Ventriculoperitoneal shunt insertion for hydrocephalus in human immunodeficiency virus-infected adults: a systematic review and meta-analysis protocol

Human immunodeficiency virus (HIV) is a lentivirus and is the causative organism for the acquired immunodeficiency syndrome (AIDS) [1]. Two major subtypes of HIV—HIV-1 and HIV-2—are known to exist [2]. Over 95% of patients infected with HIV will, without treatment, develop progressive immunodeficiency as indicated by falling CD4-positive T cell counts and/or development of AIDS defining illnesses within 10 years of seroconversion [3, 4]. In general, disease progression is slower in those infected with HIV-2 compared with the globally more common HIV-1 [1, 2].

HIV infection is associated with myriad central nervous system (CNS) sequelae. These include, but are not limited to, toxoplasmosis, primary CNS lymphoma (PCNSL), tuberculous meningitis (TBM), cryptococcal meningitis and immune reconstitution inflammatory syndrome (IRIS) [5‐8]. These may all disrupt cerebrospinal fluid (CSF) circulation causing an abnormal accumulation of CSF within the cranium, an associated rise in intracranial pressure and consequent cerebral injury. This condition is called hydrocephalus. In communicating hydrocephalus, ventricular CSF is continuous with CSF in the subarachnoid space. Communicating hydrocephalus develops because of excessive CSF production or reduced CSF resorption [9]. In non-communicating hydrocephalus, CSF passage from the ventricles to subarachnoid space is occluded. With time and successful treatment of the primary pathology, acute hydrocephalus might resolve, with normalisation of CSF pressure, or become associated with chronically raised CSF pressure [10]. Age-related cerebral tissue loss as well as HIV-associated neurocognitive disorder are associated with enlargement of CSF spaces due to generalised brain atrophy and may result in the radiological appearance of hydrocephalus but are associated with normal CSF pressure and flow, so-called hydrocephalus ex vacuo [9, 11]. This is not an indication for surgical treatment [9].

Meningeal infections including TBM, cryptococcal meningitis and bacterial meningitis can impair CSF resorption at arachnoid granulations, resulting in communicating hydrocephalus [10, 12]. TBM, which is associated with thick basal exudate, can also result in non-communicating hydrocephalus [13]. TBM in the context of HIV infection appears to be associated with substantially increased morbidity and mortality compared with HIV non-infected patients [14, 15]. This might arise because of synergistic effects of HIV and tuberculosis in the CNS causing depressed monocyte and microglial mediated immunity and dysregulated cytokine production [16‐21]. Hydrocephalus in TBM frequently persists following successful antimicrobial treatment [22]. Conversely, the majority of HIV-infected patients with raised CSF pressures secondary to cryptococcal meningitis (typically communicating hydrocephalus) experience normalising of CSF pressure and resolution of hydrocephalus following successful antimicrobial therapy [10]. A small proportion does require temporary CSF diversion for initial acute hydrocephalus [23]. Mass lesions such as toxoplasmosis, PCNSL, metastases from HIV-associated non-CNS malignancy and tuberculoma can trigger the development of non-communicating hydrocephalus by occlusion of ventricular outflow pathways [24, 25]. In these instances, hydrocephalus resolution post-treatment is dependent on restoration of normal ventricular outflow pathway anatomy. IRIS affecting the CNS is defined by a transient worsening or development of new inflammatory lesions during immune reconstitution around infective foci and might also present with hydrocephalus secondary to meningitis or mass lesions [26, 27].

As of 2015, it was estimated that 36.7 million people globally were living with HIV infection [28]. Of these, 19.0 million live in Eastern and Southern Africa [28]. HIV infection disproportionately affects certain vulnerable populations including intravenous substance misusers, sex workers and their clients, men who have sex with men and transgender people [28]. The risk of Mycobacterium tuberculosis disease is estimated to be 20 times greater in HIV-infected patients compared with their HIV non-infected counterparts, an effect which is most pronounced in low-income populations [29, 30]. As such, one cadaveric study conducted in the Ivory Coast in 1991 found that 32% of in-hospital deaths of HIV-infected patients were a result of tuberculosis with TBM affecting 20% of these [31]. The increasing availability of antiretroviral therapy (ART) has seen a decline in the incidence of new opportunistic intracranial infections, PCNSL and reduced AIDS-related deaths [5, 32, 33]. However, with improved survival and ongoing HIV transmission, the global prevalence of HIV and the associated burden of chronic disease continues to rise [28].

The incidence of hydrocephalus associated with HIV infection is poorly quantified at present. Small studies comparing clinical and radiological characteristics of HIV-associated and HIV non-associated patients with conditions associated with hydrocephalus provide some information: HIV-associated TBM appears to be associated with lower rates of non-communicating hydrocephalus than HIV non-associated TBM [34], and in some studies, over 50% of patients with cryptococcal meningitis, which occurs almost exclusively in HIV-infected patients, have exhibited raised CSF pressures [10]. However, non-communicating hydrocephalus is uncommon in cryptococcal meningitis [10, 35]. It is also important to note that any other cause of hydrocephalus may also be seen in the HIV-infected population; hence, adults might suffer the ongoing effects of infant-onset or congenital hydrocephalus. This is estimated to affect 1.1 in 1000 infants and 32 per 10,000 births and may be due to anatomical abnormalities such as neural tube defects and Chiari malformations or arise secondary to perinatal insults such as intraventricular haemorrhage and bacterial meningitis [36]. In HIV non-infected high-income populations, intracranial haemorrhage and malignancy account for 45 and 30% of new cases of adult onset hydrocephalus, respectively [37]. A lack of dedicated epidemiological research concerning HIV-associated hydrocephalus makes data from these small studies difficult to generalise, and little is known about the distribution of non-infective causes of hydrocephalus associated with HIV.

A ventriculoperitoneal shunt (VPS) is a surgically implanted device that allows flow of fluid from the cerebral ventricles into the peritoneal cavity. It consists of a silicone ventricular catheter that is inserted through a single burr hole, using either image guidance or anatomical landmarks to locate the ventricle. The catheter is sometimes rifampicin and clindamycin or silver impregnated [38]. It is connected to a valve that is usually placed in the subgaleal space. The valve can be pressure or flow regulated and may contain an antisiphon device [39]. This is connected to a length of distal tubing that is tunnelled subcutaneously from the valve site to the level of the abdomen where it is inserted through the abdominal wall into the peritoneal cavity where the distal catheter tip sits.

A functioning VPS provides a permanent pathway for ventricular CSF outflow and allows for normalisation of CSF pressure in hydrocephalus. Common complications following successful VPS include shunt infection, ventriculitis, proximal catheter occlusion or migration, shunt fracture, over drainage and distal catheter occlusion [40]. In one US cohort, 21% of adults who underwent VPS between 1990 and 2000 suffered a complication within the first year and 26% required replacement or removal of the shunt during the study period [41]. One cohort of mostly paediatric patients who underwent VPS between 2004 and 2007 in Kenya, a low-middle-income country with high HIV prevalence, there was a 2-year cumulative malfunction rate of 35% with 59.8% having “poor outcome” and 8.5% mortality [42‐44]. However, follow-up was reported as incomplete and this might not therefore been an underestimate of the complication rate in such a setting.

Alternatives to VPS include ventriculoatrial shunting where the distal catheter tip is sited in the right cardiac atrium and ventriculopleural shunting where it is sited in the pleural cavity. In some instances, an external ventricular drain (EVD) may be sited in preference to a shunt [39]. For this, a ventricular catheter is placed and the distal tip externalised through the scalp and connected without a valve to a drain mounted at a specific height above the tragus of the ear. This allows gravity to limit CSF flow via the catheter below a desired pressure. However, this is only a temporary means of CSF diversion. For some cases of communicating hydrocephalus, it may be sufficient to drain CSF through serial lumbar puncture or a lumbar drain [39]. Lumbar drains may be externalised using a similar system to an EVD or tunnelled and have the distal tubing placed in the peritoneal cavity—a lumboperitoneal shunt [39]. In cases of non-communicating hydrocephalus, it is sometimes possible to surgically divide the lamina terminalis or endoscopically conduct a third ventriculostomy (ETV) to allow for flow of ventricular CSF to the subarachnoid space [13]. However, this procedure is challenging—particularly in TBM [13]—and requires significant expertise and equipment that may not be available in emergency or resource-limited settings.

A proportion of patients infected with HIV develop hydrocephalus as a complication of CNS opportunistic infections. As described above, infectious aetiologies of hydrocephalus are over-represented in HIV-infected patients and HIV directly impacts on the pathological processes in intracranial infection. The literature from HIV non-infected patient populations therefore cannot be generalised to those with HIV infection.

VPS can be immediately life-saving in acute hydrocephalus and may be the sole treatment option for patients who are dependent on CSF diversion following emergency EVD insertion and are not candidates for ETV. However, two of the most cited studies comparing VPS in HIV-infected patients with TBM to HIV non-infected counterparts demonstrated mortality rates in excess of 60% in HIV-infected patients [45, 46]. Small numbers of HIV-infected patients in these studies achieved good outcome following VPS [45, 46]. Better outcomes have been reported following VPS insertion for other causes of HIV-associated hydrocephalus [47, 48]. To determine if the existing literature is sufficient to inform selection of HIV-infected patients for VPS in clinical practice, it is necessary for a systematic review to be conducted.

Our objectives were to compare post-VPS mortality, VPS infection, VPS malfunction and clinical outcome and rates of other complications at 1, 6 and 12 months between different aetiologies of HIV-associated hydrocephalus using quantitative meta-analysis and production of a narrative review. They would allow the identification of baseline patient characteristics which are predictive of good outcome following VPS in HIV infection.

Our study’s primary outcome will be 1-year post-VPS survival with comparison between all identified aetiologies of hydrocephalus in HIV-infected patients treated with VPS. We define separate aetiologies as being different sites of anatomical abnormality, with congenital and acquired causes considered separate: subarachnoid haemorrhage, intraventricular haemorrhage and different suspected or proven pathogens for infectious causes. Secondary outcomes will be survival at 1 and 6 months andAIDS-specific mortality, VPS failure, rates of perioperative complications and measures of outcome using validated outcome measures [49‐51] at 1 month, 6 months and 1 year post-VPS insertion. For all measures, comparison will be made between aetiologies of hydrocephalus.

Data will be extracted using standardised proformas (Appendix 2) and entered into review manager [52] by two authors (JL, NM).We will use the software package EndNote for reference management [53].

In the narrative review, we will summarise outcomes as reported by each included study separately. If data comparing outcome across different aetiologies of hydrocephalus exists, we will report it. If meta-analysis is performed, pooled odds ratios (OR) will be reported. The primary outcome of any RCTs comparing VPS with any other management strategy will be reported. If multiple trials of the same interventions report similar outcomes, meta-analysis will be conducted according to guidance by the Cochrane Handbook of Systematic Reviews of Interventions [54].

We will attempt to contact the authors of included studies electronically for any unreported details. If, following this, uncertainty remains, we will report available data and state missing data. Studies with missing mortality data will not be eligible for inclusion in meta-analysis.

On the basis of extracted data, studies will be compared for clinical and methodological heterogeneity. Where apparently homogenous studies report similar outcome measures at similar time points, the I ² (%) and T ² statistics will be calculated, with I ² greater than 50% suggesting substantial heterogeneity. For estimation of statistical heterogeneity, we will use odds of mortality between aetiologies of HIV-associated hydrocephalus as the measure of effect.

If substantial clinical, methodological or statistical heterogeneity is detected or if no comparative studies are included, we will not attempt meta-analysis and will present solely a narrative review. If homogenous comparative studies are identified, then meta-analysis of pooled mortality data will be conducted using the Peto method. OR at 12, 6 and 1 month post-VPS between pooled populations with different aetiologies of hydrocephalus will be compared. If greater than 10 homogenous studies are included for meta-analysis, meta-regression will be attempted. Covariables for this will include, where reported, age categories (16–24; 25–44; 45–59; ≥ 60), CD4⁺ cell count, time from presentation to VPS, initial Glasgow Coma Score [50], and shunt catheter type—either conventional, antibiotic or silver impregnated. Where not all variables are reported, regression will be conducted using the maximum covariables reported by all eligible studies. Where meta-analysis is performed, sensitivity analysis will be undertaken to determine the relative effects of inclusion of each individual study, as well as RCTs and non-randomised studies from different World Bank income categories: low-income, lower-middle-income, upper-middle-income and high-income economies. GRADEpro will be used to summarise strength of available data [57].

This is the original study protocol. Any subsequent protocol amendments will be published on the PROSPERO database of systematic reviews. Amendments to the protocol manuscript following peer review are documented (see Additional file 2).