Symptom Cluster Profiles Following Traumatic Orthopedic Injuries: A Protocol

Abstract

Traumatic injuries affect millions of Americans annually, resulting in $671 billion in healthcare costs and lost productivity. Post-injury symptoms, like pain, sleep disturbance, anxiety, depression, and stressor-related disorders are highly prevalent following traumatic orthopedic injuries (TOI) and may contribute to negative long-term outcomes. Symptoms rarely present in isolation, but in clusters of two or more symptoms that co-occur to affect health in aggregate. Identifying symptom cluster profiles following TOI may identify those at highest risk for negative outcomes. Dysregulation of brain-derived neurotrophic factor (BDNF) is a potential biological mechanism responsible for symptom cluster profile membership after TOI and may be targeted in future precision-health applications. The purpose of this paper is to present the protocol of a cross-sectional study designed to identify symptom cluster profiles and measure the extent to which the BDNF val66met mutation and serum concentration of BDNF are associated with membership in symptom cluster profiles. We plan to recruit 150 TOI survivors within the first 72 hours of injury. The study aims are to 1) describe TOI survivors’ membership in symptom cluster profiles, indicated by pain, sleep disturbance, and symptoms of anxiety, depression, and stressor-related disorders, immediately following a TOI; 2) examine associations between demographic and clinical factors and symptom cluster profile membership among TOI survivors; 3) test the hypothesis that low serum concentrations of BDNF are associated with membership among symptom cluster profiles following TOI; and 4) test the hypothesis that the presence of the val66met mutation on one or both alleles of the BDNF gene is associated with membership among symptom cluster profiles following TOI.

Each year, millions of people are injured, with 75% suffering traumatic orthopedic injuries (TOI): Injuries to the skeleton and associated soft tissues as a result of physical trauma (Armstrong et al., 2016; Canadian Institute for Health Information, 2012; WISQARS Nonfatal Injury Visualization, 2019). Physical trauma is the leading cause of death before age 46 and of years of potential life lost before age 70 (Trauma Statistics and Facts, 2019; WISQARS Years of Potential Life Lost (YPLL) Report, 1981 – 2017, 2019). Physical trauma also results in $671 billion in health care costs and lost work productivity annually—more than cancer, diabetes, or heart disease (Trauma Statistics and Facts, 2019).

Traumatic orthopedic injuries expose survivors to negative physical outcomes such as nonunion, malunion, loss of range of motion, osteomyelitis, and nerve damage (Armstrong et al., 2016). In addition, TOI survivors report experiencing distressing symptoms that begin immediately after injury and continue for months or years. In particular, pain, sleep disturbance, and symptoms of anxiety, depression, and stressor-related disorders are highly prevalent and often comorbid following TOI (Berube et al., 2016; McCrabb et al., 2019; Muscatelli et al., 2017; Shulman et al., 2015). Such post-injury symptoms contribute to important outcomes among TOI survivors, including lengths of hospital and intensive-care unit stays (Haupt et al., 2018), health-related quality of life (Eldin et al., 2012), and return to work (Giummarra et al., 2017).

Individual symptoms may affect long-term outcomes after TOI, but they rarely present in isolation. Instead, they often present in clusters of 2 or more symptoms that collectively influence outcomes (Kim et al., 2005; Miaskowski et al., 2007). Although, symptoms may cluster together, the degree to which a person experiences a given symptom in a cluster can vary. Identifying the common ways in which TOI survivors experience clustered symptoms may reveal subgroups of individuals at highest risk for negative symptom and functional outcomes (Katzan et al., 2019; Miaskowski et al., 2007). This can be accomplished through the use of person-oriented approaches to the study of symptom clusters, like latent profile analysis, that place individuals into mathematically determined groups called symptom cluster profiles based upon the common ways in which they experience a set of symptoms (Miaskowski et al., 2007; Tein, 2013).

However, research using such approaches among TOI survivors is sparse. To date, we have identified only two studies using these approaches with orthopedic patients (Castillo et al., 2019; Van Son et al., 2017). Van Son et al. (2017) found three latent profiles of participants with distal radius and ankle fractures determined by varying levels of health-related quality of life and described by several factors, including the severity of participants’ trait anxiety, neuroticism, extraversion, and pain characteristics. Castillo et al. (2019) found that that membership in one of 6 distinct latent profiles distinguished by the presence of certain risk and protective factors (pain, depression, PTSD, alcohol and tobacco use, resilience, social support, and self-efficacy) measured 6 weeks after injury predicted physical and mental health outcomes 12 months after TOI.

Together, these researchers showed that TOI survivors’ collective experiences with clustered symptoms influenced their long-term outcomes, however neither group included data collected prospectively during the immediate post-injury period, a potentially important time during which a person’s risk for negative outcomes can be assessed (Castillo et al., 2019; Van Son et al., 2017). Additionally, both groups of researchers included non-symptom variables when describing the latent profiles, meaning true symptom cluster profiles have yet to be described in the TOI population. Therefore, research examining symptom cluster profiles among TOI survivors during the immediate post-injury period is necessary to help identify the optimal time during which a TOI survivor’s risk for long-term outcomes can be assessed.

Symptom cluster profiles may share a common biological etiology that may be targeted in future pharmacologic and non-pharmacologic precision-health applications (Dorsey et al., 2018; McCall et al., 2018; Miaskowski et al., 2004). Dysregulation of brain-derived neurotrophic factor (BDNF), a neurotrophin that regulates neuronal development, growth, and function (Huang & Reichardt, 2001), is a biological process that may contribute to symptom cluster profile membership after TOI. One factor that influences BDNF production and regulation is the val66met mutation, which is produced by the single nucleotide polymorphism (SNP) rs6265 of the BDNF gene. Found in approximately 30% of the overall population, not accounting for racial and ethnic differences, the val66met mutation significantly increases the likelihood of developing symptoms of anxiety, depression, stressor-related disorders and sleep disturbance by down-regulating activity-dependent release of BDNF throughout the nervous systems (Bachmann et al., 2012; Pitts et al., 2019; Règue-Guyon, 2018; Yang et al., 2019).

BDNF is clinically important in the TOI population because, although inactive in healthy bone, it plays an integral role in angio- and osteogenesis at a fracture site (Kilian et al., 2014; Zhang et al., 2017). Additionally, pre-clinical studies suggest that orthopedic injury increases BDNF concentrations systemically and locally in the spinal cord, amygdala, and hippocampus, providing further evidence that BDNF plays an integral role in the post injury courses of TOI survivors (Qin et al., 2005; M. D. Zhang et al., 2016). Yet, little is known about the associations between serum BDNF concentrations, the val66met mutation of the BDNF gene, and membership among symptom cluster profiles in TOI survivors.

Study Aims

We address this gap in knowledge by pursuing the following aims in the study described in this protocol: 1) Describe TOI survivors’ membership in symptom cluster profiles, indicated by pain, sleep disturbance, and symptoms of anxiety, depression, and stress related disorders, immediately following a traumatic orthopedic injury; 2) Determine associations between demographic and clinical factors and symptom cluster profile membership among TOI survivors; 3) Test the hypothesis that low serum concentrations of BDNF are associated with membership among symptom cluster profiles following TOI; and 4) Test the hypothesis that the presence of the val66met mutation on one or both alleles of the BDNF gene is associated with membership among symptom cluster profiles following TOI. The purpose of this paper is to report the protocol we are using to address these aims.

Our study is guided by a conceptual framework (Figure 1) built by merging the National Institutes of Health Symptom Science Model and the Theory of Unpleasant Symptoms (Cashion et al., 2016; Lenz et al., 1997). The National Institutes of Health Symptom Science Model guides understanding of complex symptoms and their phenotypic characterization to identify and test candidate biomarkers which may be targeted for therapeutic and clinical intervention (Cashion et al., 2016; Cashion & Grady, 2015). The Theory of Unpleasant Symptoms describes the ways in which influencing factors interact with one another to affect the development of symptoms which interact to affect an individual’s performance (Lenz et al., 1997). The Theory of Unpleasant Symptoms does not specify the role of biomarkers in symptom development, but this is elaborated in the National Institutes of Health Symptom Science Model (Cashion et al., 2016; Lenz et al., 1997). In figure 1, the double arrows indicate that the demographic factors, clinical factors, and biomarker data interact with one another. The convergent arrow indicates that the demographic factors, clinical factors, and biomarker data collectively influence membership in a symptom cluster profile. The dotted arrows reflect the variables used to describe membership in a symptom cluster profile.

Figure 1.

Conceptual Framework

Symptom cluster profiles following traumatic orthopedic injuries: A protocol

Abbreviations: BDNF = brain-derived neurotrophic factor; SNP = single nucleotide polymorphism

Methods

Design and Setting

The current study employs a cross-sectional design with data obtained from adult participants who sustained a traumatic orthopedic injury. Participants are recruited from those under the care of the orthopedic trauma service at the level 1 trauma center at Yale New Haven Hospital (YNHH). YNHH’s orthopedic trauma service treats over 2,000 patients and performs over 1,300 surgeries annually.

Eligibility

The inclusion and exclusion criteria are designed to include participants with isolated TOIs and exclude those with multi-traumatic injuries and those in vulnerable populations. Participants are eligible for inclusion if: (1) they are aged 18 years or older; (2) the injury occurred within the last 72 hours; (3) the fracture is to the appendicular skeleton and/or pelvic ring; (4) the fracture was the result of a high energy mechanism (e.g., motor vehicle collision, fall from height, assault, etc.); (5) the participant can speak and understand English; and (6) the participant is cognitively intact as assessed via review of the medical record for conditions known to effect mental status (i.e. traumatic brain injury, intracranial hemorrhage) and in consultation with the treatment team and nursing documentation of mental status. The full inclusion and exclusion criteria are presented in Table 1.

Table 1.

Inclusion and Exclusion Criteria

Inclusion

Exclusion

Aged 18+ years Within 72 hours of injury Fracture to appendicular skeleton or pelvic ring High energy mechanism Speak and understand English Cognitively intact

Traumatic brain injury Thoracic or abdominal trauma Spinal cord injury Rib or spine fracture requiring intervention Isolated soft tissue injury (e.g. degloving) Isolated fractures to fingers or toes Peri-prosthetic fractures Under the influence of illicit substance or alcohol Attempted suicide Diagnosed mental disorder (e.g. bipolar disorder) Prescribed psychotropic medications before TOI Chronic pain Pregnant Incarcerated

Abbreviation: TOI = traumatic orthopedic injury

Sample Size

There is no standardized method for calculating sample size a priori for latent profile analyses. Experts recommend a sample size of at least 100 in order to extract the correct number of profiles (Conley, 2017; Dziak et al., 2014). Above this threshold, sample size has little effect on the ability to discern the correct number of profiles (Tein, 2013). We are recruiting a sample size of 150 because recent work shows that data from 150 participants is sufficient for extracting the correct number of profiles (Gudicha, 2016; Russell et al., 2019). We are also aiming to recruit a sample that mirrors the demographics of the state of Connecticut, the catchment area for the level 1 trauma center at YNHH, including 50% male (N=75) and 50% female (N=75) participants, with at least 30% of the study sample (N=46) being of racial and ethnic minority background (Quick Facts Connecticut, 2019).

Study Procedures

The principal investigator (PI) identifies eligible TOI survivors by rounding with the orthopedic team each morning and regularly reviewing the electronic medical record of each newly admitted patient. Once eligibility is confirmed, either the PI or a trained research nurse approaches the potential participant at their bedside, explains the study, and provides them with written information. If a TOI survivor is willing to participate, the PI or research nurse obtains informed consent and documents it electronically, with participants signing the digital consent form with a stylus on an electronic tablet. Participants are given a paper copy of the consent form for their records. Immediately after obtaining consent, the patient-reported outcomes measures and biomarkers are collected. The entire research interaction lasts no more than one hour.

All data are recorded using REDCap data management software. Demographic information and patient-reported outcomes measures are entered directly into the REDCap Mobile App by the participant on an electronic tablet. The PI or research nurse may also administer the patient-reported outcomes measures verbally. Clinical information is collected via chart review and entered into REDCap by the PI.

The biological specimens are collected via venipuncture by the PI or research nurse. Each sample is prepared by the PI following Yale Center for Clinical Investigation Standard Operating Procedures and stored at −80°C in the Orthopaedic Histology and Histomorphometry Laboratory at Yale School of Medicine until ready for overnight dry ice shipping to the University of Maryland School of Nursing for analysis (Biospecimen Management, 2019). All specimens will be analyzed by the PI in the laboratory located in the Pain and Translational Symptom Science Department at University of Maryland School of Nursing.

Variables and Measures

Demographic and Clinical Factors

We are collecting the following demographic and clinical factors: age, sex, gender, race, ethnicity, menstrual history, social support, insurance status, employment status, employment type, education level, resilience, fracture site(s), fracture classification(s), type of fracture management, weightbearing status(es), the mechanism of injury, Injury Severity Score, admission hemoglobin/hematocrit, admission lactic acid, and volume of blood transfusions received during resuscitation.

We are measuring social support using the Multidimensional Scale of Perceived Social Support. The Multidimensional Scale of Perceived Social Support, first described by Zimet et al. (1988), is a 12-item self-report instrument that measures a person’s perceptions regarding the amount of social support available to them in three domains: family, friends, and significant others. Each item is scored on a 1–7 Likert scale with higher scores on each item indicating higher perceived social support within a given domain (Zimet et al., 1990). Total scores on the Multidimensional Scale of Perceived Social Support, ranging from a possible 1–7, are obtained by averaging the scores for each item (Zimet et al., 1988) The Multidimensional Scale of Perceived Social Support is highly reliable (Cronbach’s alpha = 0.84–0.92), as are each of the subscales (Cronbach’s alpha = 0.81–0.98) (Zimet et al., 1990). The Multidimensional Scale of Perceived Social Support is valid for use among populations from which TOI survivors hail, including youths (Bruwer et al., 2008, Canty-Mitchell & Zimet, 2000), young adults (Dahlem et al., 1991), and older adults (Stanley et al., 1998). The Multidimensional Scale of Perceived Social Support has been used to evaluate the role of social support in physical and mental health outcomes among TOI survivors (Maselesele & Idemudia, 2013; Williams et al., 2004).

We are measuring resilience—the ability to bounce back or recover from stress—using the Brief Resilience Scale (Smith, 2008). The Brief Resilience Scale is a 6-item self-report instrument in which 3 items are positively worded and 3 are negatively worded (Smith, 2008). Each item is scored on a Likert scale ranging from 1, “strongly disagree”, to 5, strongly agree” (Smith, 2008). Scores on the Brief Resilience Scale are obtained by reverse coding the negatively worded questions and then averaging the scores of all 6 items (Smith, 2008). Higher scores on the Brief Resilience Scale indicate higher resilience (Smith, 2008). The Brief Resilience Scale was validated among individuals undergoing vocational rehabilitation, a population closely related to the TOI population (Tansey et al., 2015), is highly reliable, and has been used to examine resilience in TOI survivors (Beletsky et al., 2019; Smith, 2008; Verhiel et al., 2019).

The Orthopaedic Trauma Association and Gustilo-Anderson fracture classifications are internationally standardized methods to describe fracture morphology and severity (Armstrong et al., 2016). In the Orthopaedic Trauma Association fracture classification, each bone is assigned a number, and fracture morphology is described by an alpha-numeric modifier (Meinberg et al., 2018). For example, a multi-fragmentary, intra-articular, bicondylar tibial plateau fracture is recorded as 41C3.3 (Meinberg et al., 2018). The Gustilo-Anderson fracture classification differentiates the severity of open fractures, and consists of types I-III depending on the size of the open wound (Egol et al., 2010). Type III is differentiated into IIIa, IIIb, and IIIc, with IIIb fractures requiring soft tissue coverage via a rotational or free flap and IIIc fractures requiring vascular repair (Egol et al., 2010).

Injury Severity Score is a measure of physical injury in which scores range from 0–75, with higher scores indicating more severe physical injury (Association for the Advancement of Automotive Medicine, 2020; Copes et al, 1988). Each injury is coded using the Abbreviated Injury Scale and given a score of 1 (minor) to 6 (maximal) depending on severity (Association for the Advancement of Automotive Medicine, 2020). The Injury Severity Score is calculated by taking the sum of squares of the three highest Abbreviated Injury Scores across six body regions (Copes et al., 1988).

Serum BDNF

We will use the Promega BDNF EMax immunoassay system and will process the samples and assay serum BDNF levels according to the manufacturer protocols. In short, specimens will be aliquoted into a 96 well plate pre-coated with anti-BDNF monoclonal antibody. BDNF polyclonal antibody will then be added to each well and bind to the bound BDNF. Anti-IgY antibody will be added to each well as a tertiary reactant. Finally, a chromogenic agent will be added to the wells and the degree of color change measured. The amount of BDNF is proportional to the degree of color change detected. Each plate contains controls that prepare a standard curve. The Promega BDNF EMax immunoassay system produces less than 3% cross-reactivity with other neurotrophic factors and can detect BDNF at concentrations as low as 15.6pg/mL. The protocol and procedures have been previously published (BDNF Emax ImmunoAssay System, 2009; Pressler et al., 2017).

BDNF Genotyping

Each DNA sample will be assayed for SNP rs6265 using the manufacturer’s protocol for the Taqman SNP genotyping assay (TaqMan SNP Genotyping Assays User Guide, 2017). Briefly, genomic DNA will be extracted and purified from the blood sample and quantified using Nanodrop to obtain the 260/280 ratio. DNA from each subject will be diluted to a standard quantity using nuclease free water and aliquoted into a 96 well plate and reaction mix added to QC to a final volume in each well. The appropriate positive and negative controls will be included on the plate and the plate sealed with adhesive film and processed using real-time polymerase chain reaction. The procedures for performing the SNP analyses have been previously published (McGuigan & Ralston, 2002).

Symptoms

We are using instruments from the Patient-Reported Outcomes Measurement Information System (PROMIS) to assess symptoms of anxiety, depression, pain intensity, and sleep disturbance to promote the use of common data elements (Patient Reported Outcomes Measurement Information System, 2019; Redeker et al., 2015). Short forms of the PROMIS instruments are used to minimize respondent burden: A benefit of PROMIS is their brevity, taking less than 5 minutes to complete (Carlini et al., 2018; Kadri et al., 2018; Polit & Beck, 2017). Each PROMIS measure uses a Likert scale (1–5) for each item, with higher scores indicating more of a concept being measured (PROMIS Anxiety Scoring Manual, 2015). Scores are standardized using t-scores with a mean of 50 and a standard deviation of 10 (PROMIS Anxiety Scoring Manual, 2015). PROMIS is valid for use among orthopedic patients and recommended by the American Orthopaedic Foot and Ankle Society (Beleckas et al., 2018; Jildeh et al., 2018; Kitaoka et al., 2018).

Symptoms of acute stress disorder are measured with the Acute Stress Disorder Scale, a 19-item self-report measure that asks each participant if they have been exposed to a traumatic event and experienced symptoms that reflect the diagnostic domains of Acute Stress Disorder (Bryant et al., 2000). Each item is scored on a Likert scale from 1, “not at all likely”, to 5, “very much”, with a total score of ≥56 indicating the presence of clinically significant symptoms of Acute Stress Disorder (Bryant et al., 2000). A cut-off score of ≥56 has high sensitivity (0.95), specificity (0.83), as well as high positive predictive value (0.67) and negative predictive value (0.98) at predicting those who will go on to develop Post-Traumatic Stress Disorder based on the severity of their symptoms of Acute Stress Disorder (Bryant, 2012; Bryant et al., 2000). The Acute Stress Disorder Scale has been used to assess for symptoms of Acute Stress Disorder among survivors of physical trauma (Keyan & Bryant, 2019).

Participants are completing each symptom measure using time since admission to YNHH as the recall period. Using short recall times in PROMIS measures produces accurate findings (Schneider, Broderick, et al., 2013; Schneider, Choi, et al., 2013) and the Acute Stress Disorder Scale is designed to use short recall periods (Bryant, 2000). All variables and measures are presented in Table 2.

Table 2.

Summary of Study Measures

Measure	Format	Data Source
Demographics
Individual  Age, sex, gender, race/ethnicity, menstrual history, insurance type, employment status/type, education level	Interval, Nominal, Ordinal	Self-Report
Multidimensional Scale of Perceived Social Support	Scalar	Self-Report
Brief Resilience Scale	Scalar	Self-Report
Clinical Factors
Orthopedic  Fracture site, classification, management, weight-bearing status, mechanism of injury	Nominal	Chart Review
Medical  Hemoglobin/hematocrit, lactic acid, volume of blood transfusions	Interval	Chart Review
Injury Severity Score	Scalar	Chart Review
Biomarkers
Serum BDNF	Interval	Investigator Collected
BDNF SNP rs6265	Nominal	Investigator Collected
Symptoms
Acute Stress Disorder Scale	Scalar	Self-Report
PROMIS Anxiety v1.0 4a	Scalar	Self-Report
PROMIS Depression v1.0 4a	Scalar	Self-Report
PROMIS Pain Intensity v1.0 3a	Scalar	Self-Report
PROMIS Sleep Disturbance v1.0 4a	Scalar	Self-Report

Abbreviation: BDNF = brain-derived neurotrophic factor; PROMIS = Patient Reported Outcomes Measurement Information System; SNP = single nucleotide polymorphism

Data Analysis

We will compute univariate statistics to examine frequency distributions for categorical variables, normal distributions for continuous variables, and assess normality with graphs, skewness, and kurtosis data. We will assess for multivariate normality using Royston’s and Mardia’s tests which are multivariate extensions of the Shapiro-Wilk goodness of fit test and measures of skewness and kurtosis, respectively (Oppong & Yao, 2016). Descriptive statistics (e.g. mean, median) will be calculated for demographic, clinical, and symptom variables. Bivariate statistics will be performed on demographic factors, clinical factors, and biomarker data, as well as the symptoms, to assess for confounding variables. The relationships between the individual symptoms will then be explored to identify the symptom cluster profiles. Finally, we will analyze the associations of demographic and clinic factors, as well as biomarker data, with the symptom cluster profiles.

Categorical variables will be recoded as dummy variables. Scores for each PROMIS measure (pain intensity, sleep disturbance, depression, and anxiety) will be converted into normalized t-scores and used in all analyses as continuous variables. Scores on the Acute Stress Disorder Scale and Multidimensional Scale of Perceived Social Support, ranging from 19–95 and 1–7, respectively, will be used as continuous variables.

Prior to performing analyses, we will assess for missing data. all missing data will be coded as (.). Missing data on indicator variables will be handled by the model using full information maximum-likelihood information (Vermunt & Magidson, 2016). Latent profile analysis is robust to missing data on indicator variables (Passalacqua, 2013). Missing data on covariates will be handled via multiple imputation.

Aim 1

We will perform latent profile analysis using Latent Gold Version 5.1 (Latent Gold). Using continuous variables, latent profile analysis produces exclusive and exhaustive symptom cluster profiles (Tein, 2013). Latent profile analysis calculates 2 parameters: symptom cluster profile prevalence and item-response probabilities. We expect latent profile analysis to produce two or more symptom cluster profiles with unique item-response probabilities. We will run 5 models with 1–5 classes and compare them using model fit statistics. The final model will be determined by minimizing Akaike’s and Bayesian information criterion; performing the Vong-Lo-Mendell-Rubin and Lo-Mendell-Rubin likelihood ratio tests with a p<0.05 indicating a significant model; and maximizing entropy (Collins & Lanza, 2010; Katzan et al., 2019; Tein, 2013). Symptom cluster profiles will be interpreted and labeled using item-response probabilities as is standard in latent profile analysis.

Aim 2

We will use the multinomial logistic regression function in the Latent Gold software to identify demographic and clinical factors that are associated with symptom cluster profile membership (Latent Gold). The symptom cluster profiles determined in aim 1 will be used for this analysis. This model will be constrained to have the same membership probabilities as the model from aim 1. The multinomial regression produces the log-likelihood, odds ratio, and confidence interval for each variable included in the model (Collins & Lanza, 2010). Multinomial regression extends logistic regression by allowing for multiple non-ordered categorical outcomes, such as 3 or more symptom cluster profiles (Conley et al., 2017). Variables that will be addressed in aim 2 include the demographic factors and clinical factors. Variables will initially be added using forced entry via theoretical linkage and the final model will be identified using backwards step-wise selection. Should the latent profile analysis produce symptom cluster profiles with too few events per variable (Peduzzi et al., 1996), we will analyze the demographic and clinical factors in separate blocks using forward step-wise selection. Odds ratios and confidence intervals will be used to determine the significance and strengths of the associations of the covariates with symptom cluster membership.

Aim 3

To explore the associations of low serum concentrations of BDNF with membership among symptom cluster profiles, we will use ANOVA testing. Post-hoc Tukey’s test will be used to adjust for multiple comparisons. The Kruskall-Wallis test with the post-hoc Dunn’s test will be used for non-parametric data if needed. Significance for the ANOVA, Tukey’s, and/or Kruskall-Wallis tests will be determined using a p<0.05. Significance of Dunn’s test, if needed, will be determined using an overall p<0.05 that will be adjusted for multiple comparisons using Bonferroni corrections.

Aim 4

We will perform genotypic association testing using conventional χ² to examine the associations of the val66met mutation of the BDNF gene (val/val, val/met, met/met) and membership in the symptom cluster profiles (Clarke et al., 2011). We will determine significance using an overall p<0.05 that we will adjust for multiple comparisons using Bonferroni corrections (Clarke et al., 2011; Talluri et al., 2014).

Discussion

Post-injury symptoms are chronic and highly prevalent after TOI and likely to negatively influence long-term outcomes in TOI survivors. Identifying symptom cluster profiles in TOI survivors and testing the associations of a novel biomarker, BDNF, with symptom cluster profile membership early in the post-injury period will provide information foundational to the screening and treatment of post-injury symptoms in TOI survivors. Findings from this study will be used to design longitudinal studies that evaluate symptom cluster profile trajectories, the extent to which BDNF and other candidate biomarkers predict membership in symptom cluster profile trajectories, and the long-term outcomes of membership among such trajectories. Future longitudinal work will be able to evaluate if the “omics” of and membership in symptom cluster profiles contribute to orthopedic outcomes (e.g. time to union, rates of union), a question that does not appear to have been answered to date.

This study is innovative for several reasons. First, this study will include a clinically heterogenous sample of orthopedic trauma patients that includes the majority of TOIs (Y. Zhang et al., 2016). Second, the proposed study will be among the first to utilize a person-oriented approach to identify subgroups of patients within the sample population by using symptoms that have been shown a priori to affect the TOI population (Collins & Lanza, 2010; Miaskowski et al., 2007). Third, the proposed study will test the associations between serum BDNF and SNPs of its candidate gene and group membership in symptom cluster profiles following TOI to further define the genotypic and phenotypic characteristics that predispose a person to membership in a symptom cluster profile (Miaskowski et al., 2007). To our knowledge, we are the first to incorporate study of the associations of BDNF with symptom cluster profiles in this large and vulnerable group of trauma survivors.

Study Status

All research materials including REDCap surveys, informed consent documentation, study protocol, training materials, and recruitment flyer were created in September and October 2019. We obtained approval from the institutional review board at Yale University and the Nursing Research and Evidence Based Practice Committee at Yale New Haven Health System in October 2019. The PI performed multiple information sessions with clinical nurses throughout YNHH in October and November 2019. The research nurses were trained via multiple in-person sessions with the PI in October and November 2019. The first participants were enrolled in late November 2019. We paused research activities in March 2020 for COVID-19-related considerations and resumed enrolling in July 2020. As of this writing, we have screened 502 potential participants of which 119 were eligible and 80 (67.2%) enrolled.

Lessons Learned

We have learned several lessons during this study that will inform this and future work. Trauma survivors report participating in research early after injury to be a positive experience, however, obtaining informed consent during this time can be challenging (Dutton et al., 2008; Tutton et al., 2018). Our study is no different. We initially enrolled TOI survivors who were within 24 hours of their injuries, but extended the enrollment window to 72 hours post-injury after we were unable to reach potential participants for technical reasons (e.g., in surgery) and several individuals declined participation, explaining they were not in a mental state to participate in research. Extending enrollment allowed us to successfully enroll participants who were previously unavailable and those who did not feel well enough to participate during the initial interaction by scheduling a second visit 24 hours later.

Our understanding of the need for assistance with completing self-report measures in trauma survivors has been greatly enhanced. We originally decided that the PI or research nurse should be able to verbally administer the self-report measures because we anticipated participants having upper extremity injuries that prevented them from completing the measures independently. In our experience, very few participants have had upper extremity injuries that prevented them from completing the self-report measures, but many participants have requested the measures be administered verbally for reasons such as losing their glasses in the traumatic event or preference because of their current mental state. We will continue to allow for verbal administration of self-report measures in our future work.

Finally, due to the COVID-19 pandemic, we learned the importance of having a streamlined in-person research team. Much of the work on this study (e.g. screening, chart review) can be performed remotely, but key aspects (e.g. informed consent, venipuncture) must be performed in-person. The research team includes two research nurses who assist with all in-person tasks except biomarker preparation and storage, however, due to the pandemic, they are temporarily unable to work on the study. The principal investigator is trained in all aspects of the research protocol and is currently completing all in-person research activities, otherwise, this study would be indefinitely paused.