We observed that six months after switching to a less neurotoxic regimen, participants in the intervention arm had higher CSF levels of neopterin and NFL, no changes in IMT and lipid profile, and a modest improvement in liver function tests and fibrosis scores (in comparison to participants remaining on previous regimen). Our hypothesis was that switching towards drugs that had a lower in vitro neurotoxicity would be associated with better cognitive function in patients with HAND as well as an improvement in biomarkers of neuronal damage. Yet robust evidence of an in vivo neuronal protection is lacking as well as the effect on lipid profile, cardiovascular risk and liver function tests given the use of boosted darunavir, maraviroc and emtricitabine (a triple combination seldom used in clinical practice as well as in medical literature). While the results of the primary objective (neurocognition) have been submitted for publication the other endpoints were mostly exploratory and the sample size was, unfortunately, very small.

HAND is a complex phenotype and the path towards effective treatments relies on the identification of pathogenetic mechanisms and reliable biomarkers. We included IMT, lipid, liver scores, and CSF markers in order to understand the potential underlying mechanisms (and response to treatment change) of vascular, hepatic and neurological factors in the neurocognitive impairment observed in PLWH (Msoka et al. 2019; Lemoine et al. 2019; Valcour et al. 2016; Ciccarelli et al. 2014; Han et al. 2013). The majority of the enrolled subjects were diagnosed with ANI, while only a minority showed MND and none presented HAD, similarly to what is currently globally observed (Haddow et al. 2018; Calcagno et al. 2018). Despite this, almost one fourth of the subjects did not complain about neuropsychological symptoms at the screening visit. This finding confirms previous studies, which observed a lack of disease awareness and insight resulting in a discordance between objective tests results and patients’ perceptions and the need of more reliable diagnostic tools (Cysique et al. 2021). In our population we did not observe significant virological changes over time.

CSF biomarkers that reflect processes within the brain constitute an important complementary approach to evaluate ongoing CNS disease. In our study, considering the similar cognitive profiles between HAND and neurodegenerative dementias, Tau, pTau, s100b, beta40, beta42, NFL and neopterin were evaluated, as well as CSF proteins, glucose and cells.

From the literature, we know that neopterin is an early marker of intrathecal immune response, whose CSF levels are elevated in most HIV subjects compared to uninfected individuals: recent studies showed an elevated neopterin levels in patients with HAND (Edén et al. 2016). In our study we observed a statistically significant increase in CSF neopterin in the intervention arm, possibly not confirming the observation that a less neurotoxic regimen may lead to less immune-activation and that other factors that we could not identify may be additionally involved, independently. Importantly, even if rising, neopterin stayed below the threshold value that is considered normal in both groups for the majority of patients despite a few outliers. Data from literature have proven the importance of a low-level immune-activation in driving alterations in functional brain activities, possibly expressed by the finding of a raised neopterin in the CSF, but no definitive data exist for beneficial lower levels of this protein in HAND patients (Edén et al. 2016). Other hypothesis to explain that finding may be ascribable to the chosen drug regimen: eliminating an NRTI from the HAART regimen may be deleterious in patients with HAND, since it is hypothesized that neurotoxicity associated with antiretroviral drugs could be class-specific and NRTIs have been shown to block inflammasomes and lower the likelihood of β-amyloid plaque deposition in the frontal cortex thus raising the possibility of a protective effect of NRTIs on the development of neurodegeneration (De Benedetto et al. 2020); still, as assessed in prospective studies, patients in dual therapy don’t seem to underperform in this specific group (Pérez-Valero et al. 2018; Mondi et al. 2015; Trunfio et al. 2020). An alternative hypothesis is that the intervention regimen has a less potent effects on controlling neuroinflammation. Ultimately, another compatible explanation lays on a possible lower adherence due to the increased complexity of the regimen; unfortunately, we could not register directly any of these variables for our patients (dedicated questionnaires), but indirect tools (viral load) to measure adherence confirmed an overall excellent compliance (all patients remained virologically suppressed). We also observed changes in CD4/CD8 ratio since this biomarkers has been previously associated with HIV-associated non infectious comorbidities and HAND (Vassallo et al. 2013, 2017): yet no significant changes over time was recorded.

NFL is a neuronal cytoplasmic protein highly expressed in large calibre myelinated axons. One of its main roles is to maintain axonal caliber and facilitate effective nerve conduction; its level increase in CSF and plasma proportionally to the degree of axonal damage in many neurological disorders, including inflammatory, neurodegenerative, traumatic and cerebrovascular (Mielke et al. 2019). Elevated NFL is one of the markers with a more consistent correlation with NCI in HIV (Anderson et al. 2018; Guha et al. 2019; Gisslén et al. 2015). Similarly to what we observed with neopterin, we noticed a slight elevation in both groups during the study period. In our case, that may partially be explained with the fact that patients were treated with old drug combinations, which would now be considered suboptimal, or because, as already discussed, omitting NRTIs in HAND patients or complicating a drug regimen with more pills may lead to worse results and lower adherence. The link between HAND and poor adherence has been recently confirmed by a systemic review and meta-analysis (Nweke et al. 2022). In addition, another possible explanation we could not assess pertains to the viral CNS reservoir of these patients that may drive such alterations independently of treatment change: indeed intensification studies failed in demonstrating a clear beneficial effect. (Sacktor et al. 2018; Yilmaz et al. 2010). No plausible explanations for this finding can be further speculated based upon our data. Still, in the absence of unequivocal data for the NFL trajectory in HAND, these two markers may rise independently from the means applied here in this randomized clinical trial and a single intervention may not be sufficient in patients that already suffer from a low-level ongoing neuronal damage, frequently started years prior to any active medical intervention. A potential bias for having an elevated NFL in the absence of actual HIV-replicative-dependent harm may be the LP itself: in macaque monkeys an increase the NFL value by 162% was registered after LPs; that subsequently returns to normal in 1 to 2 months from the spinal tap. These data are not confirmed up-to-date in humans nor in PLWH (Boehnke et al. 2020).

Tau protein is abundant in the central nervous system and involved in microtubule assembly and stabilization. It is predominantly associated with axonal microtubules and present at lower level in dendrites where it is engaged in signaling functions. Tau it is generally phosphorylated (ptau) during its physiological functioning. Generally speaking, increased phosphorylation results in decreased binding to microtubules which is important in tau-mediated neurodegeneration. Tau protein is also involved in blood–brain barrier dysfunction that was described in numerous neurological conditions and possibly play such role in infectious pathologies of the CNS (Mietelska-Porowska et al. 2014; Calcagno et al. 2016). The finding of this protein in the CSF it is ascribable to neuronal damage generally, and particularly in what happen in dementia-like CNS pathological entities. In the setting of HAND the link with AD (characterized by amyloid deposition and tau pathology) has been repeatedly suggested, still maintaining consistent differences (Trunfio et al. 2022). On the opposite from what we observed with neopterin and NFL, ptau decreased only in the intervention arm. In view of all that factors, the finding of a lower ptau in patients undergoing a less potentially neurotoxic regimen is of great relevance given the significantly higher risk of dementia observed in elderly PLWH (Bobrow et al. 2020; Lam et al. 2021; Yu et al. 2023). especially because baseline ptau resulted higher in this same patients’ group and this protein suffer less influence from macrophage activation. A reduced level during the switch regimen may be not sufficiently significant, as we did not observe the same improvement for what pertains to total tau and beta-amyloid.

Besides cellular damage and pathway alterations we explored vascular and liver involvement and changes with treatment modifications. When it comes to peripheral markers and non-invasive tool, we observed that IMT did not vary significantly over time. This may be of no surprise, as typically IMT variation may take more time and present alterations later than the 24 weeks study period herein considered. Yet the use of maraviroc was potentially beneficial. It had been associated with significant reduction in IMT, pulse wave velocity and triglycerides, and to have additional cardiovascular and neuroprotective characteristics that may have played a role in this study but were not readily measurable (Piconi et al. 2016; Martin-Blondel et al. 2016; Francisci et al. 2019).

Finally we explored whether changes in treatment regimen could be beneficial for liver involvement as measured by non-invasive biomarker of liver function. The hypothesis was a link between liver and brain function, involving the gut-brain axis and macrophage functions in systemic inflammation. (Thomas and Apovain 2017; Spencer et al. 2010). In fact, formerly maraviroc has been associated with benefit over fibrosis progression, specifically for prolonged treatment (Gonzales et al. 2014). As previously mentioned, additional HIV-related risk factors for HAND include immune-activation, dysregulation of the gut microbiome, and the potential toxicity of ART. On the other hand, non-alcoholic fatty liver disease (NAFLD) has become increasingly significant in PLWH. Thus, there is growing recognition of the need to address underlying lifestyle factors, such as weight reduction and optimal glycaemic control, along with pharmacological interventions. As part of the expanding interest in novel approaches, maraviroc has been proposed. Maraviroc works by inhibiting the binding of HIV-1 gp120 to the CCR5 coreceptor, thereby preventing the virus from entering cells. Its ability to antagonize CCL5-CCR5-mediated interactions has sparked interest in its potential anti-inflammatory benefits, in addition to its anti-HIV activity, particularly in the liver and brain (Martin-Blondel et al. 2016; Bradshaw et al. 2020; Gonzales et al. 2014).

Notably, the chemokine CCL5/RANTES, which acts as the ligand for CCR5, plays a crucial role in hepatic inflammation and fibrosis. CCR5 facilitates interactions among intrahepatic immune cells, promoting activation and migration of Kupffer cells and hepatic stellate cells, which, in turn, trigger inflammation and hepatic fibrosis. Consequently, inhibiting this pathway could potentially reduce the progression of fibrosis (Bradshaw et al. 2020). This is relevant in view of the fact that liver fibrosis is a process damaging lymphnodes, lymphatic function and immune-regulation, and there are similarities between the intestinal and the blood–brain barrier. Since the measurement of these markers is notably convenient and non-invasive, our observation of a reduction in liver stiffness becomes particularly significant within the context of the aforementioned hypothesis. It ultimately further supports the idea of a potential role for maraviroc therapy and demonstrates that readily accessible tools can be employed to comprehensively assess PLWH with HAND.

Putting these observations together, we should highlight the importance of following-up patients comprehensively with easy-to-use, informative and reproducible non-invasive tools that may provide important information on the patient’s status.

Several limitations should be recognized. It should be emphasised that the number of patients in both intervention and control arm is small, hence more data may be required from bigger study population observed over longer time to detect additional differences. Population was also heterogenous, as many of the enrolled subjects had comorbidities, in particular some suffered from depressive disorders (Hammond et al. 2016) and hypertension, or others being active smokers; that could partially act as confounding factors (specially on IMT), which has been already widely associated to neurocognitive decline (Calcagno et al. 2018). Moreover, a minority of patients enrolled were previously HCV-positive, which is a potential bias for liver and lipid scores calculation, a potential driver of immune-activation, and may impact on stiffness results; despite that, they all had been treated long before the enrollment and no one showed liver cirrhosis. It is still unclear if HCV eradication also eliminate the burden of disease on cognition, or if it has permanent consequences on CNS (Underwood et al. 2015). In addition, the lack of well-defined markers of immune activation and general inflammation is generally recognized. These processes can persist in patients on ART, and both have been implicated in neurocognitive impairment and HAND in general. Inclusion of more inflammatory markers such as IP-10, sCD163, CD38/DR on CD8 T cells would have aided in interpretation. Unfortunately, we could not assess these in our cohort.

However, this study also has strengths: our cohort had demographic characteristics similar to other modern study groups and our results could therefore be extended to minimally confounded PLWH diagnosed with HAND. It is the first real life project on HAND coupling a non-invasive and invasive tools for the evaluation of neurocognitive field, with other classical non-invasive tool such as IMT and Fibroscan that have been largely validated across cohorts (Msoka et al. 2019; Lemoine et al. 2019; Valcour et al. 2016; Ciccarelli et al. 2014; Han et al. 2013). More data from the principal analysis are possibly going to elucidate the relationship between these tools and neurocognitive performances.