Biomarkers for diagnosis of acute appendicitis in adults

Alternative test(s)

Standard diagnostic practice involves history taking, clinical examination and often a period of active observation to elicit progression of symptoms (Andersson 2007a). Clinical scoring systems may stratify risk at this stage. However, these have been poorly validated and have not been widely adopted (Wilasrusmee 2014). Biochemical markers of inflammation, conventionally WCC and CRP, are employed to help inform the diagnosis.

Imaging modalities such as ultrasonography (US) (Van Randen 2008), computerised tomography (CT) and magnetic resonance imaging (MRI) may be utilised (Cobben 2009; Rud 2012). However these modalities have cost implications and varying availability. The diagnostic accuracy of US is operator-dependent and performances reported from high volume centres have not been reproduced in other healthcare settings (Toorenvliet 2010). A meta-analysis directly comparing graded compression US with CT in 671 predominantly adult patients gave an estimated sensitivity and specificity of 78% and 83%, respectively for US (Van Randen 2008). The poor sensitivity of US in adults as compared to children is in part due to increased adipose tissue with age (Doria 2006).

CT has the advantage of being less operator-dependent and interpretation is not impaired by bowel gas pattern, adipose tissue or patient discomfort. In adults, the sensitivity of CT is higher than US, with an estimated sensitivity and specificity of 91% and 90% in the aforementioned meta-analysis (Van Randen 2008). A disadvantage of CT is the associated ionising radiation exposure, which is especially undesirable in young patients, with concerns of increasing lifetime risk of cancer (Brenner 2007), and is relatively contraindicated in pregnancy.

The use of MRI in the diagnosis of appendicitis is a relatively recent advance and is not widely implemented. MRI protocols for appendicitis and training for interpretation of scans are under development (Cobben 2009). However, MRI is currently recommended by the American College of Radiology in suspected appendicitis in pregnancy when an US examination is negative or inconclusive and where ionising radiation should be avoided (Leeuwenburgh 2012).

The reference standard for appendicitis is histological examination of the excised appendix, giving a dichotomous marker with which to compare pre-operative diagnostic tests.

Types of studies

We will include prospective observational clinical studies measuring the accuracy of blood biomarkers relative to the reference standards for the diagnosis of appendicitis. We will exclude retrospective and diagnostic case-control studies, case reports and reviews. Studies must include a histological confirmation of acute appendicitis in operatively managed patients. The minimum dataset sufficient to construct a two-by-two contingency table of true positive, true negative, false positive and false negative rates, or data sufficient to derive these values, must be available for meta-analysis.

Participants

Participants will include adults with clinically suspected acute appendicitis. Clinically suspected acute appendicitis comprises those patients admitted to hospital with right iliac fossa pain in keeping with appendicitis, and patients undergoing appendicectomy for suspected appendicitis.

Adults are considered separately from children due to varying normal ranges of biomarkers between the two groups. The definition of adult varies in the literature from over 14 to over 18 years of age. Therefore studies that include patients younger than 14 will be excluded. It is important to evaluate the diagnostic test on the same population on which the test is applied in clinical practice as patient spectrum affects test performance characteristics (Bentley 2012).

We will exclude studies conducted in immunocompromised patients due to a theoretical alteration in biomarker response in this group. Populations from countries in which neglected tropical diseases (NTDs) are endemic will be excluded, due to the differing prevalence of pathology in these regions, particularly an increased presentation of parasitic infections and tuberculosis (Appendix 1). We will also exclude studies specifically conducted on patients with systemic conditions causing abdominal pain, such as porphyria, sickle cell anaemia and Familial Mediterranean Fever.

Index tests

The index tests will include all blood biomarkers used for the diagnosis of acute appendicitis. If combinations of biomarkers are reported in sufficient numbers, such as raised white cell count (WCC) and C-reactive protein (CRP); or raised WCC or CRP, then each defined combination will be considered separately as an index test. Similarly, if serial measurements of biomarkers are combined in a defined manner in sufficient numbers, each defined combination shall be considered as an index test. Where combinations or serial measurements of biomarkers are reported, we shall endeavour to extract individual data where available, for example on the performance of isolated index tests or admission tests.

Conventional blood biomarkers for the diagnosis of acute appendicitis include CRP, WCC and neutrophil count.

Experimental diagnostic biomarkers have also been reported in the literature and include IL-1 to IL-10 (Paajanen 2002), bilirubin (Burcharth 2013), procalcitonin (Yu 2013), calprotectin (Thuijls 2011), pancreatic stone protein (Tschuor 2012), D-lactate (Duzgun 2006), D-dimer (Kaya 2012), fibrinogen, (Kahramanca 2013), serum amyloid A (Schellekens 2013), mean platelet volume (Albayrak 2011), phospholipase A2 (Grönroos 1994), leucocyte elastase (Eriksson 1995), lactoferrin (Thuijls 2011), plasma total-oxidant capacity (Ozdogan 2006), adenosine deaminase (Öztürk 2008), lipopolysaccharide binding protein (Brănescu 2012), and nuclear factor-kappaB (Pennington 2000).

The conventional biomarkers have predefined positivity thresholds derived from population normal ranges. The positivity thresholds from emerging biomarkers, however, are likely to be poorly defined and may therefore contribute to heterogeneity.

CRP is an acute-phase protein with a large dynamic range in response to non-specific inflammation. Sensitive immunoassays for CRP have been routinely employed in clinical practice since the mid-1990s and are widely reported in acute appendicitis. In health, the median concentration is 0.8 mg/L, the 90^th centile is 3.0 mg/L and the 99^th centile is 10 mg/L (Pepys 2003).

Target conditions

The target condition is acute appendicitis. This may be categorised into simple or complicated (gangrenous or perforated) based on histological examination.

Reference standards

The following reference standards will be used:

1. Histological examination of the excised appendix or intra-operative assessment of a non-excised appendix in operatively managed patients.

2. Computerised tomography (CT) diagnosis of non-operatively managed patients or magnetic resonance imaging (MRI) diagnosis in non-operatively managed pregnant patients.

3. Clinical discharge diagnosis in those who did not receive an operation or imaging. This may include clinical follow-up where provided, such as readmissions, telephone or clinic follow-up.

Histological evidence of inflammation of the appendix represents the ‘gold standard’ diagnosis for acute appendicitis. However, not all patients presenting with clinically suspected appendicitis will undergo appendicectomy. In order to include non-operatively managed patients (true negative and false negative) in the analysis, clinical and CT reference standards will also be used.

The clinical reference standard is not well defined in the literature. It may include discharge diagnosis, re-admission to hospital or telephone or clinic follow-up.

Electronic searches

We will search The Cochrane Library (Appendix 2), MEDLINE (Ovid) (Appendix 3), EMBASE (Ovid) (Appendix 4) and the ISI Web of Knowledge (Appendix 5) using indexing terms and free text to capture the index tests and target disease as shown in the example outlined in Appendix 2. We will adapt the search terms to identify all papers included in existing systematic reviews of biomarkers in appendicitis. Relevant international surgical conference proceedings will be searched to identify abstracts. The search will be limited by date from 1990 to the current day in order to reduce heterogeneity in assays available for CRP. No language restrictions will be applied.

Searching other resources

The reference lists of retrieved articles will be searched manually for relevant studies.

Selection of studies

Two clinical review authors (MJM and NJS) will independently assess the eligibility of the titles and abstracts of all papers identified by the search strategy, using the predefined inclusion and exclusion criteria (see Criteria for considering studies for this review). Those papers considered eligible will then be read in full and assessed for inclusion. Any discrepancies will be resolved by consensus opinion on review of the full article, if necessary by a third clinical review author (IRD).

Data extraction and management

Two review authors (MJM and NJS) will independently extract data using a standardised Excel spreadsheet. Data extracted will include study design, participant characteristics, outcome measures and methodological quality. Specific data extracted will include patient presentation, selection, age, gender and pregnancy status; prevalence of appendicitis, perforation and negative appendicectomy; reference standard used; and biomarker diagnostic performance characteristics. We will seek outstanding data from the publishing authors.

Assessment of methodological quality

We will use the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) criteria to assess the quality of selected studies (Whiting 2011). Assessment will be carried out by two authors and disagreements will be resolved by consensus between senior authors. In Appendix 3 we provide the precise criteria we expect to use to make judgements about the risk of bias and problems with applicability.

Statistical analysis and data synthesis

The statistical analysis will be undertaken by MJM, advised and supported by CH and his colleagues in the Statistics Group of the University of Exeter Medical School. The general approach will be that suggested in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (Macaskill 2010).

For all included studies, we will enter the data in the two-by-two tables into Review Manager 5 software (Review Manager 5), which will allow the sensitivities, specificities and their 95% confidence intervals (CIs) to be presented in forest plots and receiver operating characteristic (ROC) space. These presentations will be used to explore the included study results, considering each biomarker individually, focusing on the test threshold in common use in clinical practice. The interaction between test accuracy results, study characteristics and methodological study quality will also be considered. In this, the first round of this systematic review, we will be focusing on clarifying the accuracy of individual tests and combinations thereof where available. Given the scale of this task, we will not extend the review to consider comparisons of tests. A further justification for this is that there is currently no consensus about the most clinically relevant alternative diagnostic strategies which should sensibly be compared. We believe an initial review focused on individual tests and combinations will inform such a debate, and that comparing alternative biomarkers is premature.

As well as accuracy, we will tabulate the number of uninterpretable results reported in each included study for each biomarker.

Where appropriate, particularly in terms of the number of included studies (minimum of four studies), we will attempt meta-analysis of each pair of sensitivity and specificity results from each included study for each biomarker. We will use the Hierarchical Summary ROC (HSROC) model implemented in SAS (SAS) or WinBUGs (WinBUGS), depending on software availability in the host institution at the time we conduct the analyses. Although the bivariate method may be appropriate for some biomarkers where the threshold is well defined, it will not be appropriate for the newer biomarkers. We thus prefer an approach which can be used consistently across all tests. The HSROC will also allow exploration of heterogeneity by incorporating co-variates into the model provided there is a sufficient number of included studies (see below).

We will interpret the results and prepare a summary of results table using Chapter 11 of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy as guidance (Bossuyt 2013).

Investigations of heterogeneity

The wider framework for investigation of heterogeneity will include aspects of the index test (specific manufacturer, version of test and threshold (if no established threshold in clinical practice)), target disorder (proportion of simple or complicated appendicitis), target population (age and prior symptoms/investigation) and study quality (nature of reference standard).

Heterogeneity will be explored in the first instance through visual examination of forest plots of sensitivities and specificities and through visual examination of the ROC plots. We will then include possible explanators of the heterogeneity as co-variates in the HSROC model. Of the potential sources of heterogeneity listed, we think that the precise nature of the biomarker and the nature of the reference standard will be, a priori, the most important influences on heterogeneity, and so we will preferentially examine the influence of these where the number of included studies limits our ability to fully explore potential sources of heterogeneity.

Sensitivity analyses

Where appropriate and if sufficient data are available, we will explore the sensitivity of any summary accuracy estimates to aspects of study quality using the number of domains where risk of bias was identified in the QUADAS-2 quality assessment. Provisionally, we will consider the effect of excluding studies where the reference standard domain was judged as having a high or unclear risk of bias, as this is considered the most relevant source of bias. We will then consider the effect of excluding studies where any two or more domains (patient selection, index test, reference standard, flow and timing) were judged as having a high or unclear risk of bias.

Assessment of reporting bias

Quantitative methods for exploring reporting bias are not well established for studies of diagnostic test accuracy. Specifically, we will not consider funnel plots of the diagnostic odds ratio versus the standard error of this estimate.

Domain 1: patient selection

Domain 2: index test

Domain 3: reference standard

Domain 4: flow and timing