Treatment outcomes in patients with pyogenic vertebral osteomyelitis who have cirrhosis

Early diagnosis and proper treatment of pyogenic vertebral osteomyelitis (PVO) in patients with cirrhosis is challenging to clinicians, and the mortality rate is expected to be high. A retrospective study was conducted to investigate the treatment outcome in PVO patients with cirrhosis and to identify the predictors of their mortality. Mortality was divided into two categories, 30-day and 90-day mortality. A stepwise multivariate logistic regression model was used to identify predictors of mortality. Eighty-five patients were identified after initial exclusion. The patients’ mean age was 60.5 years, and 50 patients were male. The early mortality rates within 30 and 90 days were 17.6% and 36.5%, respectively. Multivariate analysis revealed that increased age, CTP class C, and bacteremia at the time of PVO diagnosis were predictors of 30-day mortality, while higher MELD score, presence of combined infection, and multiple spinal lesions were predictors of 90-day mortality. Attention should be paid to the high mortality between 30 and 90 days after PVO diagnosis (18.8%), which was higher than the 30-day mortality. Liver function was consistently a strong predictor of mortality in PVO patients with cirrhosis. The high-risk patients should be targeted for an aggressive diagnostic approach, using spinal MRI and intensive monitoring and treatment strategies.

The mortality in liver cirrhosis was reported to be greater than that in the five major cancers 1 . Infection further increases the mortality of patients with cirrhosis by fourfold 2 , and infection is directly responsible for 30-50% of deaths in patients with cirrhosis 3,4 . Considering the greatly increased mortality from infection in patients with cirrhosis 2 , early diagnosis and prompt treatments should be compulsory to save patients' lives. However, adherence to such a basic principle for patients with cirrhosis is not easy for clinicians engaged in the treatment of pyogenic vertebral osteomyelitis (PVO).
A retrospective study reported that 77% of patients with end-stage liver disease had bodily pain, in the abdomen, back, head/neck, and upper and lower extremities, within 24 hours of the evaluation, and most patients (90%) received various analgesics 5 . Such a high prevalence of bodily pain in patients with cirrhosis prevents the use of the clinical symptoms of PVO as an indicator for early diagnosis. In addition, bacterial infections are common in immunocompromised patients with cirrhosis 6 , and patients with cirrhosis have a fivefold greater risk of developing infection than the general population 7 . Such high prevalence of bacterial infection in patients with cirrhosis also prevents the use of infection markers as indicators for early diagnosis of PVO. According to the guideline of the Infectious Diseases Society of America, spine magnetic resonance imaging is recommended for patients with suspected PVO who have new or worsening back pain and elevated erythrocyte sedimentation rate or C-reactive protein level 8 . However, such an approach is considered to have limitations in PVO patients with cirrhosis.
Treatment of PVO in patients with cirrhosis is challenging for clinicians. Attenuated liver function by bacterial infection threatens the life of patients with cirrhosis through variceal rupture 9 and multiorgan failure 10 . Treatment failure or recurrence is expected to be high, owing to cirrhosis-associated immune dysfunction 11 and the high prevalence of multidrug resistant organisms 12 . The strictly required long-term intravenous antibiotics 8 to reduce the recurrence of PVO potentially can paradoxically cause Clostridium difficile infection, which is associated with higher mortality 13 . Decreased bone mineral density with deteriorated bony microarchitecture in patients with cirrhosis 14 , disuse osteoporosis caused by immobilization, and long-term hospitalization aggravates skeletal destruction by the pyogenic organism, and can easily cause neurological and structural instabilities that require surgical treatment. However, the basic principles in the surgical treatment of PVO 15 , including sufficient removal  of paraspinal abscesses and firm spinal instrumentation, are technically challenging in patients with cirrhosis who  have poor bone quality and bleeding tendency with coagulopathy. As a result, difficulty in early diagnosis and prompt treatments in PVO patients with cirrhosis is expected to be related to poorer clinical outcomes, including higher mortality. However, to our knowledge, no reports have described the treatment outcome in this patient group. In addition, under the expected higher mortality, prognostic studies to identify high-risk patients, on whom intensive monitoring and treatment strategies should be concentrated, are essential for the improvement of clinical outcome. We performed a retrospective study to investigate the treatment outcome in PVO patients with cirrhosis, and to identify the predictors of their mortality.

Methods
Study design and ethics. A retrospective case review was performed in patients with cirrhosis who received treatment for PVO in our institution between January 2000 and March 2018. This study was designed and conducted using the format recommended by STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines 16 . This study was approved by the institutional review board of Hallym University Sacred Heart Hospital. The institutional review board waived the informed consent for this study. All methods were carried out in accordance with the relevant guidelines and regulations.

Study patients.
Our university medical center is one of the largest medical institutions in our country, consisting of six general hospitals. This study was performed in the main institute among the six general hospitals. As the main institute of our medical center, our hospital serves as a tertiary referral center for not only the other five general hospitals but also numerous local hospitals.
Patients with cirrhosis who received treatment for PVO were eligible for inclusion. Cirrhosis was diagnosed by liver pathological examination or a combination of laboratory biochemical, radiological, and endoscopic findings, if a liver biopsy result was not available 14 . PVO was defined using the following criteria: suggestive clinical symptoms, accompanying typical radiological features on MRI, and microbiological identification 15 . Microbiological confirmation included isolation from blood culture, CT-guided needle biopsy, or surgical biopsy. Patients were excluded if their medical records indicated that they had received a previous liver transplantation before the diagnosis of PVO. Patients were also excluded if they had received a previous spine surgery using instrumentation at the same site of the PVO. Other reasons for exclusion were incomplete medical records or imaging data. Data collection. Data were retrieved from electronic medical records using a standardized collection form.
Demographic, laboratory, and other clinical data at the time of PVO diagnosis were ascertained. Medical history was retrieved from the records, and the Charlson comorbidity index was calculated to assess comorbid medical conditions 17 . The presence of ascites, encephalopathy, and gastrointestinal (GI) bleeding at the time of PVO diagnosis were retrieved from the records, and laboratory data at the time of PVO diagnosis were retrieved. Then, liver function was determined using the Model for End-Stage Liver Disease (MELD), Child-Turcotte-Pugh (CTP) class, and CTP scores. The severity of infection at the time of PVO diagnosis was retrieved using a validated classification system by Pola et al. 18 , which divided pyogenic spondylodiscitis into three types as follows: 1) type A, cases without biomechanical instability, neither acute neurological impairment nor epidural abscesses; 2) type B, cases with radiological evidence of significant bone destruction or biomechanical instability without acute neurological impairment or epidural abscesses; and 3) type C, cases with epidural abscesses or acute neurological impairment.
Definitions. The presence of combined infection was retrospectively retrieved from the medical records and classified as follows: (1) Intra-abdominal infection: Spontaneous bacterial peritonitis was diagnosed on the basis of an ascitic fluid neutrophil count of >250/mm 3 or a positive bacteriological culture of the ascitic fluid 19 . Infectious enterocolitis was diagnosed in patients with diarrhea and leukocytes in stool or positive stool culture for pathogens, including Salmonella, Shigella, Yersinia, Campylobacter, and pathogenic Escherichia coli, or a positive Clostridium difficile stool assay 7 .

Results
Baseline patient characteristics. Eighty-five patients were identified after the initial exclusion ( Fig. 1).
The patients' mean age was 60.5 years, and 50 patients (58.9%) were male (  (Table 1). Combined infection was more frequent in the CTP class C patients (p = 0.024).
Surgical treatment was performed in 10.6% (9 of 85 patients) within one week of PVO diagnosis, and 29.4% between one and four weeks after PVO diagnosis (25 of 85 patients).

Effect of early surgery on 30-or 90-day mortality: logistic regression analysis. Early surgery
was performed in 34 patients (40.0%), of whom 9 (26.5%) underwent spinal instrumentation. The multivariate logistic regression analysis revealed that early surgical treatment was not associated with a statistically significant improvement in 30-or 90-day mortality (Table 4). A model (model 2 in Table 4) adjusted for all significant variables in the univariate analysis (Table 3) only showed a significantly lower odds ratio (0.005) for 30-day mortality in the patients who had an early surgery (p = 0.012).

Treatment outcomes in patients with at least 90-day survival.
Surgical treatment was performed in 51.9% (28 of 54 patients) of the survivors, and instrumentation was performed in 37% (20 of 54 patients) of the survivors ( Table 5). The duration of antibiotic treatment and the length of hospital stay (from the PVO diagnosis) was longer in CTP C patients, however they were statistically insignificant ( Table 5). Recurrence of PVO was identified in 11 patients (20.4%) and was more common in CTP C patients (p = 0.028) ( Table 5).

Discussion
As the first study to investigate the treatment outcome of PVO patients with cirrhosis, our study demonstrated that the 30-and 90-day mortality rates were 17.6% and 36.5%, respectively ( Table 2). Multivariate analysis revealed increased age, CTP class C, and bacteremia at the time of PVO diagnosis as predictors of 30-day mortality, whereas higher MELD score, presence of combined infection, and multiple spinal lesions were predictors for 90-day mortality (Table 3). Early surgery did not lead to meaningful differences in the survival of the PVO patients with cirrhosis with respect to early mortality (Table 4).
Previous studies reported the early mortality of PVO patients, including in-hospital mortality or 90-day mortality ranging from 2.8% to 16.8% [27][28][29][30][31][32] . A recent study investigating the clinical outcome of PVO patients with hemodialysis reported an in-hospital mortality of 14.9% and 1-year mortality of 22.4% 15 . Compared with the results of previous studies, the mortality in our PVO patients with cirrhosis was considerably higher. The remarkable finding was the increased mortality observed between 30 and 90 days after PVO diagnosis (18.8%, 16/85 patients; Table 2), which was higher than the 30-day mortality (17.6%; Table 2). Although, inferring the cause of the higher mortality between 30 and 90 days after PVO diagnosis is beyond the scope of our study, we could explain the cause as follows: first, PVO patients generally require long-term hospitalization for the administration of intravenous antibiotics for at least 6 weeks 8 , which paradoxically increases the risk of hospital-acquired infections such as Clostridium difficile infection or the risk of recurrence by multidrug-resistant organisms. In addition,  www.nature.com/scientificreports www.nature.com/scientificreports/ long-term intravenous antibiotics can attenuate liver or kidney function, which negatively influences the survival of patients. Second, pain and disability from the spinal structural instability negatively influences survival. During PVO treatment, significant bone loss occurs directly by causative organisms and indirectly by disuse-type bone loss [33][34][35] . Such bone loss can induce structural instability, which leads to neurological deficit, spinal deformity, and even death 36,37 . Therefore, permanent, and extensive stabilization using spinal instrumentation is often required after debridement or neural decompression. However, such long instrumentation is technically demanding in patients with cirrhosis, and it even fails in such patients with osteoporosis and progressive bone loss 38 .    (Fig. 2), and these results are in line with the results of other types of infection in patients with cirrhosis 39,40 . The multivariate analysis identified CTP class and MELD score as significant predictors of 30-day and 90-day mortality, respectively (Table 3). Within 30 days after PVO diagnosis, only one of the patients with CTP class A or B died (2.4%, Table 2). However, one third of the patients with CTP class C died within 30 days (31.8%, Table 2). The Charlson comorbidity index score did not show a significant association with 30-day (p = 0.125; odds ratio, 1.138), 90-day mortality (p = 0.258; odds ratio, 1.083), and late mortality (p = 0.931, odds ratio, 0.994).

Categories of mortality
In addition to liver function, the significant predictor of mortality in PVO patients with cirrhosis was the gross extent of infection indicated by the presence of multiple spinal lesions and combined infection ( Table 3). The diagnosis of PVO is frequently delayed in clinical practice 41 . Unfortunately, such delayed diagnosis of PVO is believed to cause extensive musculoskeletal involvement of the spine and neurological and structural instabilities in patients with cirrhosis. Approximately one-third of the cohort (34.1%; Table 1) had multiple spinal lesions, and two-thirds of the cohort (61.2%; Table 2) had extensive spinal involvement beyond 3 vertebral bodies. According to the classification of Pola et al., 94.1% of the cohort (80/85 patients) had structural instability (type B) or neurological compromise (type C), which theoretically requires surgical treatment 18 . Compared with the results of previous reports 15,42 , our results showed that PVO patients with cirrhosis are considered to have an even more extensive spinal involvement than other groups of PVO patients. We hypothesized that immune dysfunction 11 and impaired bony microarchitecture 14 contributes to aggressive infection.
Combined bacterial infection is common in patients with cirrhosis 7 and reported to be closely related to high short-term mortality 7,43 . Therefore, combined infection should be considered in studies about infection-related treatment outcome in patients with cirrhosis. In our study, combined infection presented in more than half of the patients (56.5%; Table 1), and the most common combined infection was urinary tract infection (30.6%; Table 1). The multivariate analysis confirmed that combined infection is closely related with the mortality of PVO patients with cirrhosis (Table 3). In this respect, clinicians should pay great attention to the presence of combined infection in PVO patients with cirrhosis. If PVO patients with cirrhosis are considered to have combined infection in other organs or if patients with cirrhosis are receiving treatment for infection in other organs show symptoms or signs of PVO, clinicians should be aware that such a combined infection is closely related to patient survival.
The establishment of prognostic factors related to mortality should be connected to early treatment strategies. In our study, aged patients with advanced cirrhosis who had combined infection or multiple spinal lesions were identified to have high mortality rates. Therefore, this group of patients should be targeted for an aggressive diagnostic approach using spinal MRI and intensive monitoring and treatment strategies. If the diagnosis is established, broad-spectrum antibiotics should be administered as early as possible; this is a prerequisite to decrease the burden of infection and to prevent early mortality in PVO patients with cirrhosis. In this respect, early surgical drainage can be theoretically suggested as a possible treatment option to rapidly remove epidural and intraosseous abscesses, which occur in relatively avascular areas where antibiotics cannot easily reach and require long-term intravenous antibiotic administration 8 . However, early surgical treatment did not show a statistically significant outcome in our study ( Table 4). The survival of the PVO patients with cirrhosis was strongly influenced by their liver function (Table 1), and early surgical treatment was believed to be insufficient for a clinically significant decrease in infection burden in these patients with such wide extent of combined or multiple spine infection (Table 1). However, a multivariate analysis revealed a significantly lower odds ratio for 30-day mortality in patients with early surgery (odds ratio, 0.002, p = 0.018, model 2 in Table 4). A large-scale multicenter study is required to confirm the effect of early surgery on the survival of PVO patients with cirrhosis.
The main limitation of our study is its retrospective design, and some unidentified confounders may have influenced the clinical outcomes of our patients. Precise clinical factors including the method of antibiotic treatment, method of surgical treatment including spinal instrumentation, surgery-related complications may have influenced the treatment outcomes, especially mortality, of our cohort. However, owing to the high early mortality in our cohort and small sample size, inclusion of such various clinical factors to estimate their association with clinical outcome was difficult. Next, due to limited population of our cohorts, we only investigated association  Table 5. Treatment outcomes in patients with at least 90-day survival. Data were presented by number (%) of patients or mean ± standard deviation.