A comparison of the Morbidities Associated with Different Early Treatments in Tubo-Ovarian Abscess Patients

OBJECTIVE: A tubo-ovarian abscess needs hospitalization and early treatment with parenteral antibiotics only or along with imaging-guided drainage. This meta-analysis juxtaposes between these interventions the length of stay in hospital in days, surgery requirement for those not responding to the initial treatment, and readmission. STUDY DESIGN: The eligible papers searched in various databases (PubMed, Central, Embase, and Scopus) irrespective of their language or date of publication. The Joanna Briggs Institute's Critical Appraisal tool and Cochrane collaboration tool were used to appraise observational and randomized controlled trials, respectively. When a comparable outcome was reported from at least three studies of similar study design, they were included in the meta-analysis (fixed-effect model). Otherwise, outcomes were reported narratively. RESULTS: From 164 studies, five eligible papers (four non-randomized studies and one randomized controlled trials) were reviewed. These studies sourced data from 609 tubo-ovarian abscess patients. Overall, all studies had at least one unclear risk of bias components. The length of stay in the hospital among the tubo-ovarian abscess patients favored the initial parenteral antibiotic only treatment (WMD= -3.26; 95% CI= -4.93 to -1.58; p<0.001; I=80.9%; p-value of Cochranes Q=0.005); however, on sensitivity analysis (meta-analysis with random-effect model) this difference disappeared. Less than three studies of a particular study design reported each of the remaining outcomes. CONCLUSION: The current evidence on how these outcomes vary between the juxtaposed interventions received by the tubo-ovarian abscess patients remains inconclusive due to the inadequate number of good quality randomized controlled trials


Introduction
A tubo-ovarian abscess (TOA) is a serious complication of pelvic inflammatory disease (PID).(1-3) It occurs in 15-30% of women hospitalised with PID. (1,2) It causes the formation of pus and inflammatory mass (of the fallopian tube and (or) ovary) which often manifests clinically as abdominal pain, pelvic mass, fever, and leucocytosis. (1,3) It can sometimes be life-threatening when accompanied by the risk of rupture and consequent severe sepsis. (1,3) To prevent these morbidities and mortality, TOA patients require an early hospitalisation and inpatient care with parenteral antibiotics (often considered as the first-line management) or dual therapy with parenteral antibiotics and imagingguided drainage (ultrasonography (US) or computed tomography (CT) guided chiefly). (1,(3)(4)(5) searchers have demonstrated the usefulness of early use of the above-mentioned dual therapy in TOA patients. (1,3,4) In a study of 302 women with TOA, the latter depicted a nominal need (in about 7% only) of subsequent surgery. (4) The relatively low invasiveness of the drainage makes this dual therapy a well-tolerated treatment for TOA patients.(1) Common routes of TOA drainage by this method are transvaginal and transrectal. (7) Since, both of these treatment modalities require inpatient care, it is important to understand the trade-off between them in terms of the required length of stay in hospital (LOS). While an increase in LOS can raise the risk of unwanted healthcare expenditure and hospital-acquired infections, premature discharges, on the other hand, might hinder the achievement of the desired health outcome and increase the number of emergency room presentations and readmissions. (8)(9)(10)(11)(12) So, a proper insight of inpatient LOS for particular illnesses is essential along with its comparison between different therapeutic modalities.
Therefore, this meta-analysis aimed to compare the average LOS between early inpatient treatment with parenteral antibiotics only and as an adjunct to imaging-guided drainage in TOA patients. Additionally, the need for surgery and readmission before and after discharge from the hospital were explored respectively.

Material and Method
The eligibility criteria for inclusion of studies comprised of the following features-1. TOA patients of any age who received the initial treatment inpatient. 2. Inpatient early treatment should have been compared between parenteral antibiotic/s (irrespective of regimen or dosage or duration of administration) and a combination of parenteral antibiotic/s and imaging-guided (US or CT or both) drainage (primary or salvage done by any route like transabdominal, transvaginal etc.) for the outcomes stated below. 3. Randomised controlled trials (RCT) and non-randomised studies (NRS) were eligible for inclusion. 4. For the respective interventions, the studies should have reported the LOS (in the hospital) in days (the primary outcome). A pre-registered protocol is unavailable for this review.
Succeeding ancillary outcomes in each of the treatment groups, when reported, were also studied-the number of TOA patients who underwent surgery due to failure of the initial inpatient intervention and the frequency of readmission with TOA post-discharge from the hospital. These secondary outcomes were not part of the eligibility criteria.
The definition of TOA and treatment failure was accepted as per the investigators. Surgical removal of an abscess (except by imaging-guided drainage or aspiration) along with the removal of adnexa, uterus, parts of bowel or pelvic clearance was considered as a surgical intervention.
Eligible papers' title and abstract were searched in the following electronic databases with no restriction to language-PubMed, CENTRAL, Embase, and Scopus. An additional search incorporated the bibliography of the papers read in full text. The initial search was done in April 2019. The last date of the repeat PubMed search was 30 March 2020. Following search terms were used for the database searches -TOA or pyosalpinx OR ovarian OR "tubo-ovarian" OR tuboovarian AND drain* OR aspirat* OR transabdominal OR transvaginal AND ultrasonograph* OR ultrasound OR imaging OR "computed tomography" OR CT OR scan AND antibiotic* OR antimicrobial. No filters were applied to narrow down the search. The literature search was not restricted to any date range.
The study selection process closely adhered to the steps commended in the PRISMA flow diagram. (13) The searched output of the electronic database was skimmed through while matching the eligibility criteria of this review. Papers seeming eligible or doubtful for inclusion were selected for full-text reading by the authors. Then, the first author extracted the following data from the papers included in this review, which was subsequently evaluated by the co-author for any unintended errors-1. study profile (first author's last name, year of publication, and country where the study was conducted) 2. study population's information (diagnosis with which patients were admitted, sample size, number of participants in each treatment group, the mean age of study population, inclusion and exclusion criteria, the frequency of attrition before discharge from hospital, and participant consent information). 3. study design 4. intervention received by the treatment groups. 5. the outcomes of interest. The authors of the reviewed papers were not contacted.
Thenceforth, the authors independently assessed the risk of bias in the NRS and RCT, using The Joanna Briggs Institute's Critical Appraisal tool and Cochrane collaboration's tool, respectively. (14,15) Conflict of opinion between the authors was resolved by discussion.
The effect of the two interventions was compared metaanalytically when at least three studies (that are not prone to high risk of bias) reported a statistically comparable outcome data. Data from NRS and RCT were not combined (for metaanalysis).
For LOS, a fixed-effect model was used for the metaanalysis as the compared NRS studies were relatively homogenous (e.g., retrospective cohort study design, conducted in the US, study population's mean age, and the antibiotics used). Furthermore, as the duration of inpatient stay was reported in the same unit (days), weighted mean difference (WMD) between the intervention groups was estimated using the inverse-variance method. (15) Statistical significance of effect estimates was determined at p<0.05 and 95% confidence interval. Heterogeneity was reported with p-value of Cochrane's Q (statistically significant if <0.1) and I 2 statistics (unimportant (0-40%), moderate (30-60%), substantial (50-90%), and considerable (75-100%)). (15) Publication bias was assessed visually using funnel plots. A sensitivity analysis repeated the meta-analysis with a random-effect model and also determined the predictive interval.
For the remaining outcomes, less than three studies were available to compare; therefore, we reported the findings narratively. Likewise, for the RCT, due to lack of additional comparable data from same study design, all of its outcome were reported qualitatively.
All statistical analysis was done with Stata statistical software (StataCorp, College Station, Texas, USA). This paper's reporting follows PRISMA reporting guideline. (13)

Results
The database search produced 164 results (42 PubMed, 87 Scopus, 34 EMBASE, and 1 CENTRAL). Manual search did not retrieve any additional record. After removing the duplicates, we skimmed through 132 titles and abstracts and selected 14 articles for full-text reading. Finally, five articles (published between 1996 and 2016) meeting the eligibility criteria were incorporated in this review. (16)(17)(18)(19)(20) Figure 1 depicts the study selection process.
Regarding the characteristics of the included studies, table I Full-text articles excluded (n = 9) Full-text articles assessed for eligibility (n = 14) Studies included in qualitative synthesis (n = 4)

Studies included in quantitative synthesis
(meta-analysis) (n = 3) Records after duplicates removed (n=132 ) Summary of study population, study design, and intervention Intervention groups=3; group 1-initially receiving antibiotic treatment only (commonly used antibiotics-gentamicin plus clindamycin, cefoxitin plus doxycycline, and a triple antibiotic regimen (ampicillin, gentamicin, and clindamycin)); group 2-iv antibiotics plus US-guided drainage; group 3 -surgical intervention In TOA patients, all studies compared the following inpatient introductory therapies-parenteral antibiotic treatments only versus imaging-guided drainage as a conjunct therapy. (16)(17)(18)(19)(20) All studies compared the LOS in the hospital between the two treatment groups. (16)(17)(18)(19)(20) Clindamycin and gen-tamycin were the two antibiotics reported to be used by all of the studies.(16-0) Only three studies (two NRS and one RCT) reported the requirement of surgeries before discharge from the hospital (due to non-response to initial inpatient treatment). (18)(19)(20) Two studies (NRS) reported the need for readmission after discharge from the hospital. (16,19) Next, the risk of bias of the studies were assessed (Table-2). (16)(17)(18)(19)(20) In the observational studies, the mechanism of confounder handling (16,18,19) and exposure determination (16)(17)(18)(19) (i.e., the rationale clinicians used to ascertain which participant receives which intervention) remained unclear. Additionally, the inter-rater or intra-rater reliability of the interventionists who performed the imaging-guided drainage Farid, 2016, USA (17) Patients diagnosed and admitted with TOA between 2001-12; sample size=113 (met inclusion criteria); treated with antibiotics only initially=61; treated with iv antibiotics and imaging-guided drainage initially=26; treated with iv antibiotics and surgery=26; mean age=40.4 years; inclusion criteria: TOA diagnosed based on imaging and clinical criteria and patients admitted more than 24 hours for treatment; exclusion criteria: TOA diagnosis not matching imaging or clinical diagnostic criteria and non-consenting participants; attrition before discharge from hospital: nil; consent from participants: obtained Retrospective cohort ntervention groups=4; group 1-received iv antibiotics only; group 2 -received iv antibiotics and imaging-guided drainage; group 3-received iv antibiotics and initial surgical intervention only; group 4those failed to respond to antibiotic therapy only were treated with imaging-guided drainage; all patients received antibiotics and the common regiments were-gentamicin/clindamycin, second-generation cephalosporins/doxycycline/metronidazole, fluoroquinolone/metronidazole, and aminopenicillin/fluoroquinolone/metronidazole Perez-Medina, 1996, Spain (20) Patients diagnosed with TOA; sample size=40; treated with antibiotics only initially =20; treated with imaging-guided drainage initially=20; mean age=29 (range 16-49 years); inclusion and exclusion criteria: not clear; attrition before discharge from hospital: nil; consent from participants: obtained RCT Intervention groups=2; group 1 -antibiotics only; group 2 -antibiotics and imaging-guided drainage; both intervention groups received clindamycin and gentamicin

Summary of primary and secondary outcome data
Outcome: duration of hospital stays

Antibiotic group Imaging-guided drainage group
Sample size (n) Mean (in days) SD (in days) Sample size (n) Mean (in days) SD (in days) Goharkhay, 2007(18) Median data, hence, data could not be pooled for meta-analysis.  Unclear risk The randomization method was not described explicitly.

Allocation concealment (selection bias)
Unclear risk It is not clear if any method was used to conceal the allocation of participants from researchers.
Blinding of participants and personnel (performance bias)

Unclear risk
We were not clear if the intervention providing personnel were blinded in some way.
Blinding of outcome assessment (detection bias) Outcome: surgery required Unclear risk It is unclear if the outcome assessors were blinded.
Incomplete outcome data (attrition bias) Outcome: surgery required and duration of hospital stay Low risk The number of TOA patients of each treatment group who required surgery was available for all participants recruited in the study.

Low risk
Although it is not clear if a protocol is available for the trial, the outcomes were reported as per the intentions in the method's section.

Other bias
Unclear risk It is not clear if any intra-rater or inter-rater reliability assessment was done for personnel involved in the imaging-guided drainage method. wasn't clear among NRS. (16)(17)(18)(19) The NRS (16)(17)(18)(19) were at low risk of bias for the following components. None of these studies had attrition in their study population (until discharge from the hospitals) (16)(17)(18)(19) and all identified a list of possible confounders at baseline (like age, sex). (16)(17)(18)(19) Moreover, most NRS had mentioned about pre-defined eligibility criteria for recruiting study participants (16,17,19) and were free of the outcome of interest at the participant recruitment stage of study. (16,18,19) In the RCT,(20) the risk of selection bias was unclear, since trialists did not mention the exact procedure used for randomization or concealing the allocation of interventions from the participants and from the researchers. Furthermore, it is unclear if the intervention providers and the outcome assessors were blind about the interventions received by the participants. (20) The RCT was at low risk of attrition and reporting bias. (20) Summarizing the risk of bias across studies was difficult due to the differences in study designs (NRS versus RCT). (16)(17)(18)(19)(20) Overall, all studies had components of unclear risk of bias. (16)(17)(18)(19)(20) The primary and secondary outcome data from the respective studies are summarised in table 1. (16)(17)(18)(19)(20) LOS: The primary outcome data were available from all of the studies. (16)(17)(18)(19)(20) However, the Goharkhay et al. (2007) study (18) and the RCT (20) were not incorporated in the metaanalysis due to their reporting of median data and being the only interventional study available for a statistical comparison respectively. In contrast to the comparison group (median=4.5 days; range=4-8), in the Goharkhay et al. (2007) study, (18) the LOS in the hospital was longer (p<0.05) in the parenteral antibiotic-treated patients (median=7 days; range=4-16). Similarly, the parenteral antibiotic-treated group of the RCT remained inpatient for a statistically significantly (p<0.001) longer duration (average 9.1 days) than the TOA patients who also received imaging-guided drainage as the initial treatment (mean 3.9 days). (20) The LOS data from the remaining studies (16,17,19) were compared meta-analytically. Meta-analysis using a fixed effect model favoured the treatment group receiving parenteral antibiotics only; however, there was considerable statistical heterogeneity (WMD=-3.26; 95% CI=-4.93 to -1.58; p<0.001; I2= 80.9%; p-value of Cochranes Q= 0.005) (Figure-2a). We could not rule out publication bias as the visual inspection of the funnel plot suggested asymmetry (Figure-2b).
The sensitivity analysis didn't replicate the findings of the preliminary meta-analysis. It did not find any statistically significant difference in LOS between the compared interventions (WMD= -3.053; 95% CI= -6.90 to 0.80; p=0.120) (Figure-2a). The predictive interval (95% CI= -49.24 to 43.13) suggested that a future study might find the combination therapy favorable in reducing the length of hospital stay than the parenteral antibiotics therapy only.
age of 35 years. While the way of handling confounders and determining exposures was unclear among the NRS, (16)(17)(18)(19) the risk of selection bias, performance bias, and detection bias was not clear in the RCT. (20) The meta-analytic comparison of LOS of three NRS (16,17,19) is not robust as they are not replicable in sensitivity analysis and have considerable unexplained heterogeneity.
The quality of the LOS-related meta-analysis finding was determined using the GRADE approach proposed by the GRADE Working Group (2004). (21) The evidence was double downgraded to low-quality evidence because it came from studies of weaker epidemiological design (observational) with unexplained heterogeneity and unclear risk of bias.
To compare the findings of this paper with existing evidence in this background, different databases (PubMed, CEN-TRAL, and Prospero) were explored for any existing, registered, or ongoing reviews. However, no such reviews were available for contrasting. Henceforth, the key strength of this paper is its conceptual uniqueness. Besides, this review is likely to be comprehensive as the database search was not restricted to any language or date range.
The primary implication of this paper is that it identifies an area of gynecological research where there is a paucity of good quality RCTs which produces an evidence gap in the context. Future, trialists may find our study useful to plan better RCTs that can address this knowledge gap. Additionally, health care professionals like gynecologists and interventional radiologists may find this review as a brief overview of some of the aspects of TOA management.
Nonetheless, this review has certain limitations. Since the reviewed studies used different antibiotic dosages and regimens, we could not extricate if these played any role in the outcomes observed. Then, at the outcome level, it was not clear how the outcomes of interest were defined and measured by the reviewed studies (Table 2). Lastly, at the study level, most studies reviewed in this paper were of weaker study design (i.e., NRS). (16)(17)(18)(19) The only RCT also suffered from certain weaknesses like small sample size, single-centric design, and unclear risk of biases. (20)

Conclusion
The evidence regarding if the LOS, surgery requirement, and the frequency of readmission in TOA patients vary between the initial parenteral antibiotic therapy recipients and combined parenteral antibiotics and imaging-guided drainage recipients remain inconclusive. Presently, there is a critical shortage of adequately powered large multicentric RCTs addressing the context.