Procalcitonin as a Diagnostic and Monitoring Tool for Bacteraemia in Patients on Haemodialysis: A Systematic Review

Abstract

Background: Infections are a major cause of mortality in haemodialysis patients. The increasing antimicrobial resistance globally further exacerbates this concern. Procalcitonin has shown potential in aiding antimicrobial stewardship and reducing the mortality and morbidity associated with infections. This systematic review aims to synthesise the existing literature on the utility of procalcitonin as a diagnostic and monitoring tool for haemodialysis patients with suspected bacteraemia. Methods: Multiple electronic databases (EMBASE, MEDLINE, and the Cochrane Library) were systematically searched to identify primary studies evaluating procalcitonin use in haemodialysis patients with suspected bacteraemia. Using a narrative synthesis approach, along with other quality assessment tools, recommendations regarding procalcitonin usage in dialysis were formulated. Results: Eleven studies were identified. The review proposes a procalcitonin-guided antibiotic protocol aimed at facilitating antibiotic use in haemodialysis patients with suspected bacteraemia. Caution is, however, advised against relying solely on procalcitonin for diagnosis, emphasising the integration of procalcitonin with other clinical and biomarkers of infection. Conclusions: Procalcitonin shows promise as a valuable diagnostic and monitoring tool for suspected bacteraemia in haemodialysis patients. While caution is advised against relying solely on PCT for diagnosis, its integration with other clinical indicators can enhance infection management. To fully establish its clinical utility, more research is needed.

1. Introduction

Renal replacement therapy (RRT) for end-stage renal disease (ESRD) patients increases infection risk, particularly through vascular access sites like arteriovenous fistulas, grafts, and central venous catheters, serving as prominent conduits for the introduction of these infections, with bacterial infections accounting for up to 70% [1]. Prompt initiation of appropriate antibiotics is critical as each hour of delay is associated with an increased mortality risk in patients with septic shock and organ dysfunction, which are well-recognised complications of bacteraemia and sepsis [2,3].

Traditional infection markers like C-reactive protein (CRP), erythrocyte sedimentary rate (ESR), and ferritin are nonspecific and vary in sensitivity. Blood cultures, the gold standard for diagnosis, are limited by delays and collection accuracy [4,5]. Thus, there is a need for a sensitive, specific biomarker with rapid turnaround times.

Procalcitonin (PCT), produced by the thyroid and adipose tissues, is released during bacterial infections by neuroendocrine cells in the gut and lungs. Detectable within 2–4 h and peaking around 6 h post-infection, PCT levels remain stable for up to 24 h [6,7]. Multiple studies indicate PCT’s utility in early bacterial infection detection, antibiotic therapy guidance, and infection course prediction [8,9,10,11].

In chronic kidney disease, it is postulated that diminished renal function leads to an increase in the serum levels of proinflammatory metabolites (in association with uraemia), triggering immune system activation and inflammatory processes, consequently leading to an increased PCT release. In HD, the invasiveness of the dialysis process, presence of commensals associated with vascular access use, and immune reaction to the dialyser can trigger an inflammatory reaction, leading to higher levels of PCT [12,13,14].

It is also hypothesised that some of the serum PCT is removed by a combination of diffusive and convective transport processes across the dialysis membrane. Interestingly, this elimination process of PCT during haemodialysis occurs concurrently with the clearance of other substances, such as β-microglobulin and various other low-molecular-weight proteins. These substances, including PCT, can be transported across the dialysis membrane due to their small molecular size, allowing them to be removed from the bloodstream during the dialysis procedure [15,16]. These mechanisms can potentially create difficulties when using PCT as a clinical diagnostic or monitoring tool in ESRD patients on HD due to the inconsistency of results.

The use of PCT for diagnosing and treating bacterial infections in haemodialysis (HD) patients is not fully developed. Although many studies have explored PCT’s role in diagnosing and managing infections, few systematic reviews focus on its application for catheter-associated and non-catheter-associated bacteraemia in HD and non-HD settings [7,17]. Our review assessed PCT as a diagnostic marker and treatment guide for HD-related bacteraemia. We aimed to create a framework to assist clinicians in using PCT to determine when antibiotic therapy is necessary for HD patients.

2. Materials and Methods

This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses 2020 (PRISMA 2020) statement [18]. The study protocol was developed following the PRISMA-P guidelines [19] and registered on PROSPERO (Registration Number: CRD42023422323).

2.1. Literature Search

The literature search was conducted across the following databases: EMBASE, MEDLINE (via Ovid), and the Cochrane Library and was aimed at identifying studies that evaluated the use of procalcitonin in haemodialysis patients in multiple clinical settings. Using a predefined inclusion and exclusion criteria, the studies included met the following conditions: (1) published from the inception of the selected databases up to March 2023, (2) written in the English language, (3) observational or interventional studies, including randomised and non-randomised controlled trials, cohort studies, cross-sectional, and case-control studies, that evaluated the use of procalcitonin as a primary or secondary outcome in haemodialysis patients, (4) involved participants aged over 18 years, and (5) provided access to the full text and complete study data. Studies were excluded if they were (1) case reports, case series, reviews, or other forms of secondary research, or (2) published in a language other than English. Supplementary approaches to identify relevant studies were employed, such as a partial grey literature search using Google Scholar and searching the reference lists of included studies. The following keywords and Medical Subject Heading (MeSH) terms were combined to create a search algorithm and used to search the prementioned databases: (Procalcitonin OR PCT) AND (“Renal Dialysis” OR Haemodialysis OR Hemodialysis OR “Renal replacement therapy”) AND (Bacteraemia OR “Bacterial Infection” OR Sepsis OR Infections). The search protocol and results from the corresponding databases can be found in the supplementary data (Appendix A).

2.2. Study Selection and Eligibility Criteria

Following an extensive database search, duplicates were removed using the Mendeley reference manager. The title and abstract content of the remaining articles were independently assessed in an unblinded standardised manner by two reviewers to evaluate their relevance to the chosen topic and their suitability based on the inclusion and exclusion criteria. Following the screening process, unrelated and unsuitable articles were removed. Any disagreement in the selected articles between the reviewers was discussed until an agreement was reached.

A final review of the selected articles, by both reviewers, was conducted to assess their eligibility for use. Research involving PCT use in the diagnosis or monitoring of bacteraemia in haemodialysis patients was preferentially selected.

2.3. Data Extraction and Quality of Studies

Data were extracted by two independent reviewers using a refined data extraction sheet based on the Cochrane Consumers and Communication review group’s template [20]. One reviewer extracted the data while another assessed it independently. Disagreements were resolved through consensus.

To determine the methodological quality of non-randomised trials, the ROBIN-I tool assessed potential biases [21]. The Newcastle-Ottawa Scale (NOS) was employed for observational studies, with scores categorising quality as high (7–9), moderate (4–6), or low (0–3). A modified Newcastle-Ottawa Scale was used for cross-sectional studies, with quality categorised from high (6–7) to low (≤3) [22,23] (Appendix C and Appendix D).

Two independent reviewers evaluated the studies, resolving significant disagreements through discussion. To standardise the process and remove bias when using the NOS for cohort studies, survey questions were developed based on the original NOS covering each domain (Appendix B).

2.4. Data Analysis and Presentation

Given the presumed heterogeneity and limited number of relevant studies, a narrative synthesis approach was selected to address the study’s complex questions and avoid issues with calculating standardised effect sizes from unavailable and non-comparable data points. The initial step involved an inductive thematic analysis with the goal of systematically identifying, tabulating, and narratively describing recurring or pertinent themes derived from the data to address the review questions. By grouping and organising the studies into meaningful categories and mapping out key consistent trends, this allowed for the exploration of relationships within and between the included studies, creating a structured framework for the synthesis and reporting of the results. The data were presented graphically in tables and narrative texts, including important study characteristics and evidence that supported the results and degree of heterogeneity [24,25,26].

2.5. Analysis of Heterogeneity and Certainty of Evidence

Regarding heterogeneity, due to the limitations in using statistical assessments of measurements, heterogeneity was informally assessed by addressing the clinical and methodological diversity between the studies by assessing the variability in participants, intervention, outcome, study design, and risk of bias. Findings from this were summarised using the terms “high” or “low” heterogeneity [25]. After extracting and synthesising data from the included studies, the relevant evidence was combined to form good practice recommendations (GPR) pertaining to the use of procalcitonin in patients undergoing haemodialysis. The criteria for defining these GPRs required each statement to be a clear and actionable clinical suggestion specifically related to the clinical use of procalcitonin in haemodialysis patients and be supported by a rationale.

A modified GRADE system based on a British HIV association archetype [26] was used to evaluate the evidence’s certainty, strength, and quality for deriving GPRs. This modification addressed anticipated challenges with assessing specific GRADE domains due to different study methods and types. A copy of the modified GRADE scale can be found in the supplementary data (Appendix E).

2.6. Ethical Considerations

Informed consent from participants was not required due to the use of secondary data from other sources. The study protocol was reviewed and approved by the University of South Wales ethics committee.

3. Results

3.1. Study Selection

A total of 569 records were identified, with 432 remaining after removing duplicates and non-English articles. Following screening, 366 were excluded for being case reports/series, involving paediatric populations, being review papers, or being non-relevant. An additional 56 were excluded due to inability to retrieve the full text or containing non-useful data. Additional searches using Google Scholar and the reference lists identified eight more records, but only one met the inclusion criteria. In total, 11 articles were included, as shown in Figure 1.

3.2. Characteristics of Included Studies

Of the included studies, ten were observational, including cohort, case-control, and cross-sectional studies. The remaining study was a non-randomised controlled trial (NRCT). This diversity in design provided varied perspectives but potentially reduced the generalisation of results. Six studies were retrospective, while five were prospective. The total cohort included 3373 participants aged 18 and above, primarily diagnosed with chronic or acute kidney injury. These individuals were from various clinical environments, such as intensive care units, general wards, and dialysis centres. There was an active step to try and exclude data from patients not on haemodialysis.

3.3. Summary of Findings

After conducting a preliminary synthesis and in-depth analysis of the data extracted from the included studies, a framework was developed based on thematic deduction derived from each individual study. The resulting framework comprised two major themes: the use of procalcitonin as a means to identify haemodialysis (HD) patients with bacteraemia and the utility of PCT as a monitoring and prognostic tool for HD patients with bacteraemia. The findings of the review demonstrated that PCT levels were significantly elevated in HD patients with bacteraemia and that, in a suitable cohort of patients, PCT can be used as an accurate indicator of bacteraemia in HD patients.

A cutoff PCT level of >1.5 ng/mL exhibited a sensitivity level of 89% and 81% specificity in detecting severe bacteraemia [27]. A similar PCT cutoff of 1.65 ng/mL was associated with a sensitivity of 80.8% in one study [28] while a PCT cutoff level of 15.5 ng/mL possessed a sensitivity of 94% for detecting bacteraemia with 100% specificity in another study [29]. Conversely, PCT levels below 1 ng/mL were less likely to be associated with bacteraemia [27,30,31].

For haemodialysis patients, the level of procalcitonin was found to be elevated, even in the absence of infection. Our observations indicate that PCT levels equal to or below the range: 0.50–0.8 ng/mL could effectively exclude the presence of bacteraemia [27,30,32]. A single study within the review linked a PCT level of less than or equal to 1.1 ng/mL to the absence of bacteraemia [29].

Our review revealed contradictory evidence in relation to the identification of Gram-positive and Gram-negative bacteraemia. A single study [33] within the review reported a greater median procalcitonin level in haemodialysis patients with Gram-positive bacteraemia compared to Gram-negative bacteraemia, specifically 61.2 ng/mL (25–75th centile: 0.9–145) versus 21.2 ng/mL (25–75th centile: 4.3–114), while another included study reported a higher mean level of PCT in patients with Gram-negative bacteraemia, as opposed to Gram-positive bacteraemia: 21.9 ± 42.2 vs. 13.7 ± 24.8. However, this result was not statistically significant [34]. Patients with Gram-negative bacteraemia had a higher proportion of the complications ascribed to bacteraemia.

In the comparison of diagnostic markers, it was found that PCT outperformed C-reactive protein and white blood cell count in diagnosing bacteraemia with or without sepsis [28,30,31]. However, one study disagreed with the premise that PCT was a better diagnostic biomarker for detecting bacteraemia than CRP in HD patients, despite the higher cost. However, this was a cross-sectional study with a retrospective design and was at high risk of selection bias [32].

The synthesis of the data also revealed that a dynamic approach involving the measurement of PCT in HD patients with bacteraemia yielded better prognostic information compared to a single measurement of procalcitonin, regardless of the absolute level. A substantial reduction in PCT levels (>63%) after 72 h of treatment was associated with lower mortality rates [35]. Furthermore, it was observed that higher PCT levels were associated with an increased risk of bacterial sepsis-induced complications, such as bacterial metastasis, septic shock necessitating intensive care admission, and death. Notably, PCT levels ranging between 2 and 4 ng/mL were linked to a higher risk of hospitalisation and death [27,33,34,35].

Regarding monitoring treatment with PCT, a corresponding rise and fall in procalcitonin levels was observed alongside changes in other inflammatory markers and clinically observed signs of infection [33,35,36,37]. In some of the studies, there was no significant correlation between procalcitonin and other inflammatory markers [29]. Pre-dialysis measurements of PCT were found to be more reliable in terms of diagnostic or prognostic value compared to the measurements taken during or immediately after dialysis, as dialysis had an impact on the clearance of procalcitonin [27,34].

The studies included implemented exclusion criteria targeting certain medical conditions based on their potential to introduce confounding effects on the serum procalcitonin levels. These identified medical conditions comprise severe burns, recent major surgery, malignancy, immunodeficiencies, patients receiving immunosuppressants, pregnancy, and acute pancreatitis. Consequently, these conditions were intentionally excluded from the study population to ensure a more accurate and unbiased assessment of the relationship between procalcitonin levels and the investigated factors.

Based on the findings, it was discouraged to rely solely on procalcitonin as a diagnostic tool due to some overlap in the levels observed between patients with bacteraemia and those without. Instead, a comprehensive approach integrating multiple clinical and laboratory parameters was recommended for the accurate diagnosis and management of bacteraemia in HD patients [31,32,34].

Table 1. Summary of the characteristics and findings of the included studies.

Author, Year	Study Design	Patient Group/Clinical Setting	Primary Endpoint of Study	How PCT Was Used	Summary of Findings
Agrawal, 2019 [33]	retrospective cohort study	169 HD patients (>18 yrs) with non-tunnel haemodialysis catheters 52 had CASBI- taken from catheter hub and blood culture	Incidence rate, survival, risk factors and complications of CABSI	PCT levels in Gram-positive vs. Gram-negative infections vs. CoNS	-High procalcitonin corresponds to higher % of neutrophil left shift and differentiates Gram-positive and Gram-negative infections from those by CoNS–61.2 (0.9–145) vs. 21.2 (4.3–114) vs. 1.9 (0.9–3.4) < 0.001 [median (25–75th percentile)] -The vast majority of individuals (75%) with catheter-associated bacteraemia in the study had PCT levels > 0.9 ng/mL. -Higher PCT level > 4 ng/mL was associated with higher rates of hospitalisation, ITU admissions, metastatic complications, and death
Demir, 2018 [30]	retrospective cohort study	1110-HD Patients (>18 yrs) with infections (including septic patients with positive blood culture) and no infection. 113 had sepsis + bacteraemia.	PCT and CRP cutoff values that determine presence of bacterial infection	-PCT level in patient with documented and non-documented bacterial infection (including bacteraemia + sepsis). -Monitoring response to treatment.	-Mean PCT with no infection: 0.51 ± 0.96 ng/mL (p = 0.01). -Mean PCT with bacteraemia + sepsis 33.9 ± 35.9 (p = 0.001); patients with sepsis + bacteraemia had the highest mean PCT level. -Notable decline in PCT level in HD with treatment, which corresponded to a decline in CRP -Mean PCT following treatment 0.5 ± 0.59 (p = 0.001) similar to Mean PCT without infection (mean value following Rx for sepsis = 0.62 ± 0.47) -CRP ≥ 100 mg/dL and PCT ≥ 5 ng/mL = 100% PPV and 94% NPV for showing the presence of sepsis + bacteraemia in HD patients.
Hamada and Gamal, 2017 [29]	Prospective cohort	31 HD patients with signs of suspected CRBSI	PCT use a diagnostic marker for CRBSI	PCT levels in culture-positive vs. culture-negative HD patients	-Small number of patients in study -PCT level was significantly higher in the positive blood culture group 40.0 ± 21.9 ng/mL (95% CI 28.4–54.8) than the culture-negative grp, 1.1 ± 1 (95% CI 0.54–1.8)—p<0.001 -PCT showed a positive correlation with CRBSI, while CRP did not. -At PCT cut off level of 15.5 ng/mL, sensitivity = 94%, specificity = 100% for bacteraemia (p < 0.001).
Herget-Rosenthal, 2001 [27]	Prospective cohort	68 (>18 yrs) ESRD patients (n = 48) and AKI patients (n = 20) on high-flux and low-flux membranes HD- 12/68 sepsis	PCT as a diagnostic marker of severe infections/sepsis for ESRD/AKI patient on HD when compared to CRP and WBC	PCT levels measured before starting HD, during and after HD in patients with and without severe infections/sepsis	-PCT (no infection)-0.8 ng/mL (0.3–1.4) -PCT (severe infection/sepsis)-5.9 ng/mL (2.5–11.5) (p < 0.01). -PCT cutoff level to discriminate between severe bacterial infections/sepsis and non-infection: 1.5 ng/mL (sensitivity/specificity/PPV/NPV/accuracy 89/81/84/87/85% respectively), better than crp: 5 mg/dL sensitivity/specificity/PPV/NPV/accuracy 89/48/68/78/71%, respectively. -In High flux groups, PCT reduced by 83 ± 25% (lowest value) after 4 h into HD—reduced PCT sensitivity, PPV, accuracy-only return to normal level at 48 h—check PCT before high flux HD
Kim, 2022 [35]	Retrospective cohort	SIAKI on CRRT-649 patients	PCT use as a prognostic marker to predict survival and recovery from dialysis	PCT at the start and 72 h of CRRT. %PCT decrease was calculated	-Non-survivors had a higher PCT level than survivors at initiation and 72 h ± 12 h post CRRT initiation: 14.2 ng/mL [7.3–22.8] vs. 10.4 ng/mL [4.8–17.5], p < 0.001 and 13.5 ng/mL [6.4–24.9] vs. 3.6 ng/mL [0.8–9.8], p < 0.001 -Survivors had a higher %PCT reduction 63.0% [4.0–90.0] vs. −19.0% [−46.0–15.3], p < 0.001 (% PCT decrease = [(PCT0-PCT72 h)/PCT0] × 100.) -A dynamic approach in assessing PCT over 72 h is better at predicting survival, progression of sepsis, and reflects the effectiveness of sepsis management, than a single PCT level, regardless of the initial severity assessed by clinical scores
Yunus, 2018 [34]	Retrospective cohort	364 Patients in ICU (20 yrs–79 yrs), with sepsis. 16 CKD/HD patients	Determine the correlation between initial PCT values and mortality, illness severity and prognosis; also, association with site, type and extent of infection	PCT level measured in different groups with different clinical outcomes.	-Small positive correlation between PCT and SOFA/end organ dysfunction -Patients with higher mean PCT > 19.3 ± 38.3 [median PCT > 2.0 (0.1–173.1)] were slightly more likely to die on ward or ICU, when compared to those who survived and were subsequently discharged. However, there was a lack of significant correlation between initial PCT level and clinical outcome, consideration for a dynamic approach. -Patients with mean PCT < 7.0 ± 17.0 [median PCT < 1.5 (0.1–131)] were less likely to be transferred to ICU or die. Those with mean PCT > 19.3 ± 38.3 (median ≥ 2) were more likely to have a higher mortality in ITU. -Mean PCT associated with bacteraemia: 18.3 ± 33.7 vs. non-bacteraemia infections 12.8 ± 31.0 [median PCT 4.7 (0.1–173.1) vs. 1.4 (0.1–252.5)] p = 0.04 -There was a statistically significant association between PCT and severity of sepsis: mean/median PCT in sepsis = 9.6 ± 22.7/1.4 (0.1–200); PCT in septic shock = 32.7 ± 52.2/8.1 (0.1–252) -Gram positive vs. Gram negative vs. viral mean PCT: 13.7 ± 24.8 vs. 21.9 ± 42.2 vs. 3.7 ± 3.3—no significant correlation between PCT and Gram staining
Zhai, 2021 [37]	Non-randomised control trial	53 ICU patients (aged 18–65) with sepsis. Control group received conventional CRRT, observed group treated with oxiris-endotoxin adsorption membrane for CRRT. Other treatments remained the same	Application value of oxiris–endotoxin adsorption technology in patients with sepsis	PCT was measured before and after treatment in both groups	A noticeable decrease in PCT and both the control group and observed group following treatment, corresponding to a reduction in other biochemical and physiological/clinical markers: lactate, endotoxin, SOFA score, and heart rate. Observational group: PCT level 41.62 ± 13.98–9.87 ± 2.15 (before treatment-after treatment) Treatment group: PCT level 40.27 ± 15.20–15.64 ± 4.29 (before treatment-after treatment)
Jiang, 2019 [28]	Retrospective cohort	373, CKD 5 Patients (20–93 yrs.) on HD, PD and not on dialysis. HD patients made up 92.3% of the sepsis group.	To estimate the predictive value of PCT in diagnosing sepsis in CKD stage 5 patients on and off dialysis	PCT was measured in different groups of patients: non-infection, local infection, and sepsis groups.	-AUC of PCT for predicting sepsis in CKD5 ± HD: 0.838 (95% CI 0.797–0.874) p < 0.05—CRP and WBC were unable to predict sepsis. -Specificity and sensitivity of PCT at 1.650 ng/mL for predicting bacterial sepsis: 80.8% and 75.2%—independent risk factor of sepsis (OR = 6.926, p = 0.002)
Schneider, 2020 [31]	Prospective cohort	54 Chronic HD patients (>18 yrs) with fever: temp >38 in hospital	Diagnostic potential of PCT in HD patients with infections	Pct was measured in patients with presumed infection and those without an infection	-For patients clinically deemed to have an infection, 68% had a positive blood culture test. Mean PCT 4.3 ± 5.5. -77% of patients with infection had PCT levels > 1 (p = 0.02). -For patients with positive blood culture, 86% had PCT > 1 -PCT was better at diagnosing bacterial infections than CRP. -The sole use of PCT as a diagnostic tool is not advised due to significant overlap between patients with and without bacterial infections
Navas, 2018 [36]	Prospective case control	18 ICU patients with acute (<48 h) septic shock (suspected bloodstream infections + vasoactive drugs) receiving haemofiltration + broad-spectrum antibiotics (control) Haemofiltration + haemoperfusion with polymyxin B + broad spectrum antibiotics (case)	The clinical and biological effects of hemoperfusion with polymyxin B in patients with endotoxic shock and AKI	PCT was used to monitor the impact of treatment in both the case and control groups	-A consistent decrease in PCT was noted within the 5-day course of treatment in both groups; this corresponded to a similar drop in CRP, lactate, and endotoxins noticed. PCT reduction (mean) 56.7 ng/mL–21.3 ng/mL in 5 days (group 1) 56.4 ng/mL–17 ng/mL (group 2)
Park, 2014 [32]	Retrospective cross sectional-493	493 patients (>18 yr) with suspected infections, including bacteraemia	Diagnostic value of PCT compared to CRP	PCT levels were measured and grouped by renal function and infection status	-PCT was noticeably higher in HD group with suspected infection/bacteraemia when compared to HD group without infection: median values measured 7.37 vs. 0.55 ng/mL (this difference corresponded to changes in CRP). -The best cutoff values of PCT for diagnosing infection were 1.1 ng/mL for patients with impaired renal function. -PCT is not a better diagnostic marker of infection, than CRP, in renal dysfunction

Note: HD, haemodialysis; CABSI, catheter-associated bloodstream infection; CRBSI, catheter-related bloodstream infection; ITU, intensive therapy unit; ICU, intensive care unit; CKD, chronic kidney disease; SIAKI, sepsis-induced acute kidney injury; ESRD, end-stage renal disease; PCT, procalcitonin; CRP, C-reactive protein; CoNS, coagulase-negative staphylococci; CRRT, continuous renal replacement therapy; PPV, positive predictive value; SOFA, sequential organ failure assessment; WBC, white blood cell count.

summarises the characteristics and findings from the included studies.

3.4. Quality Assessment

The observational studies in this systematic review were of moderate-to-high quality based on the Newcastle-Ottawa Scale, as shown in

Table 2. Quality assessment data using Newcastle-Ottawa (NOS) quality assessment tool for cohort and case control studies, and modified Newcastle-Ottawa tool (adapted from the original NOS tool) for cross-sectional studies.

Study	Selection ****	Comparability **	Outcome ***	Overall
Agrawal, 2019 [33]	2	2	1	5
Demir, 2018 [30]	3	1	2	6
Hamada and Gamal, 2017 [29]	2	1	2	5
Herget-Rosenthal, 2001 [27]	2	2	3	7
Kim, 2022 [35]	3	2	3	8
Yunus, 2018 [34]	3	2	2	7
Jiang, 2019 [28]	3	2	2	7
Schneider, 2020 [31]	3	1	2	6
Navas, 2018 [36]	2	1	3	6
Park, 2014 [32]	2	1	3	6

**** Selection—total of 4 points. ** Comparability—Total of 2 Points. *** Outcome—Total of 3 Points (for NOS Cross sectional study—total of 2 points/stars).

. The main limitations were non-representative samples, inadequate selection controls, insufficient control for confounding variables, and lack of rigorous follow-up or comprehensive outcome assessment.

The NRCT risk of bias assessment (

Table 3. Risk Of Bias in Non-randomised studies—of interventions (ROBINS-I) assessment tool—for Zhai 2021 [37].

BIAS DOMAINS	ROB Assessment
Bias due to confounding	3
Bias in selection of participants	2
Bias in classification of intervention	1
Bias due to deviation from intended intervention	2
Bias due to missing data	1
Bias in measurement of outcome	4
Bias in selection of the reported results	2
Overall bias	3

Risk of bias assessment: 1—low; 2—Moderate; 3—Serious; 4—Critical.

) indicated a high potential for bias due to confounding and outcome measurement, with scores of three and four, respectively. The overall bias score of three reflects a moderate-to-high risk of bias. These findings necessitate cautious interpretation of the findings and highlight the need for future research to address these biases to enhance study quality.

3.5. Heterogeneity

The systematic review revealed significant heterogeneity among the included observational studies and non-randomised control trials in terms of participants, exposure, outcomes, and methodology. Participant differences such as diverse ages, genders, and comorbidities, along with varying types of renal failure, introduced confounding factors and limit generalisability. There was significant variation in exposures and interventions, including different diagnostic tests, treatments, and the types of organisms causing bacteraemia. The outcome measures varied, with studies focusing on diagnostic potential, prognostic indicators, or treatment efficacy. Methodological differences, such as study design, sample size, and statistical analyses, further contributed to heterogeneity, affecting the reliability of the results. Thus, the findings should be interpreted with caution, considering the context and specific study characteristics.

3.6. Grading Good Practice Recommendations

The objective of this review was to develop a framework for clinicians on using procalcitonin (PCT) for haemodialysis patients suspected of having bacteraemia. We have summarised the recommendations, with an attached modified GRADE score, below:

(a)

Check procalcitonin levels when there is clinical and biochemical evidence of bacteraemia. Our review found a correlation between procalcitonin and bacteraemia, both related to access and non-access related bacteraemia (1B).

(b)

In cases of uncertainty regarding a bacterial infection, it is advisable to consider initiating broad-spectrum antibiotics if the procalcitonin level is greater than or equal to 1.5 ng/mL and there are clinical or other biochemical signs of bacterial infections. Our review indicated that a procalcitonin cutoff value of 1.5 ng/mL had a sensitivity of 89% and an accuracy of 85%. A recent meta-analysis also supports this conclusion. Antibiotics can be started with a procalcitonin range greater than 0.5–1 ng/mL if there is a strong clinical suspicion of a possible bacterial infection (1B).

(c)

We recommend against using procalcitonin alone for diagnosing bacteraemia due to significant overlap between patients with and without bacteraemia (1B).

(d)

Dynamic measurements of procalcitonin are more useful for monitoring disease progression or recovery. We suggest measuring (PreHD) procalcitonin levels every 72 h, and a reduction of more than 60% in procalcitonin level indicates recovery (1C).

(e)

We recommend discontinuing antibiotics when the procalcitonin level is below 0.5 ng/mL and there is resolution of clinical symptoms (1B).

4. Discussion

We hypothesised that procalcitonin (PCT) is a valuable diagnostic and monitoring tool for haemodialysis patients with bacteraemia. The review consistently found elevated PCT levels in haemodialysis patients with bacteraemia. A PCT cutoff of 1.5 ng/mL effectively distinguished HD patients with severe bacteraemia, demonstrating a sensitivity and accuracy exceeding 80%. PCT was found to be more specific than C-reactive protein in this regard, while maintaining a comparable level of sensitivity to bacteraemia.

A meta-analysis [38] highlighted that raising the PCT cutoff from 0.68–0.85 ng/mL to 1.5–15 ng/mL decreased sensitivity from 96% to 86% but increased specificity from 77% to 89%. Our review supports a PCT cutoff of 1.5 ng/mL. While this meta-analysis focused primarily on diagnostic accuracy, our review takes a more clinically applied perspective by exploring how procalcitonin can be used to guide antimicrobial decision-making in practice.

Multiple studies found that procalcitonin levels below 1 ng/mL in HD patients usually indicate the absence of bacteraemia, with a reliable cutoff range of 0.5–0.8 ng/mL [27,30,31,32]. One excluded study agreed with a cutoff level of <0.5 ng/mL [16]. For access-related bacteraemia (ARB), PCT levels ranged from 0.9–1.1 ng/mL, with higher levels observed in more severe infections. The review highlighted that relying solely on PCT as a diagnostic marker is insufficient due to overlapping levels in patients with and without bacteraemia. Combining PCT with CRP improves diagnostic accuracy, as shown by one of the included studies with a combined cutoff of CRP > 100 mg/dl and PCT > 5 ng/mL, yielding a positive predictive value of 100% and a negative predictive value of 94% [39].

The ability of procalcitonin to distinguish between Gram-positive and Gram-negative bacteraemia yielded conflicting results. Agrawal et al. [35] found higher median PCT levels in HD patients with Gram-positive bacteraemia, while Yunus et al. [34] reported higher but not clinically significant PCT levels in Gram-negative bacteraemia. Despite this, Gram-negative bacteraemia consistently had worse outcomes. Li et al. [40] identified a PCT threshold of 2.44 ng/mL to differentiate between Gram-negative and Gram-positive bacteraemia, with a specificity of 77.4% and a sensitivity of 68.4%. This suggests that PCT could aid in detecting Gram-negative bacteraemia and guiding antibiotic therapy, but further research is needed to refine these thresholds for HD patients.

Our review found that PCT levels above 2–4 ng/mL correlated with higher risks of hospitalisation, bacterial complications, multi-organ failure, and death. Patients with PCT levels below 1.5 ng/mL had lower chances of needing intensive care or experiencing fatal outcomes.

Procalcitonin is a valuable predictor of infection severity and outcomes. A 2008 study showed that PCT was superior to CRP and white blood cell count in predicting survival and mortality, and it was as effective as CRB65 (confusion, respiratory rate > 30, hypotension, age > 65) for identifying low-risk pneumonia patients [41]. Similarly, a 2015 systematic review demonstrated that higher PCT levels and reduced PCT clearance were associated with increased all-cause mortality in sepsis [40]. A 2018 meta-analysis found that PCT-guided antibiotic treatment in intensive care significantly reduced mortality rates and treatment durations compared to control groups, with consistent benefits across various sepsis severity subgroups [42].

This review highlighted the connection between changes in PCT levels and other biological and clinical infection indicators. Increased or decreased PCT levels generally corresponded with similar changes in CRP, white blood cell count, lactate, and clinical markers like the Sequential Organ Failure Assessment (SOFA) score. Other studies have provided evidence of this correlation [41,42].

A single PCT level showed a weak correlation with survival, regardless of disease severity or initial PCT level [34]. A more accurate survival predictor was the percentage of PCT clearance after 72 h of treatment. Survivors showed a reduction in PCT levels exceeding 63% after 72 h of treatment initiation. Adopting a dynamic approach to monitor disease progression and treatment efficacy using PCT could guide therapeutic decision-making, but further high-level evidence is needed to support this approach [35].

Procalcitonin levels should be measured before a haemodialysis session to ensure optimal diagnostic and prognostic accuracy, as dialysis can reduce circulating PCT levels and compromise the reliability of the result [27].

Based on the synthesised recommendations from this review, a hypothetical procalcitonin-guided antibiotic framework was developed for haemodialysis patients presenting with symptoms suggestive of bacteraemia. This proposed approach is intended to support clinical decision-making around antibiotic use by incorporating procalcitonin levels. Presented as a flow chart in Figure 2, it outlines when to consider starting, stopping, re-evaluating, or escalating antibiotic therapy. It is important to note that this framework has not been clinically validated and should be viewed as a hypothesis generated from the current evidence base.

An important clinical trial evaluated procalcitonin-guided antibiotic therapy in ICU patients, using procalcitonin levels as the primary determinant for adjusting the antibiotic regimen. The study found no clinical benefit and, in fact, reported increased mortality and greater antibiotic use in the procalcitonin group [2]. In contrast, our protocol uses procalcitonin as a supportive adjunct alongside clinical judgment and other lab tests. It incorporates a 60% reduction in procalcitonin after 72 h as an indicator of treatment response and highlights the need for reassessment before an escalation of antibiotics if there is no clinical improvement. Unlike the critically ill ICU population studied in the trial, where factors such as fungal infections may interfere with procalcitonin accuracy, our protocol is designed for non-ICU inpatients, offering a more targeted and context-appropriate application. Further rigorous evaluation is crucial to assess its effectiveness.

This review synthesised the existing data on procalcitonin use in haemodialysis (HD) patients with bacteraemia using a rigorous systematic approach to develop evidence-based guidance. However, our study has several limitations. First, our search protocol and criteria might have excluded some relevant studies. Also, our inability to access the full papers of relevant studies may have limited our evaluation due to their exclusion, especially considering that most included studies were observational and susceptible to biases that could compromise our findings [43,44]. The lack of randomised controlled trials prevented us from establishing causal relationships. Finally, the significant heterogeneity among studies in design, methods, participant characteristics, and outcomes made comparisons difficult, leading to a qualitative, subjective approach that may not have fully addressed this variability and could introduce reviewer biases [45,46]. Unmeasured confounding factors, such as other treatments, fluctuations in procalcitonin levels due to timing of dialysis, chronic kidney disease, and other factors, for which no stratification was carried out, also increased the bias risk. These limitations highlight the need for more rigorous research, including well-designed RCTs, to better investigate procalcitonin use in HD patients with bacteraemia.

5. Conclusions

This review underscores the potential of PCT as a diagnostic and monitoring tool in haemodialysis patients with bacteraemia. Despite current limitations, the consistent association between elevated PCT levels and bacterial infection supports its clinical utility. Addressing knowledge gaps through well-designed RCTs, cost-effectiveness analyses, and clinical impact assessments will optimise PCT’s use in diagnosing and monitoring bacteraemia in haemodialysis patients. The proposed hypothetical procalcitonin-guided framework provides a foundation for future research and clinical implementation.