Introduction

Due to population aging, the number of elderly patients presenting to trauma centers will continue to increase. Advanced age is a known risk factor for morbidity and mortality after traumatic injury [1, 2] but it is clear that other factors such as medical comorbidities also contribute strongly to outcomes [3, 4]. Recently, attention has been focused on the concept of frailty as a driver of outcomes across a wide range of disease states in the elderly, including chronic kidney disease, malignancy, and aortic valve disease [5–7]. The original description of frailty includes difficulty with activities of daily living, reduced speed of ambulation, subjective feelings of exhaustion, unintentional weight loss, and weakness as measured by grip strength [8]. The association of frailty with outcomes after trauma is not well characterized, perhaps in part because frailty evaluation using conventional metrics in the trauma population is not practicable. Trauma is by its nature unpredictable, and for this reason most injured patients will present without baseline measurements of frailty. Attempts to measure frailty after injury using conventional metrics may be limited by altered mental status or inability to participate in physical activity. An attractive solution to this problem is to measure a surrogate for frailty that does not rely upon active patient participation, such as radiographically determined central sarcopenia [9]. Central sarcopenia, as measured by psoas cross-sectional area, has been associated with poor outcomes after liver transplantation [10] and open repair of ruptured aortic aneurysms [11], but the impact of central sarcopenia has not yet been well characterized in the elderly trauma population.

We sought to evaluate whether central sarcopenia, as measured by psoas cross-sectional area on admission imaging, is associated with outcomes in elderly trauma patients. We hypothesized that lower psoas cross-sectional area would be associated with increased morbidity and mortality in a cohort of elderly, severely injured trauma patients.

Materials and methods

After obtaining institutional review board approval, a query of our institutional trauma registry across a 5-year time period (2005–2010) was performed. Data for our trauma registry are collected prospectively and reported to the Pennsylvania Trauma Outcomes Study (PTOS) database by trained nurse abstractors. Study inclusion criteria were age ≥55 years, severe injury (Injury Severity Score (ISS) >15), and ICU length of stay >48 h. Patients were excluded if they suffered a critical head injury (defined as a Head/Neck Abbreviated Injury Scale (AIS) score ≥5), did not receive admission cross-sectional imaging of the abdomen, had fractures or preexisting hardware of the 4th lumbar vertebral body, or had a retroperitoneal hematoma that distorted the cross-sectional area of the psoas at the level of the L4 vertebral body. Patient demographics (age, sex, and race), physiologic variables on presentation, mechanism of injury, Injury Severity Score, hospital length of stay (HLOS), ICU length of stay (ILOS), ventilator days, and comorbidities were abstracted from the institutional registry. Morbidity was measured by PTOS-defined occurrences (see “Appendix 1”). Patients meeting all inclusion criteria with no exclusion criteria were then uploaded into a RED Cap database [12]. Computed tomography (CT) studies of the abdomen were obtained from the medical record and evaluated for each patient by one of three trained reviewers (DG, LE, DH). Prior to abstracting the study CTs, a sample of studies were independently abstracted by the three reviewers and results were compared in order to assess for inter-rater reliability. For each study, the right and left psoas muscle cross-sectional areas (PCSA) were measured at the level of the L4 vertebral body immediately inferior to the origin of the posterior elements. To normalize for body habitus, the cross-sectional area of the L4 vertebral body was also recorded at this level (Fig. 1). Mean PSCA was calculated for each patient, and the ratio between mean PSCA and L4 vertebral body area was calculated using the following formula:

Fig. 1
figure 1

Psoas:lumbar vertebral index was calculated as the ratio between the mean psoas cross-sectional area and the vertebral cross-sectional area at the level of the L4 vertebral body just inferior to the insertion of the posterior elements

Full size image
$$ {\text{Psoas:L4 vertebral index = ([right PCSA (mm}}^{ 2} ) {\text{ + left PCSA }}\left( {{\text{mm}}^{ 2} } \right) ] {\text{ / 2)/ L4 vertebral CSA }}\left( {{\text{mm}}^{ 2} } \right) . $$

The 50th percentile of psoas:L4 vertebral index (PLVI) value was determined and patients were grouped into high (>0.84) and low (≤0.83) categories based on their relation to the cohort median. Univariate analyses of patient demographic variables, admission vital signs, comorbidities, and outcome measures between the two groups were performed.

Primary outcomes

The primary outcomes of interest of the study were in-hospital mortality and morbidity. Morbidity was defined as a composite endpoint of any of 37 PTOS-defined complications (see “Appendix 1”) and as subgroups of respiratory complications (acute respiratory distress syndrome, acute respiratory failure, aspiration/aspiration pneumonia, atelectasis, or pneumonia), infectious complications (sepsis, septicemia, acute sinusitis, soft tissue infection, urinary tract infection, pneumonia, wound infection, central nervous system infection, or empyema), organ failure (acute respiratory distress syndrome, coagulopathy, acute renal failure, or liver failure), thromboembolic complications (pulmonary embolism or deep venous thrombosis), and hemorrhagic complications (coagulopathy or postoperative hemorrhage).

Secondary outcomes of interest were hospital length of stay (HLOS), ICU length of stay, and total ventilator days, which were examined between groups. In addition, the PLVI category was examined via mechanism of injury.

Statistical analyses

The associations of baseline characteristics with morbidity and with hospital mortality were tested using t test for continuous normally distributed variables and Mann–Whitney or Kruskal–Wallis for continuous non-normally distributed variables. Categorical variables were compared using χ 2 or Fischer’s exact tests, as appropriate. Univariate logistic regression was used to assess the association between independent variables and the primary outcomes of interest.

Multivariable logistic regression models were developed to adjust the association of PLVI with morbidity and mortality for potential confounders. Variables found to be associated with the outcome of interest in univariate analysis with a p value of <0.10 were included in the multivariate logistic models. All statistical analyses were performed using SPSS version 19.0 (IBM, Chicago IL).

Secondary outcomes of interest were hospital length of stay (HLOS), ICU length of stay, and total ventilator days. In addition, mechanism injury by PLVI category was examined.

Results

At total of 234 patients met the inclusion criteria, of whom 54 were excluded (Fig. 2), leaving 180 patients for analysis. Median age in the cohort was 74 years (IQR 63–82), and median ISS was 24 (IQR 18–29). Patients were 58 % male and 66 % Caucasian.

Fig. 2
figure 2

Study inclusion criteria. PTOS Pennsylvania Trauma Outcomes Study, CTAP computed tomography of the abdomen/pelvis

Full size image

There was no correlation between L4 vertebral body CSA and age, but psoas CSA did have a weak negative relationship with age (R 2 = 0.15). The correlation between right and left psoas CSA was high but not complete (R 2 = 0.75), so mean psoas CSA was used to calculate the PLVI. PLVI values were found to be normally distributed within the study population, with a mean value of 0.85 ± 0.25. PLVI values were higher in male patients than in female patients (0.91 ± 0.26 vs. 0.77 ± 0.21, p < 0.001). Lumbar–psoas vertebral index rating reliability was found to be good between raters, with a maximum deviation of 23 %, which resulted in no misclassifications based on groups. Compared to patients in the high-PLVI group, patients in the low-PLVI group were older (median age 79 (IQR 72–85) vs. 70 (IQR 60–77), p < 00.1), less likely to be male (43 vs. 71 %, p < 0.001), and less likely to be Caucasian (63 vs. 69 %, p = 0.04). No significant differences were seen between the two groups in terms of number of baseline comorbidities, mechanism of injury, ISS, or admission vitals (Table 1). In unadjusted analysis, patients in the low-PLVI group had higher morbidity (83 vs. 58 %, p < 0.001), but no difference was seen in mortality (16 vs. 10 %, p = 0.26) or any of the secondary endpoints of hospital length of stay, ICU length of stay, or ventilator days.

Table 1 Baseline patient variables and outcomes
Full size table

In univariate analysis (Table 2), PLVI category was not significantly associated with mortality. Only age in years (OR 1.05, 95 % CI 1.01–1.10), comorbidities (OR 1.30, 95 % CI 1.01–1.60), and AIS head (OR 1.44, 95 % CI 1.02–2.02) were found to be associated with mortality. In contrast, low PLVI was strongly associated with development of morbidity (OR 3.66, 95 % CI 1.83–7.32), as was ISS (OR 1.05, 95 % CI 1.01–1.10).

Table 2 Univariate analysis of baseline factors, mortality, and morbidity
Full size table

Multivariable logistic regression analysis (Table 3) did not demonstrate an association between PLVI and mortality, but even after controlling for baseline comorbidities, ISS, and admission SBP, low PLVI was found to be strongly associated with morbidity (OR 4.91, 95 % CI 2.28–10.60).

Table 3 Multivariable analysis of baseline factors, mortality, and morbidity
Full size table

In the unadjusted analysis of morbidity subcategories of complications (Table 4), we found that infectious complications were statistically more likely to occur in the low PLVI group (60 vs. 39 %, p = 0.03). Mechanism of injury also differed between the two groups (Table 5), with patients in the low PLVI category more likely to have sustained a fall (59 vs. 43 %, p < 0.05) and less likely to have sustained an MVC (20 vs. 33 %, p < 0.05).

Table 4 Subgroup complication rates by PLVI group
Full size table
Table 5 Mechanism of injury by PLVI group
Full size table

Discussion

With an increasing elderly population living with greater numbers of chronic medical conditions and decreased physiologic reserve, the concept of frailty is emerging as a significant syndrome in the elderly patient population. Frailty further diminishes the older patient’s ability to compensate when stressed. Fried and colleagues [8] described frailty as a phenotype that includes three of the following five characteristics: unintentional weight loss, slow walking speed, self-reported exhaustion, low physical activity, and weakness as measured by grip strength. The physiology of aging or an inherent decline in various organ systems accounts for this phenotype, and includes increased levels of inflammatory markers (CRP, IL-6), diminution in bone density (osteoporosis), cognitive changes (delirium and dementia), and sarcopenia [13].

In this cohort of elderly, severely injured trauma patients, we found that central sarcopenia as measured by PLVI was associated with morbidity but not mortality. The PLVI has several advantages over other metrics of frailty available for use in the trauma population. Because the majority of blunt trauma patients will undergo cross-sectional imaging upon admission [14], the ability to measure PLVI is present early on for most elderly trauma patients. Since imaging occurs extremely early in the hospital course of trauma patient, admission PLVI is unlikely to be greatly affected by injuries, interventions undertaken to treat injuries, or complications developing thereafter. The admission PLVI can therefore be thought of as a baseline “snapshot” that represents the patient’s condition at the time of injury. Because this metric is objective, does not require patient cooperation or recall, and does not require the ability to ambulate, it may be far easier to assess in patients suffering traumatic injury.

The use of central sarcopenia as a proxy for frailty and predictor of outcomes has been studied in other groups of surgical patients in recent years. Decreased psoas cross-sectional area has been shown to be associated with mortality after liver transplantation [10], elective abdominal aortic aneurysm repair [15], and esophagectomy [16]. In a cohort of patients undergoing hepatic resection for colorectal metastases, Peng et al. [17] found that a total psoas cross-sectional area of <500 mm was associated with increased risk of postoperative complications and hospital length of stay, but not mortality. Of note, our methodology differs from the above-referenced studies in that, rather than reporting the combined cross-sectional area of the psoas muscles at the L4 vertebra (or “total psoas area”, TPA), we report the ratio of psoas CSA to the L4 vertebral body CSA. The intent of this methodology is to provide a PSCA value that is normalized to individual patient body habitus, but further research will be necessary to determine the optimal measurement of central sarcopenia.

Injury Severity Score is one of the best known predictors of mortality in trauma patients, and so it may seem surprising that this failed to predict mortality in our models. However, the distribution of ISS in our cohort is relatively tightly distributed around the median of 24 (range 16–59), with 80 % of our cohort having an ISS between 16 and 32. Given that the predicted mortality in this range of ISS for patients greater than 55 years of age would be expected to be between ~5 and 13 %, it is not entirely surprising, considering the relatively small sample size, that no difference is seen in our model.

In the subgroup analysis, the only category of complications to reach statistical significance between the two groups was infectious complications. Frailty has been associated with alterations in both the innate and adaptive immune system [18–20], and it is thus possible that in the presence of frailty as reflected by low PLVI, host immune responses may be compromised. Alternatively, it may be that the frailty phenotype necessitates interventions which result in increased risk of infections. For instance, patients who are unable to be mobilized from bed may be subjected to prolonged use of indwelling urinary catheters, resulting in increased risk of urinary tract infection.

Despite our findings, this study has some important limitations. The utility of using “any complication” as a combined primary endpoint may be questioned, as different complications have different risk factors which may not be equally distributed between our two groups. Given our relatively small sample size and the degree of complexity that would be introduced in order to control for all of the risk factors for each individual complication, we have chosen to look at an aggregate measure instead. However, the use of morbidity as a combined endpoint may be more meaningful than rates of specific complications, as the aggregate reflects the overall baseline condition and host response to trauma. Additionally, as no specific treatments are available to protect frail patients from posttraumatic morbidity, the therapeutic benefit of a frailty index is currently very limited. However, the addition of central sarcopenia to the traditional predictors of outcome may help inform clinicians when discussing likely outcomes with family patients and their families. Future prospective studies may identify specific opportunities for improvement in this cohort. Finally, this study represents a single-institution experience with a limited number of patients, and as such may not be applicable to other centers. Specifically, the LPVI median of 0.84 from which we derive our cut-point to divide the cohort into “high” and “low” PLVI groups may not be valid in other populations. Further work with larger cohorts may help clarify this issue.

In conclusion, the PLVI is an easily-obtained metric which may help physicians caring for trauma patients to identify a subpopulation of elderly patients at increased risk of morbidity, independent of age and comorbidities. Further investigative efforts should be directed towards translating risk identification into risk reduction.