Use of Synofast® in routine rheumatology practice: diagnostic performance and impact on management of an acute arthritis

Abstract

Background

Septic arthritis (SA) is a rare but serious disease. There is no reliable marker for rapid diagnosis. Synofast® is a new device that analyses synovial fluid and can accurately rule out SA within 15 minutes.

Objective

To evaluate the diagnostic performance of Synofast® in routine practice in patients with suspected SA and to assess its impact on patient care.

Method

This was a monocentric prospective study of patients with acute arthritis of possible septic origin. In addition to the usual blood and synovial bacteriological tests, a drop of synovial fluid was taken for the Synofast® test, which provides three SA risk probabilities: “low”, “possible” or “high”. The physician recorded the management he would have given before the test and the management he actually gave after the test, knowing the diagnostic performance reported in previous studies.

Results

Over one year, 43 patients were included, of whom eight had SA (18.6%). Sensitivity, negative predictive value (PV) and negative likelihood ratio (LR) were respectively of 75%, 93% and 0.32 when analysing for “high” and “possible” septic risk while specificity, positive PV and positive LR were respectively of 94%, 67% and 8.75 when analysing for “high” septic risk only. The test helps to avoid hospitalisations in 1/29 patients and to defer antibiotic in 1/7 patients. Follow-up and drug management remained unchanged in 95% and 89% of patients, respectively.

Conclusion

Synofast® seems to be a valuable tool for the exclusion of SA. The results need to be confirmed in a larger population, especially the impact on management.

Supplementary Information

The online version contains supplementary material available at 10.1186/s42358-025-00506-y.

Introduction

Bacterial septic arthritis (SA) is caused by the joint invasion of a bacterium that triggers a local and systemic inflammatory response. It usually presents as an acute arthritis with pain and swelling in one or a few joints, with reduced range of joint motion and sometimes associated with fever, and develops rapidly over a few days [1]. Its incidence rate varies from 4 to 16 cases per 100,000 population per year, increasing with age to 73/100,000 after 90 years of age [2, 3]. SA can cause severe morbidity with a mortality rate of 10%, prolonged hospitalisation, a 6-fold increased risk of having joint replacement and irreversible functional loss in up to 50%, making it a diagnostic and therapeutic emergency [2, 4–6].

The clinical presentation of SA is non-specific and can be confused especially with gout or pseudogout in particular, but also with rheumatoid arthritis, which are the main differential diagnoses and risk factors for SA [2, 7]. For example, fever, often considered as a specific sign of SA, is present in half of all non-SA patients [8]. Synovial fluid (SF) analysis is neither specific nor sensible: joint leukocyte count may be higher in SA, but without a discriminatory threshold; microcrystals are found in 25% of SF aspirates in SA, and Gram stain is positive in only 30–70% of cases [2, 3, 8, 9]. Definitive diagnosis is therefore based on the detection of a cultivable germ, most commonly in SF but also in peripheral blood. This synovial culture takes a long time, usually 48–72 hours to several days, and remains negative in up to 35–40% of cases [2, 6, 10, 11].

Treatment of SA is based on antibiotics, started intravenously and then adapted to the germ, together with drainage of the affected joint. Given the potential serious morbidity of SA, the usual approach is therefore to hospitalise patients presenting with possible SA while awaiting the results of the bacterial cultures. There is therfore a lack of a discriminatory test that can quickly rule out the diagnosis of SA and help avoid unnecessary hospitalisation. According to the French National Health Service reference cost, the average cost of a day of hospitalisation for ‘arthropathy’ is €1,700 to €2,800, and it often takes three days of hospitalisation to obtain an aetiologic diagnosis of arthritis [12, 13]. There is therefore also an economic issue.

To this end, a new device called Synofast®, which analyses the mid-infrared spectrum of a drop of SF (7 µl) in search of a bacterial metabolic signature, as illustrated in Fig. 1, can rule out the diagnosis of SA within 15 minutes, with a reported NPV of 98% [14–16]. These results imply significant potential consequences, such as shorter hospital stays, fewer invasive procedures and less antibiotic therapy. However, this new diagnostic test has not yet been evaluated in routine practice, nor has its impact on the physician’s decision.

Fig. 1.

Operating diagram of the Synofast® device

The main objective of this study was to confirm the validity of Synofast® in routine care at the Rheumatology Department of the University Hospital of Clermont-Ferrand, and to study its impact on the management of patients with acute arthritis.

Method

Study design, population, data

This is a prospective diagnostic study including patients with acute arthritis of possible septic origin, eevaluated by a rheumatologist at the University Hospital of Clermont-Ferrand, France, during a consultation or hospitalisation. Acute arthritis was defined as a hot swollen joint evolving for less than six weeks. Patients were included between April 30, 2023, and April 30, 2024. Exclusion criteria were prosthetic joint effusion, post-traumatic effusion and patients under 18 years of age.

Data collected included patient characteristics (sex, age, immunosuppression risk factors, history of rheumatic disease or microcrystal arthropathy), clinical features (number and location of joints affected, duration of symptoms, presence of fever or chills, macroscopic appearance of the SF), and biological features (CRP level, white blood cell count, SF leukocyte level and neutrophil percentage, presence of crystals).

Protocol

After aspirating the synovial fluid and filling the usual tubes required for diagnosis (cytology and culture), the observer added another drop of SF to the Synofast® device. The infrared spectroscopic analysis provided three different risk assessements in about fifteen minutes: “low”, “possible” or “high” septic risk. Meanwhile, the physician noted his own septic risk assessment, based on clinical and biological features, and reported the management he would have given before receiving the Synofast® result. Then, he reported the management he actually carried out after receiving the result of the test. Management was assessed in terms of referral (hospitalisation vs follow-up consultation vs no follow-up) and in term of treatment (colchicine, antibiotic, anti-inflammatory drugs [oral NSAID, oral corticosteroid, joint corticosteroid injection], multiple choice). The protocol was approved by the local ethics committee (DRCI of Clermont-Ferrand University Hospital, IRB00013412) on April 19, 2023 (IRB number 2023-CF047) in accordance with the Public Health Code and French data protection laws.All participants were informed about the study and retained the right to oppose the use of their data. Only participants who did not express opposition were included. The authors declare that they have no relationship with, or conflicts of interest regarding the developer or the manufacturer of Synofast®. The study was conducted at the authors’ own initiative and received no funding or material support from the company.

Case definition

Definitive diagnosis was made after obtaining the results of the SF culture and blood culture. SA was considered if there was bacterial growth in the SF culture (Newman’s criterion 1) or in the blood culture only (Newman’s criterion 2), after exclusion of possible contamination [10]. Arthritis of other aetiologies (e.g., microcrystalline, inflammatory, etc.) was classified as non-septic.

Data analysis

Statistical analysis was performed using Stata software (version 16; StataCorp, College Station, TX, USA). All tests were two-sided, with an alpha level set at 0.05. Categorical data are expressed as number of patients and associated percentages, and quantitative data as mean ± standard deviation or median [25th; 75th percentiles]. Comparisons between SA and non-SA patients were made by the chi-squared test or the Fisher’s exact test for categorical data, and by the Mann-Whitney test for quantitative data. Diagnostic performance was assessed using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR +), and negative likelihood ratio (LR −). These measures were expressed with 95% confidence intervals (CI).

Sensitivity, NPV and LR − were calculated for patients who received a “high” septic risk or a “possible” septic risk response, to estimate the ability to rule out SA when a “low” septic risk result is found by Synofast® (screening test). On the other hand, specificity, PPV, and LR + were calculated for the “high” septic risk result to estimate the ability to confirm the diagnosis of SA (diagnostic confirmation). A secondary analysis was performed for SA diagnosed by SF culture alone (Newman criterion 1). A third analysis was performed to determine whether a rheumatologist’s septic risk assessment, in addition to the Synofast® result, improved diagnostic performance, particularly in ruling out SA. A predefined septic risk threshold of 40% was chosen. This corresponds to a “possible” risk in clinical practice, where hospitalisation for SF culture control can reasonably be avoided and the patient reassessed in consultation. Sensitivity, specificity, PPV, NPV, LR + and LR − were therefore analysed for patients with either rheumatologist’s septic risk assessment ≥ 40% or a Synofast® result of “high” or “possible” risk.

Results

Demographics

During this one-year study, 51 patients were included. Eight patients were excluded from the analysis: three due to missing data, and five due to arthritis lasting more than six weeks (Fig. 2). The study population thus consisted of 43 patients with a mean age of 70.9 ± 17.8 years, of whom 74.4% were male. The characteristics of the patients are shown in Table 1

Fig. 2.

Flowchart of the study population

Table 1.

Patients demographic characteristics

	Total (N = 43)	SA (N = 8)	NON-SA (N = 35)			p
Sex, female	11 (25,6)	1 (12,5)		10 (28,6)	0,66
Age (years)	70,9 ± 17,8	81,3 ± 11,7		68,5 ± 18,2	0,053
Symptoms duration (days)	4 [2; 7]	6 [4; 9]		4 [2; 7]	0,13
Clinical signs
Fever (≥38 °C)	14 (32,6)	4 (50)		10 (28,6)	0,40
Chills	5/41 (12,2)	1 (12,5)		4/33 (12,1)	1,00
Number of articulations					1,00
Polyarticular	14 (32,6)	3 (37,5)		11 (31,4)
Monoarticular	29 (67,4)	5 (62,5)		24 (68,6)
Knee	18	0		18
Gross appearance					0,02
Purulent	9 (20,9)	5 (62,5)		4 (11,4)
Turbid	29 (67,4)	3 (37,5)		26 (74,3)
Haemorrhagic	1 (2,3)	0 (0)		1 (2,9)
Clear	4 (9,3)	0 (0)		4 (11,4)
Biological tests
CRP (mg/L)	96 [36; 206]	193 [103; 249]		76 [34; 150]	0,10
Leucocytes, blood (G/L)	9,76 ± 3,81	8,47 ± 3,86		10,05 ± 3,79	0,30
Neutrophils, blood (G/L) (n=40)	6,8 [5,0; 8,7]	6,3 [2,2; 8,0]		6,8 [5,1; 8,7]	0,42
Leucocytes, synovial fluid (/µL) (n=40)	38 500 [11 350; 76 100]	135 450 [40 000; 165 000]		31 050 [11 000; 69 100]	0,12
Neutrophil (%) (n=37)	80,4 ± 22,4	73 ± 39,1		81,5 ± 19,4
Microcrystal presence	24/42 (57,1)	3/7 (42,9)		21 (60,0)	0,44
MSU	11	1		10
CPP	17	3		14
Risk factors
Infectious risk factor	23 (53,5)	5 (62,5)		18 (51,4)	0,70
Immunodeficiency	10 (23,2)	1 (12,5)		9 (25,7)	0,66
Diabetes	13 (30,2)	4 (50,0)		9 (25,7)	0,22
History of rheumatic disease	21 (48,8)	4 (50,0)		17 (48,6)	1,00
Chronic inflammatory arthritis	6 (14,0)	0 (0)		6 (17,1)	0,57
Cristal-induced arthritis	17 (39,5)	3 (37,5)		13 (37,1)	1,00
Other (SA)	2 (4,7)	2 (25,0)		0 (0)	0,03
Priori treatment
Colchicine	13 (30,2)	3 (37,5)		10 (28,6)	0,68
NSAID	6 (14,0)	0 (0)		6 (17,1)	0,57
Antibiotics	10 (23,3)	4 (50,0)		6 (17,1)	0,07
Clinician estimated risk for SA (%)	44 ± 28	83 ± 17		35 ± 23	≤0,001
Synofast® RESULTS
Low risk	29 (67,4)	2 (25,0)		27 (77,1)
Possible risk	8 (18,6)	2 (25,0)		6 (17,1)
High risk	6 (14,0)	4 (50,0)		2 (5,7)

Express in number (%), mean (± SD) or median (Q1; Q3)

SA = septic arthritis; non-SA = non septic arthritis; CRP = C-reactive protein; MSU = monosodic urate crystals; CPP = calcium pyrophosphate; NSAID = non-steroidal anti-inflammatory drug

Of the 43 patients, eight had SA, including two with negative SF culture but positive blood culture, giving a prevalence of SA of 18.6% (95%CI: 8.4 to 33.4). Patients with SA tended to be older than non-SA patients (81.3 ± 11.7 vs 68.5 ± 18.2 years; p = 0.053), were presenting more often with purulent fluid (62.5% vs 11.4%; p = 0.006) and had a higher likelihood of SA as estimated by the clinician (83 ± 17% vs 35 ± 23%; p < 0.001). Thirty percent of patients had diabetes and 48.8% had a rheumatological history (14.0% with inflammatory rheumatism, 37.2% with microcrystalline disease, and 4.7% with a history of SA). Seventy-four percent of patients were already hospitalised at the time of testing. Bacteraemia was observed in 6/8 cases of SA. Characteristics of the patients with SA are shown in Supplementary Appendix (Table S1). The most commonly affected joint in cases of SA was the shoulder (n = 3), and three patients had polyarticular involvement. Half of the cases of SA were treated with antibiotics at the time of the joint aspiration, for a mean of five days (range from two to eight days). Of the non-SA cases (n = 35), 22 were microcrystalline, 11 were inflammatory arthritis and two were of another aetiology (one osteoarthritis flare-up and one aseptic reactive arthritis after intra-articular injection).

Diagnostic performances of Synofast® alone

The distribution of Synofast® test results by final diagnosis is shown in Table 2 and its diagnostic performance is shown in Table 3. For arthritis classified as “possible + high” septic risk, the Synofast® test had a sensitivity of 75% (95%CI: 35 to 97) and a NPV of 93% (95%CI: 77 to 99). For arthritis classified as “high” septic risk, the test had a specificity and a PPV of 94% (95%CI: 81 to 99) and 67% (95%CI: 22 to 96), respectively.

Table 2.

Synofast® according to final diagnosis

Synofast® results Final diagnosis	“Low risk”	“Possible risk”	“High risk”	Total
Septic	2	2	4	8 (19%)
Cristal-induced arthritis	16	5	1	22 (51%)
Chronic inflammatory arthritis	9	1	1	11 (25%)
Other	2	0	0	2 (5%)
Total	29 (67%)	8 (19%)	6 (14%)

Table 3.

Diagnostic performances of Synofast®

Synofast® response	p	Se	NPV	LR-	Sp	PPV	LR +
“possible septic risk” + “high septic risk”	0,186 (0,084–0,334)	75% (35–97)	93% (77–99)	0,32 (0,10–1,09)	77% (60–90)	43% (18–71)	3,28 (1,58–6,80)
Synofast® response	p	Sp	PPV	LR +	Se	NPV	LR-
“high septic risk”	0,186 (0,084–0,334)	94% (81–99)	67% (22–96)	8,75 (1,93–39,75)	50 (18–84)	89 (57–97)	0,53 (0,26–1,07)

p: prevalence; Se: sensibility; Sp: specificity; PPV: positive predictive value; NPV: negative predictive value; LR +: positive likelihood ratio; LR-: negative likelihood ratio

For the two patients with SA but a Synofast® test assessing a “low” septic risk, they were also treated with antibiotics for at least six days before the diagnosis. One of the two had a negative CRP and a low SF cellularity (1133 leucocytes per microliter). Considering only the six SA cases with positive SF culture (Newman’s criterion 1), the NPV and sensitivity both increased to 100% both (95%CI: 88 to 100 and 95%CI: 54 to 100, respectively) for a 14.0% prevalence of the disease (95%CI: 5.3 to 27.9) (Supplementary Appendix, Table S2).

Diagnostic performances of Synofast® combined with rheumatologist septic estimation

The level of agreement between Synofast® and clinical judgement for the diagnsosis of SA was 65.1% (Cohen’s kappa coefficient of 0.34). When grouping patients who either had a pre-test probability of SA estimated by the rheumatologist to be greater ≥ 40% or were classified by Synofast® as “possible” or “high” septic risk, the sensitivity was 100% (95%CI: 63 to 100), the NPV was 100% (95%CI: 79 to 100) and the LR − was 0 (Table 4). The comparison of the sensitivities of Synofast® alone and in combination with the rheumatological septic estimation remained non-significant (p = 0.5).

Table 4.

Diagnostic performances combining the rheumatologist risk appreciation (estimated septic risk > 40%) and the Synofast® results (combination of “possible septic risk” and “high septic risk”)

p	Se	Sp	PPV	NPV	LR +	LR-
0,19 (0,08–0,33)	100% (63–100)	46% (29–63)	30% (14–50)	100% (79–100)	1,84 (1,36–2,50)	0

p: prevalence; Se: sensibility; Sp: specificity; PPV: positive predictive value; NPV: negative predictive value; LR +: positive likelihood ratio; LR-: negative likelihood ratio

Impact on management of acute arthritis

This analysis was performed on 42 patients due to missing data for one patient (Table 5)

Table 5.

Patients management before and after Synofast® test results

	Total (n = 42)			“Low risk” (n = 28)	“Possible risk” (n = 8)	“High risk” (n = 6)
Unchanged referral after Synofast® test	40 (95%)			26 (93%)	8 (100%)	6 (100%)
Hospitalisation initially planned	29			17	6	6
Maintained	28 (97%)			16 (94%)	6 (100%)	6 (100%)
Changed to follow-up consultation	1 (3%)			1 (6%)
Follow-up consultation initially planned	12			11	1	0
Maintained	11 (92%)			10 (91%)	1 (100%)
Changed to “no follow-up”	1 (8%)			1 (9%)
No follow-up initially planned	1			0	1	0
Maintained	1 (100%)				1 (100%)
Unchanged treatment after Synofast® test	39/44 (93%)			28/29 (97%)	7/9 (78%)	7/7 (100%)
Colchicine initially planned	31			24	6	1
Maintained	29 (94%)			23 (96%)	5 (83%)	1 (100%)
Changed to CSI or colchicine + CSI		2 (6%)		1 (4%)	1 (17%)
NSAID or OCS or CSI initially planned			5	3	2	0
Maintained	5 (100%)			3 (100%)	2 (100%)
Antibiotics initially planned	7			2	1	4
Maintained	6 (86%)			2 (100%)	0 (0%)	4 (100%)
Changed to colchicine	1 (14%)				1 (100%)
No treatment initially planned  Maintained	1 1 (100%)			0	0	1 1 (100%)

Percentages are rounded to the nearest unit. All the theorical possibilities of management changes are not shown, only effective changes are represented. NSAID = non-steroidal anti-inflammatory drug; OCS = oral corticosteroid; CSI = corticosteroid injection

Of the 29 initially planned hospitalisations, 28 were maintained despite a “low septic risk” result in the majority of patients (n = 17/29, 58.6%). The management of only one patient was changed from hospitalisation to a follow-up visit, when the test result was “low septic risk”. Of the 12 patients considered for a follow-up visit, 11 had a “low risk” of SA and the management of one of them was changed to systematic outpatient follow-up.

Of the 44 medications initially considered, 39 did not change (88.6%). One of the seven patients for whom the clinician had planned to start antibiotic therapy was finally treated with colchicine alone. Two other patients of the 31 who would have received colchicine had joint steroid injection.

Discussion

This first prospective routine practice study reports relatively good diagnostic performance of the Synofast® test, particularly with a high NPV (93%), allowing rapid exclusion of SA in patients with new-onset arthritis. These results are slightly lower than those of the Synofresh validation study where the NPV was estimated at 98% [16]. In our study, the sensitivity was moderate (75%), with a negative likelihood ratio of only 0.32, which is unlikely to change the post-test probability significantly [17]. The Synofresh study showed better results with a sensitivity of 89% and negative likelihood ratio of 0.12. These discrepancies may be partly due to the smaller sample size (43 patients vs 359), which makes it more susceptible to sampling variation, as evidenced by the wide confidence intervals. The reduced sensitivity is explained by two false-negative results among the eight SA cases. In these two cases, SA was diagnosed by blood culture as the SF remained sterile, and both patients were taking antibiotics for at least six days. As the Synofast® device analyses SF spectra to detect a bacteriological metabolic signature, it may be ineffective in cases where the SF culture is negative, particularly when the patient has been taking antibiotics for several days which may clear up the bacteria and so its metabolic signature. However, always in the Synofresh study, comparable diagnostic performance was reported between Newman’s criterion 1 and 2 [18]. This limitation should be considered when using the test, as it is common: 35–40% of SA cases have negative SF culture, while blood and SF cultures are both sterile in 19% of cases [2, 19].

For a positive diagnostic value, the PPV of Synofast® was poor, but the specificity was reasonable given the low prevalence of the disease and the small sample size. We chose to analyse NPV/sensitivity/LR − separately from PPV/specificity/LR + because clinicians are primarily interested in the ability of the test to exclude SA when the result is “low” septic risk (i.e. NPV and sensitivity for “possible” risk and “high” risk results together) and to confirm the diagnosis when the result is “high septic risk”.

We hypothesised that combining Synofast® with a rheumatologist’s estimate of the probability of SA risk, which considers all the available clinical and biological evidence as well as the clinician’s experience, would improve its diagnostic performance. However, our results were not statistically significant, most likely due to a lack of power. In our study, agreement between Synofast® and clinical judgement was relatively low. Therefore, studies are necessary to recommend the use of Synofast® alone, by example in emergency department, or as one element of diagnostic management among all the clinical and biological parameters.

There is currently no clinical sign or biological marker that is accurate enough to distinguish SA from other causes of arthritis. Several biomarkers have therefore been investigated. Synovial interleukin-6, calprotectin, and procalcitonin showed sensitivities, specificities and NPVs of 93%/64%/89%, 96%/44%/95% and 87%/94%/96% respectively [20–22]. SF PCR16s is effective, with a specificity close to 100% but sensitivity varies from 30% to 95% [23–26]. These tests are relatively expensive, analyses may be outsourced, as in our centre,and results could take several days. Cheaper and faster, synovial lactate measurement appears to be a good indicator, often studied, with good diagnostic performance but with considerable variability between the different studies [27–30]. The lactate/glucose ratio > 5 showed an excellent specificity of 98% but a low sensitivity of 52% [27]. The RESAS score, a composite score based on synovial lactate and glucose dosage, the presence of microcrystals, and synovial neutrophil count, with a threshold ≥ 4 to confirm SA, seems to be more suitable, with strong diagnostic performance (specificity 98%; PPV 77–90%; NPV 95–98%), but variable sensitivity between 56% in the calibration cohort, in contrast to 92% in the validation cohort [31]. These tests are slower than Synofast®, with results available after one hour, and require a laboratory.

In this study, Synofast® did not significantly change patient management and only prevented one out of the 29 initially planned hospitalisations. In the Synofresh study, management would have been changed based on the test result in 78% of cases but according to theoretical scenarios [32]. There are several possible explanations for this. First, the majority of our patients were already hospitalised at the time of enrolment, which may have affected management. Therefore, future studies should evaluate the Synofast® test in emergency consultations prior to potential hospital admission. Furthermore, apart from the risk of infection, factors such as age, underlying conditions, frailty and disability are the main determinants of patient referral [33]. The patients in our study are quite old and have a high rate of comorbidities. Arthritis, regardless of its origin, may reduce their autonomy and lead to hospitalisation. A third hypothesis is that this is a new device in its first real-world study and despite the available information on its theoretical performance, users may be reluctant to change their practice. To avoid such a potential bias, future studies should incorporate randomisation between different management strategies.

The main limitation of our study is the small number of patients involved, which restricts its statistical power and limits subgroup analysis. Although the tests were significant, confidence intervals are wide due to the small number of subjects, thus limiting its precision and power. This reduces the reliability of the sensitivity and specificity values and increases the risk of a type II error. Consequently, the results should be interpreted with caution and the true diagnostic performances may be lower or greater. This is partly due to the monocentric nature of the study and the low incidence of SA. Also, as this was a pilot study, no formal sample size calculation was performed, and patients were consecutively enrolled over a one-year period. This methodological approach may have reduced the statistical power and constitutes a limitation of the study.

In terms of strengths, this is the first prospective real-world study to evaluate Synofast® in patients with suspected SA. The test’s indication was respected, with enrolled patients having a true suspicion of SA, as reflected by the clinician’s average pre-test probability estimate of 44%, which incresed to 83% in the SA group. The test was find to be easy to use in routine practice, with no technical difficulties. Furthermore, the prevalence of SA in our study is consistent with literature data, supporting its external validity and its integration into routine clinical practice [8, 15, 25, 27, 31].

Conclusion

This single-centre, prospective, pilot study, conducted in real-life practice, suggests that the Synofast® test is an easy-to-use, fairly reliable method of excluding SA, with a negative predictive value of 93%. This is somewhat lower than in previous studies. However, it appears to be less effective in cases involving prior antibiotic treatment. A multicentre randomised study with a larger sample size is essential to determine its impact on patient management and its role in the diagnosis strategy for acute arthritis.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors contributions

M. Couderc and F. Badaire contributed equally to the conceptualization of the study and the drafting and revision of the manuscript. C. Lambert also participated in the conceptualization, data curation, validation and contributed to the writing. All other authors reviewed the manuscript and approved the final version for submission.

Data availability

The datasets generated and analyzed during this study are available from the corresponding author on reasonable request.

Declaration

Competing interests

The authors declare no competing.