Intracranial infectious aneurysm (IA) is a rare but potentially life threatening condition. It accounts for 0.7–6.5% of all intracranial aneurysms, according to different series.1 2 IA occurs because of destruction of the vessel wall secondary to infection of a segment of the artery. Spread of infection can be endovascular,3 as in infective endocarditis (IE), or via extravascular contiguous spread,3^–6 as in meningitis or cavernous sinus thrombophlebitis (CST). Rarely, it can occur from an unknown source (cryptogenic type). IA carries a high mortality.7 Hence it is very important to make an early diagnosis and initiate treatment on an emergent basis. Because IA is a rare disease, a high index of suspicion is essential to ensure its prompt diagnosis. In this context, it is essential to have a simple and reliable set of criteria for the diagnosis.

However, establishing the diagnosis of IA is often difficult. The gold standard currently involves demonstration of the presence of an infectious organism along with inflammatory mediated destruction of the vessel wall. This histopathological standard, although ideal, is fulfilled in <20% of the cases in the published series and is possible only when IA is excised during surgery or when examined during autopsy.8 Even then, fibrosis may be the only appreciable pathological manifestation, especially if the IA is resected after prolonged antimicrobial treatment. Histological confirmation is rather retrospective as the specimen becomes available only after surgical intervention or death of the patient. This standard is therefore difficult to apply in a strictly clinical context in which there is a need to guide management, but a tissue based diagnosis is not possible. There is a need for reliable and simple diagnostic criteria that can be applied early to select patients for different treatment protocols such as medical management, intervention procedures or surgery.

Unfortunately, at the present time, the clinical diagnosis of IA is incomplete; it is based on documentation of the aneurysm by angiography in the context of a predisposing infection such as IE, meningitis or CST (see fig 1). This standard is general and non-discriminatory, and may lead to either over diagnosis (as could occur when a pre-existing berry aneurysm is incidentally detected in the context of a relevant infectious disease) or under diagnosis (when the infection is not obvious or the aneurysm is atypical). Both of these issues could potentially be addressed by considering additional data in the diagnostic workup. After reviewing both our own clinical material as well as the published literature, we have developed and validated a new set of criteria for the clinical diagnosis of IA.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Cerebral angiogram showing two tandem intracranial infectious aneurysms (IA) (black arrows) on the posterior inferior cerebellar artery (A), one large IA on the distal segment of the posterior cerebral artery (B), one on the branches of the middle cerebral artery (C) and a giant lobulated IA of the cavernous portion of the internal carotid artery (D).

METHODS

This study was carried out at Sree Chitra Tirunal Institute for Medical Sciences and Technology (SCTIMST), in Trivandrum, South India, which is a tertiary care centre for neurological and cardiac disorders. The institution’s emphasis on cerebrovascular disease provided us with the opportunity to care for a large number of patients with IA over the past three decades.

Patients

Two cohorts of patients were identified by retrospective case review, and utilised to both develop and validate our proposed criteria. The IA cohort consisted of 25 patients with a confirmed diagnosis of IA treated in our hospital between 1976 and 2003. They all satisfied the following criteria: (1) aneurysm demonstrated by neuroimaging and/or autopsy; (2) presence of predisposing infection (IE, meningitis, CST or orbital cellulitis) and/or demonstration of pathogenic organism and documented inflammation and destruction of the vessel wall. The non-infectious aneurysm cohort consisted of all patients with intracranial aneurysms of non-infectious aetiology treated at SCTIMST from 1 January 2002 to 31 December 2002 (n = 111). They satisfied the following inclusion criteria: (1) aneurysm demonstrated by neuroimaging and (2) no evidence of predisposing infection. Relevant demographic data for both cohorts are presented in table 1. There were 1740 cases of intracranial aneurysms seen in our institute between 1976 and 2003, of which 1.44% were IAs. Clinical features and outcomes of patients with IA in this series have been published elsewhere.8

Identification of salient clinical data

We abstracted relevant clinical characteristics for each patient onto a standardised proforma. Data were transferred to a spreadsheet and appropriate descriptive statistics were calculated. The data were further analysed with SPSS V.14 (SPSS Inc, Chicago, IL, USA) to calculate sensitivity, specificity and generate receiver operator characteristic (ROC) curves. Positive and negative predictive values of the proposed criteria were calculated using the prevalence of 1.44% for IA calculated as described above.

Diagnostic criteria

We generated a set of discriminative clinical criteria by analysing permutations of relevant clinical data and comparing the resultant clinical predictive ability of each combination (see below). The proposed diagnostic criteria we developed include a mandatory criterion (demonstration of an intracranial aneurysm by neuroimaging) and 12 supportive criteria drawn from three domains (box 1). Each positive supportive criterion is given 1 point. The score under each domain (A_sum, B_sum and C_sum) as well as the total score were calculated for each patient. Category scores were calculated for each domain as 1, if the sum score for that domain was ⩾1.

RESULTS

The total score for the IA group ranged from 2 to 6 (mean 4.68 (1.03); median 5) while that for the non-IA group ranged from 0 to 2 (mean 0.73 (0.67); median 1). The mean A_sum for the IA group was 1.2 (5) and for non-IA group 0. The mean B_sum for the IA group was 1.16 (0.62) and for the non-IA group 0.27 (0.49). The mean C_sum for the IA group was 2.32 (0.75) and for non-IA group 0.46 (0.58). These differences were statistically significant (p<0.0001).

ROC curve analyses for various combinations of salient characteristics were generated. The area under the ROC curve, sensitivity and specificity for each domain and score are given in table 2. A score of 1 under domain A had high sensitivity and specificity but did not have the best area under the ROC curve and hence it was excluded from further analysis. The ROC curve area was highest (0.997) for the total score (fig 2) when compared with that for the sum for different domains (table 2). The category scores alone did not yield useful ROC, sensitivity or specificity scores (data not shown). A total score of 3 had maximum sensitivity (96%) and specificity (100%), as well as a positive predictive value of 100% and negative predictive value of 99.4%. A cutoff value of 2 for the total score had high sensitivity (100%) but its specificity was comparatively low (87.4%). For a total score of 1, the sensitivity and specificity for a diagnosis of IA were 100% and 39.6%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Receiver operator characteristic (ROC) curve for total score under the diagnostic criteria.

DISCUSSION

The diagnosis of IA requires a high index of suspicion. Currently, the diagnosis of IA is suggested in an appropriate clinical context but a definitive diagnosis is possible only when a pathological specimen is available or at autopsy. Previously, there have been no proposed or validated criteria to clinically diagnose IA. In this study, we have proposed a simple schema, consisting of one mandatory criterion with 12 supportive criteria under three major domains (predisposing infection, angiographic findings and other features contributing to diagnosis). Our results indicate that a total score of 3 or more, regardless of the domain, has high specificity, sensitivity (96%) and positive and negative predictive value for the diagnosis of IA. We therefore propose that a total score of 3 is confirmative of clinically definitive IA (see box 2). A total score of 2 would be consistent with clinically probable IA with a sensitivity and specificity of 100% and 87.4%, respectively. For a total score of 1, the sensitivity and specificity for the diagnosis of IA are 100% and 39.6%, respectively, consistent with possible IA. Although A_sum scores with a cutoff value of 1 also had high sensitivity (96%) and specificity (100%), we chose to rely on total scores for diagnosis as it had the highest area under the ROC curve. ROC curve characteristics have several advantages over sensitivity and specificity while evaluating the precision of a diagnostic test. ROC curve is more comprehensive and precise than sensitivity and specificity in assessing the usefulness of a new test. The main limitation of this study is that histological confirmation of IA was limited to 20% in the series that we used for this analysis. As mentioned earlier, histological proof is not possible in several instances. We would like to emphasise that these proposed criteria require further testing and validation on a different sample of data.

Previous reports on IA have included cases using solely clinical criteria, with a lower proportion of biopsy proven cases.1 3 4 9 10 We have extended this approach and selected several characteristics typical of IA drawn from clinical, imaging and other domains while preparing this proposed set of criteria. Our IA cohort consisted of 25 cases of IA of diverse clinical profiles. There were five cases of biopsy proven IA, one that showed a new aneurysm on follow-up angiogram, six cases that revealed a change in size of the aneurysm on follow-up angiogram and seven cases where the IA resolved with medical therapy. These features further strengthen the diagnosis of IA in the patients included in the IA cohort. In general, patients with IA were younger and exhibited disease in locations atypical for berry aneurysm (table 1).

Chapot and colleagues1 used the following criteria to include patients in their study of the endovascular treatment of IA. The presence of endocarditis and a positive blood culture were the core features for diagnosis. Each aneurysm had to involve segments 2, 3 or 4 of the middle cerebral artery or the posterior cerebral artery or involve segments M1 or P1 in association with at least two of the following criteria: (a) change in aneurysm size or morphology on consecutive angiograms, (b) presence of another intracranial or extracranial mycotic aneurysm, (c) rupture of the aneurysm, (d) arterial occlusion or stenosis adjacent to the aneurysm and (e) cerebral infarction caused by arterial occlusion at the level of the lesion.

Drawbacks of these criteria include the limitation of relevant infectious diseases to IE alone. In addition, these rules exclude all other predisposing factors, especially extravascular ones. Furthermore, these criteria insist on microbial culture positivity, a condition that restricts the applicability to less than half of IA in general. Organisms are isolated in only 30–47% of IA.9 10 Also excluded are aneurysms located in the anterior cerebral artery or vertebrobasilar circulation.

The cohort that we used in this study was large, contained a variety of aetiologies (bacterial, fungal, tubercular organisms) and predisposing conditions (infective endocarditis, meningitis and cavernous sinus thrombophlebitis). Hence the proposed criteria are applicable to IAs of both intravascular (ie, IE related) and extravascular (ie, meningitis, CTS, orbital cellulitis) origin. They include aneurysms in all locations within the CNS, and take into account 12 parameters to provide wider clinical generalisability. One possible advantage of these proposed criteria is that they can be applied even in the absence of microbial isolation. The diagnosis of cryptogenic IA is more challenging, as one important criterion (presence or history of predisposing infection) is lacking in such instances. However, with these criteria, we diagnosed clinically definite IA in the only case of cryptogenic IA in our series. The strength of these criteria to diagnose cryptogenic IA needs confirmation in another set of data from a different background. We included fever or history of recent fever of 1 or more weeks’ duration in domain three. This strategy enabled us to include other infections, pneumonia or other sepsis that may possibly contribute to the pathogenesis of IA.

CONCLUSION

The proposed criteria appear to be reliable and easily applicable to a wide range of IA patients. They are likely to have wider generalisability as they are based on analysis of all subtypes of IA, including cryptogenic variant. This set of diagnostic criteria would probably be very useful in clinical research and practice. Its clinical reliability needs to be confirmed further by validation on another set of IA under different conditions.