The prediction of surgical intervention in patients with tubo-ovarian abscess

Abstract The aim of this study was to compare the clinical characteristics of patients with tubo-ovarian abscess (TOA) who responded to medical treatment and those who underwent surgical intervention due to medical treatment failure. Electronic medical records were evaluated retrospectively to identify patients who were diagnosed with TOA. Demographic, clinical, and laboratory data including white blood cell (WBC) count, erythrocyte sedimentation rate (ESR), and C-reactive protein (CRP) were compared between the medical treatment group and the surgical intervention group. Patient age, TOA diameter, WBC count, CRP, and ESR were significantly different between the groups. On multiple regression analysis, significant correlations were identified between age (p = .001), ESR (p = .045), and failure of medical treatment. TOA diameter (p = .065) showed a borderline association with surgical intervention. The risk of needing surgical intervention in TOA patients can be predicted using ESR in addition to age and TOA size as risk factors. IMPACT STATEMENT What is already known on this subject? For patients diagnosed with a tubo-ovarian abscess (TOA), the size of TOA and the patient’s age are helpful for early identification of patients who are likely to need surgical treatment. Inflammatory markers such as C-reactive protein and white blood cell are also associated with the risk of surgical intervention. What do the results of this study add? Erythrocyte sedimentation rate (ESR) in addition to the size of TOA and the patient’s age is a useful marker in determining whether to undergo surgery in patients with TOA. What are the implications of these findings for clinical practice and/or further research? ESR combined with the patient’s age and the size of TOA is clinically useful in predicting the need for early surgical intervention in patients with TOA. Large prospective controlled studies are required to establish relationship between inflammatory markers and the risk of surgical intervention.


Introduction
Pelvic inflammatory disease (PID) causes inflammation in the uterus, ovaries, fallopian tubes, and adjacent pelvic organs including bowel (Granberg et al. 2009). Tubo-ovarian abscesses (TOAs) may form in cases of severe PID and can lead to sepsis if the abscess ruptures. TOAs are thought to occur when pathogens spread from the cervix to the endometrium and through the fallopian tube into the peritoneal cavity, causing pelvic peritonitis and abscess formation. TOA affects sexually active women of reproductive age and is associated with repeated hospitalisation, infertility, ectopic pregnancies, pelvic adhesions, and chronic pelvic pain (Granberg et al. 2009;Gradison 2012;Bugg and Taira 2016).
To diagnosis TOA in patients with suspected PID, imaging methods such as ultrasound, CT, and MRI are used. Gynaecologic ultrasound is the most common imaging modality for differentiating between haemorrhagic ovarian cysts, endometriosis, and TOA in patients with clinically suspected TOA. Treatment of TOA includes broad spectrum antibiotics or surgical interventions such as drainage or more invasive procedures. Broad spectrum antibiotics targeting anaerobes and gram-negative aerobes is the first line treatment and is effective in 70% of TOA patients (Wiesenfeld and Sweet 1993;McNeeley et al. 1998). Surgical intervention is needed in approximately 30% of TOA patients who do not respond to antibiotics or whose abscesses rupture. In patients diagnosed with TOA, prediction of antibiotic treatment failure may help in the early identification of patients with a high likelihood of needing surgical treatment. Body temperature, white blood count (WBC), erythrocyte sedimentation rate (ESR), and C-reactive protein (CRP) are useful in determining the severity of the inflammation. Dewitt J et al (Dewitt et al. 2010) reported that high WBC counts or high CRP was associated with risk of surgical intervention. However, the inconsistency exists among these studies (Akkurt et al. 2015;Alay et al. 2019). The aim of this study was to determine the clinical characteristics including inflammation marker associated with the need for surgical intervention.

Methods
This retrospective study was performed at the International Saint Mary's Hospital, Incheon, South Korea. The electronic medical records of patients admitted with suspected tuboovarian abscess (TOA) between March 2014 and June 2019 were reviewed. The Institutional Review Board of Catholic Kwandong University, International Saint Mary's Hospital approved this study as a minimal risk and waived the need for informed consent (IS19RISI0057).
A diagnosis of TOA was made in patients who presented with lower abdominal and pelvic pain, vaginal discharge, and cervical motion tenderness during gynaecologic examination. Additional findings included fever; elevated white blood cell (WBC) count, erythrocyte sedimentation rate (ESR), and Creactive protein (CRP); and a complex adnexal mass on pelvic ultrasonography, computed tomography, or magnetic resonance imaging. Hospitalisation and intravenous antibiotics were recommended for all patients diagnosed with TOA. The antibiotic regimen included ceftriaxone 2.0 g iv every 24 h or ceftriaxone 1.0 g iv every 12 h, metronidazole 500 mg iv every 8 h, and isepamicin 400 mg iv every 24 h.
Surgical intervention was considered if the patient's symptoms did not improve after 72 h of antibiotic administration. Specifically, the criteria for considering surgical treatment were: persistent fever, enlarging TOA, worsening pelvic tenderness, and acute clinical deterioration such as any sign of sepsis, ruptured abscess, or generalised peritonitis. Surgical treatment was also considered to rule out malignancy when the patient was elderly and the tumour marker levels remained high even though symptoms improved after antibiotic treatment.
Conservative treatment was provided with follow-up imaging if the clinical signs improved and tumour marker levels decreased. Postoperative antibiotics were continued until clinical improvement was achieved.
Demographic, clinical, and laboratory data were compared between patients who were managed conservatively and patients who underwent surgical treatment. Demographic data included age, parity, body mass index (BMI), marital status, menopause status, smoking, alcohol, number and type of previous pelvic surgery procedures, and medical history of intrauterine device (IUD) insertion or previous PID. Clinical and laboratory data included vital signs, length of hospital stay, mass location, imaging modality for diagnosis, the largest diameter of the TOA mass, and results of blood panels including a serum tumour marker (CA 125) and inflammatory markers such as ESR and CRP. The TOA size was defined as the largest diameter measured by transvaginal ultrasonography.
Peak fever was defined as the highest body temperature during treatment in the conservative group and the highest preoperative body temperature in the surgical intervention group. The highest WBC counts (Â10 3 /mL), ESR, CRP, and neutrophil and lymphocyte percentages were selected during hospitalisation in the conservative treatment group and the highest values before operation were selected in the surgical intervention group. Absolute neutrophil count (ANC) and neutrophil-to-lymphocyte ratio (NLR) were calculated from the complete blood cell count (CBC) and recorded. In the surgical intervention group, type and technique of surgery and pathology results were described.
The data were entered into Microsoft Excel and analysed using SPSS statistical software, version 20 (SPSS Inc, Chicago, IL, USA). Median and interquartile range were used for descriptive statistics. The characteristics of the two groups were compared using Student's t-test and the v 2 test. Fisher's exact test was used for categorical variables with small samples. A binary logistic regression was performed for significant variables in the univariate and multivariate analysis to assess the correlation between clinical findings and the risk of surgical intervention in patients with TOA. All statistical tests were 2-sided, and a p value of <.05 was considered significant. A receiver operating characteristic (ROC) curve was used to determine the cut-off of clinical characteristics with regard to the need for surgical intervention. The 95% confidence interval (CI) of the area under the curve (AUC) was determined. Optimum diagnostic threshold was chosen by using coordinates from the ROC curve. The sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), and odds ratio of significant variables were calculated to predict the risk of surgical intervention in patients with TOA. Significant variables in the multivariate analysis were also combined to maximise the predictability of surgical intervention.

Results
During the study period, 244 patients were diagnosed with PID. Of these, 72 patients received a final diagnosis of TOA; 40 were treated with intravenous antibiotics (conservative treatment group), and 32 did not respond to medical treatment and needed further surgery or abscess drainage (surgical interventional group). Table 1 shows the patient demographic and clinical data for the two groups. The overall median age was 31.1 years (interquartile range: 25.2-42.5 years) in the conservative treatment group and 47 years (interquartile range: 41.8-50 years) in the surgical intervention group; the difference was significant (p < .001). There were no significant between-group differences in BMI, IUD use, history of previous PID, alcohol, or smoking. The numbers of patients who were nulliparous (p < .001), and premenopausal (p ¼ .003) were significantly higher in the conservative treatment group.
In the 32 patients comprising the surgical intervention group, laparoscopy and laparotomy were performed in 17 (53.1%) and 14 (43.8%) patients, respectively. One patient was treated with abscess drainage via culdotomy and pigtail insertion, 15 patients underwent unilateral salpingectomy or unilateral salpingo-oophorectomy, and 12 patients underwent bilateral salpingectomy or bilateral salpingo-oophorectomy. Four cases of total hysterectomy in addition to adnexal surgery were identified. Four patients underwent appendectomy because the TOA spread to the periappendiceal area. One patient underwent low anterior resection because the TOA resulted from colon cancer with perforation. Fistulectomy was performed in one patient with a fistula between the TOA and abdominal skin. Two patients were diagnosed with actinomycosis.
Univariate analysis was performed by binary logistic regression. Multivariate analysis was performed using variables found to be significant in the univariate analysis. As shown in Table 2, five variables remained significantly associated with the risk of surgical intervention: age (OR, 1.134; 95% confidence interval [CI], 1.065-1.209); mass size (OR, 1.947; 95% CI, 1.313-2.887); ESR (OR, 1.037; 95% CI, 1.016-1.059); CRP (OR, 1.012; 95% CI, 1.004-1.019); and WBC (OR, 1.000; 95% CI, 1.000-1.000). Only age (p ¼ .001) and ESR (p ¼ .045) remained significant variables in the multivariate analysis. Mass size showed borderline significance (p ¼ .065). Table 3 shows cut-off values and AUC based on the ROC curve for variables found to be significant in the univariate analysis. Sensitivity, specificity, accuracy, PPV, and NPV were calculated using the cut-off values. As a single variable, age (cut-off value > 34.3 years and AUC 0.822) showed the highest sensitivity (93.8%) and NPV (87.5%) for predicting surgical intervention in TOA patients. However, the specificity of age was only 35%. The accuracy of mass size (cut-off value > 5.9 cm, AUC 0.780) was the highest and ESR (cut-off value 45 mm/h and AUC 0.750) showed the highest PPVs (Figure 1).
To increase the predictability, age and ESR, which were significant variables in the multivariate analysis, were combined. The combination index of age > 34.2 years and ESR > 45 mm/h had an accuracy of 81.9%, which was higher than that of the single variables. Risk scoring was performed based on age, ESR, and mass size. Each variable with a value higher than the cut-off value received a score of 1, and the combination of variables resulting in a total score of more than 2  points was categorised as the risk group. The AUCs of the risk group and combination index were 0.844 (sensitivity 93.8%, specificity 75.0%, accuracy 83.3%, PPV 75%, and NPV 93.8%) and 0.819 (sensitivity 81.3%, specificity 82.5%, accuracy 81.9%, PPV 78.8%, and NPV 84.6%), respectively ( Figure 2).

Discussion
Early surgical intervention can reduce morbidity and mortality in TOA patients who do not respond to antibiotics. Several studies of demographic and clinical characteristics associated with antibiotic treatment failure in TOA patients have shown that the TOA size and age of the patient are significantly related and proportional to the need for surgical treatment (Halperin et al. 2003;Dewitt et al. 2010;Habboub 2016;Kinay et al. 2016;Alay et al. 2019;Fouks et al. 2019). Early recognition of a patient's risk of antibiotic treatment failure alerts clinicians to the need to change the treatment strategy, thereby reducing acute and long-term complications of TOA.
In accordance with the results of previous studies, older patients (over 34 years) with large TOAs (more than 5.9 cm) were less likely to respond to antibiotics and more likely to undergo invasive treatment. Age was the strongest predictor of failed antibiotic treatment in the present study. TOA was associated with increased ESR, CRP, and WBC. Previous studies reported that these biological markers help to predict response in TOA patients treated with antibiotics. High WBC counts or high CRP have been reported to lead to antibiotics failure (Dewitt et al. 2010). However, the results are inconsistent. Akkurt et al. (2015) found no difference in CRP levels between TOA patients who responded to antibiotics and those who did not. Yildirim et al. (2015) reported that WBCs and neutrophils were elevated in TOA patients, and neutrophils were elevated even when the WBC count was normal; they also found that lymphocytes were relatively decreased, resulting in an increase in the neutrophil-tolymphocyte ratio (NLR). Alay et al. (2019) reported that neutrophil counts were significantly higher in TOA patients with no response to medication and that the NLR could be used to predict non-response to medication in TOA patients.
Most of these studies were evaluated based on the patient's clinical characteristics on the day of admission (Alay et al. 2019;Fouks et al. 2019). However, some patients who  did not have a fever on the day of hospitalisation may have had a fever later. CRP and ESR levels can also increase after hospitalisation and starting antibiotics. WBC and neutrophil counts also may increase on the second or third day of hospitalisation. Therefore, we analysed the highest values recorded during hospitalisation. Because the level of inflammation increases due to surgery, we used the highest values obtained before surgery in the surgical intervention group. In our study, ESR, CRP, and WBC count were statistically significant in the univariate analysis, but only ESR was significant in the multivariate analysis. Neutrophil counts and the NLR had only borderline significance in the univariate analysis. Simultaneous consideration of multiple variables can improve predictability. Fouks et al. scored risk by weighting age at hospitalisation (> 35 years), abscess size (! 70 mm), WBC count (> 16,000/mm 3 ), and bilaterality (Fouks et al. 2019). They excluded the CRP value because it was not significant in the univariate analysis. Four groups were defined based on the risk score, and the risk of antibiotic treatment failure was positively correlated with a higher risk score. In the highest risk group, there was a 92% chance of invasive treatment such as surgery or drainage. We reconfirmed that age (> 34.3 years) and the size of the abscess (> 59 mm) are risk factors, and the cut-off values are also similar. The risk group sensitivity in our study was 93.8%.
We observed higher mean values for WBCs, CRP, and ESR (all inflammatory markers) in the surgical intervention group than in the conservative group. The frequency of fever above 38 C and peak fever during admission were not statistically different between the groups, which accords with the results of previous studies (Lee et al. 2015;Fouks et al. 2019). To our knowledge, this is the first report of the value of inflammatory markers obtained during hospitalisation for predicting antibiotic treatment failure in patients with TOA. Previous studies were based on inflammatory markers obtained on the day of admission (Alay et al. 2019;Fouks et al. 2019) and mainly referred to WBC and CRP as useful inflammatory markers for predicting conservative treatment failure. In this study, the WBC, CRP, and ESR levels were all higher in the surgical intervention group than in the conservative treatment group. However, only ESR was a significant predictor in the multivariate analysis. This difference may be due to the fact that, unlike previous studies, our study used the highest inflammatory marker levels recorded during the entire hospital stay in the conservative treatment group and during hospitalisation before surgery in the surgical intervention group.
ESR and CRP are widely used inflammatory markers. Because of their lack of sensitivity and specificity, ESR and CRP should be interpreted in the context of clinical history and physical exam for the diagnosis of TOA. Discrepancies between ESR and CRP measurements are common, especially in chronic inflammatory diseases (Bray et al. 2016). TOA is a late complication of PID. CRP is preferred as a serologic marker for acute inflammatory conditions (Keenan et al. 2008). CRP levels tend to increase rapidly after the onset of inflammation and drop quickly with the administration of antibiotics (Shusterman et al. 1985). The ESR changes more slowly than the CRP level during antibiotic therapy (Shusterman et al. 1985;Litao and Kamat 2014), which may explain why ESR was more useful for predicting surgical intervention than CRP in this study.
In conclusion, the addition of ESR to the previously identified risk factors of age and TOA size predicts the likelihood of surgical intervention. The limitations of this study relate to its retrospective design and small study population derived from a single institution. Future prospective controlled studies are required to verify our findings.