4 ]. In the hematopoietic cell transplant population, age, lymphopenia, high-dose total body irradiation, and the presence of co-pathogens are significant risk factors for progression of a viral upper respiratory tract infection to a lower respiratory tract infection [ 5 – 7 ]. Previous reports concerning cancer patients and COVID-19 highlighted the following predisposing factors for severe complications such as intensive care unit (ICU) admission and death: Eastern Cooperative Oncology Group Performance Status (ECOG PS) ≥ 2 and progressive oncologic disease, presence of metastases, anti-cancer treatment administrated within 14 days, hematological malignancies, contact with health care providers with COVID-19, and patients with solid tumors [ 8 – 12 ].

Reverse transcription polymerase chain reaction (RT-PCR) is currently the reference test to diagnose patients with SARS-CoV-2 infection [ 13 ]. Computed tomography (CT) scan has a reported high sensitivity [ 14 ] but intrathoracic radiological signs can be delayed after clinical disease onset, with up to 56% CT negativity in the first 3 days of symptomatic infection [ 15 ]. Imaging is indicated in case of moderate to severe disease manifestations, but can be omitted in case of mild symptoms consistent with COVID-19 and no risk factors for disease progression. Even though CT scans may have a higher sensitivity than PCR (71–95% for PCR and 97–98% for CT), their specificity is lower and radiologists may experience difficulties for differentiating SARS-CoV-2 from non-SARS-CoV-2 pneumonia [ 16 ]. Moreover, false-negativity of PCR, a negative CT, and SARS-CoV-2 pneumonia mimickers on CT may lead to inaccurate diagnoses.

In these situations, a Bayesian approach of the diagnosis with the calculation of a pre-test probability combining clinical and biological features has proven to be a very useful tool and is already used in clinical practice in the management of patients with suspected pulmonary embolism [ 17 ]. Such an approach could help a better identification of the population at risk which need to be tested and can improve patient management in the pretest and time-to-result periods but can also improve diagnostic confidence in PCR-negative cases.

A new prediction score called “Pre-test probability for SARS-CoV-2 Infection based on Scoring (the “PARIS score”)” has been developed by Tordjman et al. [ 18 ]. The score (ranged from 0 to 5) is based on the evaluation of complete blood cell count (lymphocytes < 1.3 G/L, eosinophiles < 0.06 G/L, basophiles < 0.04 G/L, and polymorphonuclear < 5 G/L) and has been reported to have a good performance to evaluate the pre-test probability of SARS-CoV-2 infection and to identify groups at low and high risk of infection. This approach could help to avoid or delay diagnostic testing in low-risk patients and quarantining the high-risk patient group until obtaining a negative RT PCR test as well as negative CT scanner. The original publication of Tordjman et al. claimed a negative predictive value (NPV) for a low score/low risk (0–1) of 98% while the positive predictive value (PPV) for a high score/high risk (4–5) was between 80 and 92%. The overall disease prevalence in the studied population was 64% but analysis of different time periods showed a similar discriminant capability down to a disease prevalence of 31%. However, it has to be remembered that NPV and PPV values are strongly affected by the prevalence of the disease in the tested population [ 18 ].

To date, a number of other similar approaches have been proposed [ 19 ] but the PARIS score is simple to use in clinical practice and is requiring readily available data. However, this score has never been evaluated in cancer patients whose biological parameters used to calculate the score are often altered by the disease and/or the therapy.

The aim of our study is to evaluate if the PARIS score remains useful in cancer patients for evaluating the pre-test probability of SARS-CoV-2 infection.

Patients and methods

We performed a retrospective study by including patients presenting at the emergency department of the Jules Bordet Institute between March 10 and May 18, 2020. All patients irrespective of their tumor type (solid tumor or hematological malignancy) and cancer status (diagnostic period, treatment, or follow-up) were eligible provided they had an unplanned visit at the emergency department whatever the reason and were tested with a SARS-CoV-2 RT-PCR test on a naso-pharyngeal swab.

The following data were extracted from medical records: symptoms, comorbidities, performance status, type and stage of cancer, anti-cancer treatment, blood count, results of thoracic imaging, and outcome.

The evaluation of the CT scans and the quantification of disease extent were performed according to the recommendations of the European society of radiology [ 20 ]. A grading of the pulmonary involvement was coded as follows: grade 0, no involvement; grade 1, < 10%; grade 2, 10–25%; grade 3, 25–50%; grade 4, 50–75%; grade 5, > 75%.

The study was approved by our Ethical Committee on 21/09/2020 (n°CE3213).

Statistical analysis

Based on the results of the SARS-CoV-2 PCR test, positive and negative patients groups were compared in terms of symptoms, comorbidities, performance status, type and stage of cancer, anti-cancer treatment, blood count, imaging results, and outcome (vital status at 30 days after emergency department admission). For categorical variables, a likelihood chi-square test, Mantel–Haenszel chi-square test, or Fisher exact test was used while continuous variables were assessed with the Mann–Whitney test. A p < 0.05 was considered significant.

We assessed the diagnostic performance of the PARIS score (considering the 3 classes: low-probability, intermediate, and high probability) in terms of sensitivity and specificity. The NPV and PPV were also calculated on the same population with a focus on the NPV in the low probability class. The 95% confidence intervals (CI) for all parameters were calculated using the Wilson method. We calculated the area under the ROC curve with 95% CI.

The sample size calculation performed before starting the study was as follows: statistical assumptions were made for n = 170 patients expected to be included and a 15 to 20% prevalence of positive SARS-CoV-2 test: 100 patients in the low probability class, 42 in the intermediate, and 7 in the high probability class. Assuming an undegraded sensitivity of 80% as in the general population [ 18 ]: with n = 30 SARS-CoV-2-positive patients, the 95% CI for sensitivity would be 63 to 90%.

For an expected NPV of 99%, in an n = 100 population with a low-probability PARIS score, the lower 95% CI limit would be 95%. For a PPV of 92%, assuming n = 27 cases with high-probability PARIS score, the 95% CI limits would be 77 to 98%.

Results

A total of 164 patients were admitted in the emergency department (ED) between March 10 and May 18, 2020, and tested with an RT-PCR test on a naso-pharyngeal swab for SARS-CoV-2 RNA presence for a total of 181 tests with 16 patients tested negative twice and one patient tested thrice (Table 1 ). The median number of days between two presentations was 17 days (min 9 days) with a mean 19.1 ± 7.6 days. We thus considered these as separate events. No previously positive patients were tested twice. Demographic data of recruited patients is presented in Table 2 .

Out of the 181 SARS-CoV-2 tests performed: 155 were negative and 26 were positive of which 8 were treated as outpatients, 18 were hospitalized, and 6 of them died while 12 were discharged alive. Out of these, 139 had solid tumors and 42 hematological malignancies (Table 2 ). The proportion of COVID-19-positive tests was higher in hematological patients: 26% (11/42) vs. 11% (15/139) in solid tumor ( p = 0.02).

Cough, myalgia, and digestive symptoms were statistically more frequent in SARS-CoV-2-positive patients, 62%, 19%, and 35% respectively compared to 37% ( p = 0.02), 5% ( p = 0.02), and 13% ( p = 0.02) in patients without SARS-CoV-2 infection. There was no statistical association between a PCR SARS- CoV-2-positive result and other symptoms such as headache ( p = 0.21) or temperature > 37.5 °C during the last 72 h ( p = 0.16).

A total of 75 patients (41%) received an anti-cancer treatment in the last 14 days before PCR testing and 29 (16%) within 14 to 30 days. We found no association between administration of an anti-cancer treatment and a positive swab.

For each biological parameter of the PARIS score, a comparison was made between the SARS-CoV-2-positive and SARS-CoV-2-negative population but only the total neutrophil count differed between positive and negatively tested individuals (median 2965 cells/mm 3 and 4840 cell/mm 3 respectively p = 0.02). The other laboratory parameters were not different between PCR SARS-CoV-2-positive and SARS-CoV-2-negative patients (Table 1 ).

Data available permitted the calculation of a PARIS score for 166 of the 181 visits (Table 3 ): 18 (11%) were in the low probability risk group (0 and 1), 66 (40%) in the intermediate probability risk group (score 2 and 3), and 82 (49%) in the high probability risk (score 4 and 5). There was no statistical associated between the PARIS score and the SARS-CoV-2 test result ( p = 0.09). The PARIS score had an area under the ROC curve of only 0.61 (95% CI, 0.48 to 0.73) (Fig. 1 ). Only 15/24 cases, i.e., 63% of the SARS-CoV-2-positive cases were classified as a “high probability risk” by the PARIS score (95% CI, 43 to 79%). Only 16/142, i.e., 11% (95% CI, 7 to 18%) of the patients with a negative SARS-CoV-2 PCR were classified as “low probability risk”; 53% were in the “low or intermediate probability risk” class: 75/142 (95% CI, 45 to 61%). The predictive value of “low probability risk” score to predict a negative SARS-CoV-2 PCR result was 89% (95% CI, 67 to 97%) ( n = 16/18); the predictive value of “high probability risk” score to predict a positive SARS-CoV-2 PCR result was 18% (95% CI, 11 to 28%) ( n = 15/82).

A chest CT scan was performed in 130 out of the 181 presentations. The median grade of pulmonary involvement was 1 in COVID-19-negative patients and 2 in COVID-19-positive patients (Table 4 ). The chest CT scan was considered suspect for 70 cases of which 53 had a negative RT-PCR test. Out of the 60 negative CT scans, only 5 patients had a positive PCR test. The diagnostic performance of the CT scan was as follows: sensitivity 77% (95% CI, 57 to 90%), specificity 51% (42 to 60%), PPV 24% (16 to 36%), and NPV 92% (82 to 96%).

Discussion

During epidemic episodes, a rapid diagnosis and triage of COVID-19 suspect patients are of paramount importance. This relies mainly on clinical grounds but simple tools like clinical/biological/radiological parameters/scores are of great utility in selecting an appropriate attitude toward acutely ill patients in order to maximize the timely use of the more complex diagnostic methods. The PARIS score was proposed as having excellent operational characteristics in the general population, is very simple to use, and can be calculated in any emergency department in less than 1 h. Thus, it was alleged that patients can be correctly managed while waiting the results of molecular testing.

In our retrospective analysis, these findings were not confirmed and thus the PARIS score is probably not applicable to an oncological population.

When comparing our population with the originally published series, the prevalence of positive cases is much lower (14.4%) compared with an average of 68% and a minimum of 31% in the publication of Tordjman [ 18 ]. However, our figures are in line with the disease prevalence in Belgium in the Brussels area during the same period (March 10–May 18, 2020) which was between 9 and 10% for most of the period except during May 2020 when it slowly decreased to around 5% [ 21 ]. Oncologic patients are very frequently immunosuppressed and despite taking a lot of precautions, they frequently cannot afford to delay hospital visits (and thus transportation and minimal social contact) which probably explain the higher incidence of positive tests. This prevalence of diseased patients better approaches the real situation in an emergency department than in the population Torjdman studied. The higher prevalence of the disease in the Tordjman series is probably explained by the inclusion of patients hospitalized in a ward.

Among oncological patients, we found that a hematological cancer is a risk factor for having a positive PCR test for SARS-CoV-2 and is very probably explained by the intrinsic state of more profound immunosuppression as well as the treatments received which are often more aggressive. In our population, few symptoms were statistically associated with the diagnosis of SARS-CoV-COVID-19 pneumonia, probably because these symptoms can be explained by primary or metastatic neoplasm, paraneoplastic conditions, or can be related to the treatment of the oncological condition.

One can postulate that the hematological toxicity secondary to the high proportion of patients receiving a chemotherapy prior to emergency department admission is associated with changes in hematological parameters involved in the calculation of the PARIS score and thus explain the difference of our results with those in Tordjman’s study in the general population. However, in our analysis, neither treatment received nor time from treatment to ED admission was associated with an increased risk of a positive PCR. Thus, the use of an immunosuppressive treatment alone cannot explain the lack of validity of the PARIS score. It would therefore be interesting to study the PARIS score in samples from a non-oncological population but benefiting from immunosuppressive therapy as for example the organ transplant population.

We performed a simulation of the positive predictive values at different levels of disease prevalence (not presented here) that shows that a high PARIS score will be no more useful at higher disease prevalence levels in an oncological population.

The negative predictive value of the PARIS score calculated from our sample was 0.91, which could be very interesting in excluding COVID-19, but at a low disease prevalence rate, this concerns a minority of patients.

Another issue could be an initial overestimation of the validity of the PARIS Score as the population used both for score derivation and score validation had a very high disease prevalence biasing toward a more favorable operational characteristics by blunting the number of false positive. Moreover, the initial score included patients from medicals wards with a possible bias toward more severe COVID-19 cases.

When analyzing each component of the PARIS score individually, only neutropenia was associated with an increased risk of a SARS-CoV-2-positive PCR in our group. However, the median neutrophil count in the negative SARS-CoV-2 PCR population is 4840/mm 3 and is below the threshold associated with an increased risk of positive PCR as defined by the PARIS score (< 5000/mm 3 ). Thus, it is probable that a specific threshold needs to be determined for oncological patients under therapy. Our series was not large enough to permit that kind of analysis.

Interestingly, in our sample, a suspicious chest CT scan was not associated with an increased risk of having a positive SARS-CoV-2 PCR and the sensitivity and specificity were low. This is probably also explained by the fact that the ground glass opacities found in COVID-19 are also found in a number of infections frequently encountered in oncological patients (such as viral pneumonia, pulmonary aspergillosis; carcinomatous lymphangitis and toxicities of anti-cancer treatments). One can postulate that combining CT with the PARIS score may be eventually useful but given the suboptimal operational characteristics of both tests in the oncologic population, this is unlikely to be true.

A number of other similar approaches have been promoted and a recent systematic review has been published [ 19 ]. However, data availability precluded us to make a complete comparison between the different scores in our population. It is however probable that similar studies in the oncologic population need to be considered for each of these approaches before they can be promoted in clinical practice.

In conclusion, we showed that the PARIS score has little applicability in an oncological population. Specific score/approaches should be derived for this population.

Author contribution

Ameye Lieveke, Loizidou Angela, Grigoriu Bogdan, and Meert Anne-Pascale contributed to the study conception and design. Data collection and analysis were conducted by Ameye L, C Gueuning, A Loizidou, and AP Meert. All authors contributed to the interpretation of the findings. The first draft of the manuscript was written by C Gueuning, B Grigoriu, and AP Meert. L Ameye and A Loizidou provided supervisory support and reviewed this paper. All authors contributed to the revision of the manuscript and approved the final manuscript.

Funding

We thank “Les Amis de l’Institut Jules Bordet.”

Data availability

On demand

Code availability

Not applicable

Declarations

Ethics approval

This study was approved by the ethic committee of the Institut Jules Bordet (21/09/2020).

Consent to participate

Not applicable

Consent for publication

Not applicable

Conflict of interest

The authors declare no competing interests.

Footnotes

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Whimbey E, Englund JA, Couch RB. Community respiratory virus infections in immunocompromised patients with cancer. Am J Med. 1997;102(3A):10–8. doi: 10.1016/S0002-9343(97)80004-6.

2. Hijano DR, Maron G, Hayden RT. Respiratory viral infections in patients with cancer or undergoing hematopoietic cell transplant. Front Microbiol. 2018;9:3097. doi: 10.3389/fmicb.2018.03097.

3. Desai A, et al. COVID-19 and cancer: lessons from a pooled meta-analysis. JCO Global Oncol. 2020;6:557–559. doi: 10.1200/GO.20.00097.

4. Meng Y, et al. Cancer history is an independent risk factor for mortality in hospitalized COVID-19 patients: a propensity score-matched analysis. J Hematol Oncol. 2020;13(1):75. doi: 10.1186/s13045-020-00907-0.

5. Fareed K, Roy FC. Respiratory syncytial virus in hematopoietic cell transplant recipients and patients with hematologic malignancies. Haematologica. 2019;104(7):1322–1331. doi: 10.3324/haematol.2018.215152.

6. Atilla E, et al. Upper respiratory viral infections in patients with haematological malignancies after allogeneic haematopoietic stem cell transplantation: a retrospective study. Antivir Ther. 2018;23(6):523–527. doi: 10.3851/IMP3224.

7. Pochon C, Voigt S. Respiratory virus infections in hematopoietic cell transplant recipients. Front Microbiol. 2019;9:3294–3294. doi: 10.3389/fmicb.2018.03294.

8. Kuderer NM, et al. Clinical impact of COVID-19 on patients with cancer (CCC19): a cohort study. Lancet. 2020;395(10241):1907–1918. doi: 10.1016/S0140-6736(20)31187-9.

9. Dai M, et al. Patients with cancer appear more vulnerable to SARS-CoV-2: a multicenter study during the COVID-19 outbreak. Cancer Discov. 2020;10(6):783–791. doi: 10.1158/2159-8290.CD-20-0422.

10. Zhang L, et al. Clinical characteristics of COVID-19-infected cancer patients: a retrospective case study in three hospitals within Wuhan, China. Ann Oncol. 2020;31(7):894–901. doi: 10.1016/j.annonc.2020.03.296.

11. He W, et al. COVID-19 in persons with haematological cancers. Leukemia. 2020;34(6):1637–1645. doi: 10.1038/s41375-020-0836-7.

12. Yang K, et al. Clinical characteristics, outcomes, and risk factors for mortality in patients with cancer and COVID-19 in Hubei, China: a multicentre, retrospective, cohort study. Lancet Oncol. 2020;21(7):904–913. doi: 10.1016/S1470-2045(20)30310-7.

13. Corman VM, et al. Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR. Eurosurveillance. 2020;25(3):2000045. doi: 10.2807/1560-7917.ES.2020.25.3.2000045.

14. Ai T, et al. Correlation of chest CT and RT-PCR testing for coronavirus disease 2019 (COVID-19) in China: a report of 1014 cases. Radiology. 2020;296(2):E32–E40. doi: 10.1148/radiol.2020200642.

15. Bernheim A, et al. Chest CT findings in coronavirus disease-19 (COVID-19): relationship to duration of infection. Radiology. 2020;295(3):200463. doi: 10.1148/radiol.2020200463.

16. Bai HX, et al. Performance of radiologists in differentiating COVID-19 from non-COVID-19 viral pneumonia at chest CT. Radiology. 2020;296(2):E46–E54. doi: 10.1148/radiol.2020200823.

17. Le Gal G, et al. Prediction of pulmonary embolism in the emergency department: the revised Geneva score. Ann Intern Med. 2006;144(3):165–171. doi: 10.7326/0003-4819-144-3-200602070-00004.

18. Tordjman M, et al. Pre-test probability for SARS-Cov-2-related infection score: the PARIS score. PLoS ONE. 2020;15(12):e0243342. doi: 10.1371/journal.pone.0243342.

19. Locquet M, et al. A systematic review of prediction models to diagnose COVID-19 in adults admitted to healthcare centers. Arch Public Health. 2021;79(1):105. doi: 10.1186/s13690-021-00630-3.

20. Revel MP, et al. COVID-19 patients and the radiology department - advice from the European Society of Radiology (ESR) and the European Society of Thoracic Imaging (ESTI) Eur Radiol. 2020;30(9):4903–4909. doi: 10.1007/s00330-020-06865-y.

21. Sciensano. Belgium COVID-19 epidemiological situation - interactive dashboard https://epistat.wiv-isp.be/covid/ and https://datastudio.google.com/embed/reporting/c14a5cfc-cab7-4812-848c-0369173148ab/page/cUWaB . 2021 [cited 2021 11/07/2021]

Associated Data

Data Availability Statement

On demand

Not applicable