The dynamics of procalcitonin in COVID-19 patients admitted to Intensive care unit - a multi-centre cohort study in the South West of England, UK.

Sir,

In this Journal, Seaton and colleagues recently demonstrated the importance and impact of antimicrobial stewardship during the current pandemic. Severe COVID-19 infections are characterised by a systemic inflammatory response, and frequently present with pyrexia, raised C-reactive protein (CRP), hypoxia and lung infiltrates. Clinicians have struggled to determine which COVID-19 patients have super-added bacterial infection requiring antibiotic treatment, leading to widespread antibiotic use.

Microbiological culture is a relatively insensitive technique, especially during antibiotic treatment. It can be difficult to distinguish infection and colonisation in non-sterile sites, and even in patients with sepsis only 30–50% will have a positive blood culture. We cannot therefore rely on positive microbiology alone as an indicator of bacterial infection.

Procalcitonin (PCT) is an inflammatory biomarker that rises in bacterial infection and falls in response to antibiotic treatment, and has greater sensitivity and specificity for bacterial infection than CRP. , PCT has been used to distinguish between influenza with and without secondary bacterial infection and is of potential value in identifying COVID-19 patients with genuine bacterial infection.

Previous studies have investigated the role of PCT in COVID-19 infection. Williams et al. described a retrospective analysis of PCT use in COVID-19 patients, concluding that PCT led to a reduction in antibiotic use without impacting on 28 day outcomes.

Van Berkel et al. measured PCT and CRP in intensive care unit (ICU) patients with COVID-19, diagnosed with secondary bacterial infection based on a positive culture and the opinion of two ICU physicians. They concluded that low PCT could be used to exclude secondary bacterial infection.

PCT has been identified as marker of poor prognosis in COVID-19 infection, and it is unclear if a raised PCT is part of the inflammatory syndrome associated with COVID-19 or primarily reflects bacterial co-infection requiring antibiotic treatment.

We hypothesise PCT raised as an innate part of COVID-19 infection would be unresponsive to antibiotics, while that due to bacterial co-infection would respond to appropriate antibiotic treatment. If PCT is low in many COVID-19 patients and responsive to antibiotic treatment in others, then PCT could provide a useful marker of super-added bacterial infection in COVID-19 and in conjunction with the overall clinical picture can guide antibiotic use.

We have undertaken a retrospective observational study describing the dynamics of PCT and CRP, including the response to antibiotic treatment, in adults with severe COVID-19 infection requiring ICU admission (n = 99) during the first wave of the pandemic. For comparison we selected two better-understood groups of patients from historical data, adult ICU patients with either bacteraemia representing proven bacterial infection (n = 113), or influenza representing viral infection at risk of super-added bacterial infection (n = 32). Microbiology, inflammatory markers, and antibiotic use, were recorded for the 3 cohorts for 14 days from the first positive blood culture or viral PCR test.

Bacterial co-infection rates in the COVID-19 and the influenza cohorts (7.1% and 18.7%, respectively) were similar to those found in other studies and co-infection rates for both viral infections are higher in ICU patients than in other hospitalised patients. ,

CRP was initially raised in the COVID-19 cohort and continued to rise during week 1, falling during week 2 (Fig. 1 ).

Fig. 1

CRP and PCT mean and interquartile range (IQR) by day for the 3 cohorts. BC=BSI cohort; COV=COVID-19 cohort; FLU=influenza cohort.

Elevated PCT in the first 48 h of admission was rare in COVID-19 patients. Where PCT was recorded it was <1.0 ng/L in 68.9% of COVID compared to 38% influenza patients.

In an attempt to produce an objective assessment of antibiotic response we have adopted the following definitions “a priori”. We have defined likely bacterial infection group as PCT>1.0 ng/l and have defined a response to antibiotic treatment as a 40% reduction from peak PCT by day 3, or a 60% reduction by day 4 or an 80% reduction by day 5 of treatment or a reduction to below 1.0 ng/l. Any PCT reductions up to 24 h after an antibiotic regime was stopped was included as part of the attributable response.

Patients with insufficient PCT data to determine response where excluded. The remainder were placed in 3 groups; group 1: PCT below 1.0 ng/L on days 0 to 13; group 2: PCT raised above 1.0 ng/L, but a response to antibiotic treatment was observed; group 3: PCT unresponsive to antibiotic treatment. The characteristics of the 3 groups are summarised in Table 1 .

Table 1

SD=standard deviation; p values relative to the COVID-19 cohort; Abx=antibiotic.

Low PCT group	COVID-19 Group 1	Influenza Group 1	BSI Group 1
Number (n)	27 (36.0%)	10 (43.4%)	6 (8.2%)
Age-years (SD)	56.2 (13.2)	55.5 (17.2) p = 0.89	62.8 (8.8) p = 0.25
Gender (n (%male))	20 (71.4%)	5 (50%) p = 0.16	4 (66.6%) p = 0.71
Days to ICU discharge (SD)	13.4 (6.1)	6 (2.5) p<0.001	7.5 (4.8) p<0.001
Mortality <14 days (n (%))	1 (3.6%)	2 (20%) p = 0.11	1 (16.6%) p = 0.23
Mortality <28 days (n (%))	4 (14.8%)	2 (20%) p = 0.70	1 (16.6%) p = 0.90
Co-infection (n (%))	2 (7.4%)	1 (10%) p = 0.79	2 (33.3%) p<0.078
Early infection (n (%))	3 (11.1%)	0 (0%) p = 0.27	0 (0%) p = 0.39
Late infection (n (%))	5 (18.5%)	0 (0%) p = 0.14	1 (16.7%) p = 0.91
PCT responsive Abx	COVID-19 Group 2	Influenza Group 2	BSI Group 2
Number (n)	22 (29.3%)	9 (39.1%)	56 (76.7%)
Age-years (SD)	59.6 (10.6)	48.4 (19.6) p = 0.048	63.4 (14.9) p = 0.28
Gender (n (%male))	15 (68.1%)	4 (44.4%) p = 0.22	35 (62.5%) p = 0.64
Days to ICU discharge (SD)	23.0 (15.7)	14.1 (8.8) p = 0.1	15.5 (14.2) p = 0.01
Mortality <14 days (n (%))	3 (13.6%)	2(22.2%) p = 0.55	5 (8.9%) p = 0.53
Mortality <28 days (n (%))	4 (18.1%)	2(22.2%) p = 0.79	8 (14.2%) p = 0.66
Co-infection (n (%))	0 (0%)	3 (33.3%) p = 0.004	21 (37.5%) p = 0.001
Early infection (n (%))	3 (13.6%)	0 (0%) p = 0.22	8 (14.3%) p = 0.43
Late infection (n (%))	4 (18.2%)	3 (33.3%) p = 0.36	16 (28.6%) p = 0.52
PCT not responsive Abx	COVID-19 Group 3	Influenza Group 3	BSI Group 3
Number (n)	26 (34.6%)	4 (17.4%)	11(15.1%)
Age-years (SD)	60.3 (10.4)	52.2 (15.3) p = 0.19	63.5 (13.5) p = 0.43
Gender (n (%male))	21 (80.7%)	2 (50%) p = 0.17	9 (81.8%) p = 0.94
Days to ICU discharge (SD)	18.3 (11.2)	35.7 (35.6) p = 0.40	10.8 (13.5) p = 0.16
Mortality <14 days (n (%))	6 (23%)	1 (25%) p = 0.99	5 (45.5%) p = 0.33
Mortality <28 days (n (%))	14 (53.8%)	1 (25%) p = 0.56	7 (63.6%) p = 0.85
Co-infection (n (%))	4 (15.4%)	1 (25%) p = 0.63	7 (63.6%) p = 0.003
Early infection (n (%))	4 (15.4%)	1 (25%) p = 0.63	1 (9.1%) p = 0.83
Late infection (n (%))	9 (34.6%)	0 (25%) p = 0.16	1 (9.1%) p = 0.11

Low PCT is expected in viral infection. In bacterial infection PCT is typically raised, with higher values seen in systemic compared to localised infection, and with more pathogenic organism.

In keeping with this only 8.2% of the BSI cohort had low PCT (group 1) while 76.7% showed a good PCT response to antibiotics (group 2), and 15.1% a poor response to antibiotics (group 3), with associated high mortality in this group. In contrast 43% of influenza patients, and 36% COVID-19 patients had a low PCT from admission to day 13 (group 1), with 39.1% of influenza patients, and 29.3% COVID-19 patients having a raised PCT that responded rapidly to antibiotics consistent with super-added bacterial infection (group2).

In all cohorts a proportion of patients showed a poor PCT response to antibiotic treatment (BSI=15.1%, influenza=17.4% and COVID-19=34.6%). The higher proportion of COVID-19 patients in groups 3 is likely to be due to late infection with rising CRP and PCT, and positive microbiology common after day 6. A partial response to antibiotic treatment by day 14 was also seen in this group.

In summary, the dynamics of PCT in the COVID-19 cohort were similar to that of the influenza cohort with 65.4% and 82.6% respectively having low PCT, or PCT that responded rapidly to antibiotics. The influenza cohort had higher rates of co-infection while the COVID-19 cohort had higher rates of late hospital acquired infection. The dynamics of PCT in COVID-19 patients are consistent with a response to secondary bacterial infection and are not consistent with an inflammatory response to COVID-19 alone. In contrast to CRP (which was raised and unresponsive to antibiotics during week 1), PCT appears to be a useful biomarker in identifying COVID-19 patients with super-added bacterial infection, and supports antibiotic treatment in COVID-19 patients with a significantly raised PCT including those without positive microbiological cultures.