The role of total cell-free DNA in predicting outcomes among trauma patients in the intensive care unit: a systematic review

The inclusion criteria for studies in this review were cohort human studies that investigated levels of cfDNA in plasma or serum in trauma patients aged ≥18 years in the ICU. Trauma was defined as all grades of trauma ranging from minor to severe trauma, including isolated traumatic brain injury (TBI) that resulted in ICU admission. Studies exclusively evaluating circulating RNA as well as studies conducted outside the ICUs were excluded. A thorough assessment of the quality of DNA sampling and processing was conducted for all included studies using previous definitions [21] (see Additional file 1). We included studies that analyzed DNA by specific sequencing (beta-globin, GAPDH, NADH dehydrogenase) or fluorescent methods. We only included published studies and only studies published in the English language.

A computerized comprehensive search strategy was conducted using four databases (PubMed, EMBASE, SCOPUS, and the Cochrane Central Register for Controlled Trials) from January 1974 to January 2016. The search was performed on 20 January 2016. The following literature search was used in PubMed: “(circulating cell free dna) OR cfdna) OR circulating nucleic acids) OR cell free mitochondrial DNA) OR nDNA) AND (injury) OR trauma) OR stress) OR surgery) OR intensive care unit) OR perioperative) OR postoperative) OR intraoperative) OR preoperative)”.

The literature search was adapted from the PubMed literature search to EMBASE, SCOPUS, and the Cochrane Central Register for Controlled Trials. We supplemented the structured literature search with searching of the reference lists from the included articles in order to find additional eligible studies.

The Cochrane systematic review tool Covidence.org was used in the screening process. Two reviewers (MG and JB) independently screened titles and abstracts until the full-text articles were found. Two authors (MG and JB) independently assessed the full-text articles. Whenever different opinions emerged a third author (IG) was included in the discussion until consensus was reached.

All included articles were assessed for the following information: publication details, study method, patient details, specific time points of cfDNA measurements, mortality, post-traumatic complications, and trauma severity. This data was collected using a data sheet.

Risk of bias assessment was performed by two authors (MG and JB) using an adapted version of the Scottish Intercollegiate Guidelines Network (SIGN) (http://​www.​sign.​ac.​uk/​methodology/​checklists.​html) checklist for cohort studies [22]. The SIGN checklist evaluates selection bias, performance bias, attrition bias, detection bias, and statistical analysis. All studies were assigned up to 8 points if they met the SIGN checklist criteria. Studies with 8 points were considered to be high quality (HQ) studies with little or no risk of bias, while studies between 6 and 7 and 0 and 5 were considered acceptable quality (AQ) with an associated risk of bias and unacceptable quality with a high risk of bias, respectively.

One point was given for every checkpoint met, while studies with insufficient information regarding a checkpoint were given 0 points. Checkpoints not applicable for certain studies were given 1 point as it was not considered a shortcoming of the quality of the study.

Different techniques and methods have been developed over the past years to process and analyze cfDNA. This was taken into consideration by a thorough assessment of DNA sampling and processing in all included studies using previously described methods [21] on the quality of cfDNA processing and analyses (see Additional file 1). We adopted these recommendations of cfDNA processing into a list of checkpoints for each study. Checkpoints considered the type of medium for analysis (plasma or serum), the amount and speed of centrifugation, type of sample tube, freezing of samples and under which conditions (–20 or –80 °C), and finally if blood had been processed before 4 hours. The results of these assessments were collected on the same data sheet as outcomes. Greater than 5 points were deemed as high-quality circulating DNA processing, 3–4 points were deemed as acceptable quality, while 0–2 points were considered unacceptable quality. The quality of the DNA processing was taken into account in evaluating the robustness of study conclusions.

The search identified 2728 potential studies (see Fig. 1), where 2710 studies by title and abstract screening were deemed irrelevant. Of the remaining 18 studies, six studies were excluded for the following reason: four studies were based on nontrauma patients [23‐26] and two studies included both ICU and nonICU patients [27, 28]. A total of 15 studies from 14 references were included in this systematic review of which two studies were added after examining the included studies’ references lists (see Fig. 1). Ten studies were designed as prospective cohort studies, three studies included selected patients from a cohort, while one study included retrospective patients.

The main characteristics and cfDNA processing score (see Additional file 2 for information regarding specific points) of the included studies are summarized in Table 1. A total of 904 patients were included in the period 1996–2013 with a sample size range of 25–188. One reference presented results from two studies leading to two separate DNA processing and risk of bias assessments [29]. Ten studies included nonspecific trauma patients while five studies included patients with TBI with or without extracranial trauma [16, 30‐36]. Nine of the nonspecific trauma studies based severity of trauma on the Injury Severity Score (ISS) [14, 15, 29, 34‐38], while one used both the Acute Physiology and Chronic Health Evaluation (APACHE) II and Sepsis-related Organ Failure Assessment (SOFA) score [39]. All TBI studies used the Glasgow Coma Scale (GCS) to classify the effects of the trauma while using either the abbreviated injury scale (AIS), ISS, or APACHE II to describe the severity of trauma. The DNA processing score ranged from 2–5 (Table 1). All studies measured cfDNA on ICU admission, while eight studies also measured cfDNA at different time points during the admission period.

The patient population differed to some extent between the studies. Ten studies included all trauma patients in the ICU [14, 15, 29, 34‐39], while five studies required patients to be diagnosed with TBI [16, 30‐33]. One study only included male patients [32]. Treatment strategies during ICU admission were not described, which mean that different treatments might have been used. There is a risk of bacterial DNA contamination in DNA assessment methods, but only one study reported on bacterial DNA analyses [38].

Risk of bias assessments are presented in Additional file 3. According to the adapted version of the SIGN checklist (Additional file 4), one study was rated as a high-quality study [16] while 13 studies were rated as being of acceptable quality and one was rated as of unacceptable quality. Eleven of 15 studies had control groups [14, 16, 30‐33, 35‐39]; 12 studies failed to deliver adjusted confidence intervals [14, 15, 29‐31, 34, 36‐39], and a total of five studies were deemed to have a high risk of selection bias as a consequence of a retrospective design [14], failure to provide exclusion criteria [39], and nonconsecutive patient inclusion [29, 32]. A possibility of detection bias was assessed in 11 of 15 groups as a consequence of these not providing information regarding the blinding of essential risk factors such as severity of trauma or clinical outcome [14, 15, 30‐32, 34‐39].

Two studies were considered as high-quality studies based on DNA processing [15, 39] while nine were deemed of acceptable quality [14, 16, 29, 33, 35‐38] and three presented with an unacceptable quality of DNA processing [30, 32, 34]. One study was not assessed as no information regarding blood sampling and processing was available [31]. Twelve of 15 studies used plasma [14‐16, 29, 32‐39]. All studies centrifuged blood samples prior to freezing. None of the TBI studies stored the samples at –80 °C while 9 of 10 nonspecific trauma studies did [15, 29, 34‐39].

The main outcomes are presented in Table 2. In a study of 56 patients, no correlation was found between severity of trauma scored with ISS and levels of cfDNA [37]; however, four other studies found a significant correlation between cfDNA and ISS score. Patients diagnosed with acute lung injury (ALI) or acute respiratory distress syndrome (ARDS) had significantly higher cfDNA compared to patients without these diagnoses [14]. This corresponds to a study which proved that repeatedly measured levels of cfDNA 1 h, 2 h, and 3 h after ICU admission in 25 patients at all time points were significantly correlated with a higher degree of injury [29]. ISS scores were also correlated to cfDNA in a prospective study from 2008 on 37 patients [15] which investigated cfDNA originating from neutrophils using a fluorescent method, which was able to identify three different types of cfDNA kinetics. Type 1 kinetics were characterized by an initial value of cfDNA below 800 ng/mL followed by a rapid decrease. Type 2 kinetics were characterized by initial values above 800 ng/mL followed by a rapid decrease. Type 3 kinetics of cfDNA was characterized by high initial values above 800 ng/mL and a rapid decrease on days 1–5 followed by a significant increase of cfDNA on days 7–10 compared to type 1 and 2 kinetics. Type 3 kinetics were also associated with a higher ISS score, but not significantly. A significant association between type 3 kinetics and sepsis was found.

SOFA and APACHE II scores were investigated in a study on 96 patients which found that higher levels of cfDNA significantly correlated with higher SOFA scores but not with APACHE II score [39]. A study measuring mtDNA in trauma patients found a significant correlation between mtDNA levels in the plasma and APACHE II and ISS scores [35]. Similar results were found in two other studies [34, 36], but one study did not find an association between mtDNA levels and APACHE II score [34].

In a study on patients with severe TBI, high APACHE II and ISS scores were significantly correlated with higher levels of cfDNA 24 h after ICU admission. This study also showed that patients with high GCS score (11–15) had significantly lower cfDNA at both ICU admission and after 24 h [30], which was confirmed in a similar study [33]. However, a significant correlation between patients with isolated TBI and patients with extracranial injuries was not found for cfDNA levels in 41 male patients with TBI [32].

Regarding nDNA, similar results have been presented where patients with high nDNA had lower GCS scores and significantly higher ISS scores [16]. However, one study, reporting on both nDNA and mtDNA, found no correlation between levels of nDNA and mtDNA in relation to SIRS or multi-organ failure (MOF) [38].

Survival data in relation to cfDNA was presented in ten studies. One study reported not being powered to present survival data [38], while three references did not present survival data analyzed in relation to cfDNA levels [29, 35, 37]. In the earliest study measuring cfDNA in a retrospective population of trauma patients, it was found that all patients who died in the ICU had significantly higher cfDNA than survivors [14]. Likewise, it has been reported that nonsurvivors had a 2.3-fold higher cfDNA level than survivors [39]. In a prospective cohort of 37 patients, lower cfDNA levels at ICU admission had a negative predictive value (NPV) of 100% in relation to death [15]. In another study, survivors presented significantly lower cfDNA levels on ICU admission and 24 h after, but not 48 h after, ICU admission [30]. Other authors only found this significant correlation at 24 h after ICU admission [31‐33]. Two of the studies measuring mtDNA presented survival data, both showing nonsurviving patients with significantly higher mtDNA than survivors [34, 36]. Finally, in a prospective study of 88 patients where nDNA was measured at ICU admission and on days 4 and 7, patients with a fatal outcome had significantly higher nDNA at all time points [16].