Role of systemic inflammation scores for prediction of clinical outcomes in patients treated with atazanavir not boosted by ritonavir in the Italian MASTER cohort

Background Atazanavir (ATV) not boosted by ritonavir (uATV) has been frequently used in the past for switching combination antiretroviral therapy (cART). However, the clinical outcomes and predictors of such strategy are unknown. Methods An observational study was carried out on the Italian MASTER, selecting HIV infected patients on cART switching to an uATV-containing regimen. Baseline was set as the last visit before uATV initiation. In the primary analysis, a composite clinical end-point was defined as the first occurring of any condition among: liver, cardiovascular, kidney, diabetes, non AIDS related cancer or death events. Incidence of AIDS events and incidence of composite clinical end-point were estimated. Kaplan-Meier and multivariable Cox regression analysis were used to assess predictors of the composite clinical end-point. Results 436 patients were observed. The majority of patients were males (61.5%) and Italians (85.3%), mean age was 42.7 years (IQR: 37.7–42), the most frequent route of transmission was heterosexual intercourse (47%), followed by injection drug use (25%) and homosexual contact (24%); the rate of HCV-Ab positivity was 16.3%. Patients were observed for a median time of 882 days (IQR: 252-1,769) under uATV. We recorded 93 clinical events (3 cardiovascular events, 20 kidney diseases, 33 liver diseases, 9 non AIDS related cancers, 21 diabetes, 7 AIDS events), and 19 deaths, accounting for an incidence of 3.7 (composite) events per 100 PYFU. At multivariable analysis, factors associated with the composite clinical end-point were intravenous drug use as risk factor for HIV acquisition vs. heterosexual intercourses [HR: 2.608, 95% CI 1.31–5.19, p = 0.0063], HIV RNA per Log10 copies/ml higher [HR: 1.612, 95% CI 1.278–2.034, p < 0.0001], number of switches in the nucleoside/nucleotide (NRTI) backbone of cART (performed to compose the uATV regimen under study or occurred in the past) per each more [HR: 1.085, 95% CI 1.025–1.15, p = 0.0051], Fib-4 score per unit higher [HR: 1.03, 95% CI 1.018–1.043, p < 0.0001] and Neutrophil/lymphocytes ratio (NLR inflammation score) per Log10 higher [HR: 1.319, 95% CI 1.047–1.662, p = 0.0188]. Conclusions Intravenous drug users with high HIV RNA, high Fib-4 levels and more heavily exposed to antiretroviral drugs appeared to be more at risk of clinical events. Interestingly, high levels of inflammation measured through NLR, were also associated with clinical events. So, these patients should be monitored more strictly.


Background
Atazanavir (ATV) is a protease inhibitor (PI) for the treatment of HIV infection, prescribed with ritonavir (atazanavir/ritonavir, ATV/r) in combination antiretroviral therapy (cART) [1][2][3][4]. Differently from other PIs, ATV not boosted by ritonavir (unboosted ATV, uATV) has been rather frequently used in the past as an off-label switching option [5,6], for toxicity reasons [7,8]. Previous studies indicated safety and effectiveness of uATV prescription in selected cART experienced patients, with undetectable HIV RNA and without drug resistances [9]. This is why regimens including uATV have been experimented with and were used in clinical practice [6], and subsequently uATV prescription has been formally approved in the meantime by European Medicine Agency (EMA) and by Italian National Health authorities in 2014. At the same time, national guidelines recommended caution because of the lower genetic barrier of regimens including uATV [10].
Clinical outcome of patients receiving ATV (and particularly of those received uATV) is still not completely clear. In a previous study of the Italian MASTER cohort [11], patients treated with ATV/r more frequently achieved a composite outcome of success (HIV RNA <500 copies/mL, no AIDS events, CD4 + T cell count >500 cell/ mm 3 ) than patients prescribed other PIs [12] and reported a lower probability of switching the regimen than patients treated with other PIs [12].
However, there are some unknowns about cardiovascular risk and renal and liver profile of patients treated with uATV. Firstly, a moderate increase of lipid values (total cholesterol, HDL and LDL fractions, triglycerides) has been reported in patients prescribed ATV/r [13,14]. Similarly, the expected increase in cardiovascular risk [15] was not confirmed by previous studies [16][17][18]. A recent study of the Italian MASTER cohort reported that the increased total cholesterol was balanced by the increased HDL fraction in patients prescribed ATV/r or uATV [19] and other studies observed a less impact on lipid profile in patients prescribed uATV [20].
Some degrees of acute and chronic interstitial nephritis, crystal deposition and nephrolithiasis were described when ATV/r was co-administered with other potentially nephrotoxic drugs, as tenofovir (TDF) [21][22][23]. However, an improvement of filtration rate in patients prescribed ATV/r or uATV was reported, either after switching to a regimen without TDF [19] or in patients prescribed ATV/r as a part of simplification regimens [13,14]. Therefore, other factors than use of ATV may be involved in renal failure development.
Liver toxicity is a well-known side effect of ATV/r, so prescription of uATV is often adopted in case of severe hyperbilirubinaemia and cholelitiasis [7]. However, effects of uATV prescription on liver fibrosis progression are not yet known.
Further, an increased risk of clinical events was associated with systemic inflammation levels in HIV infected patients. As recently described, neutrophil/lymphocyte ratio (NLR) measured at baseline and during follow up was independently associated with incidence of cardiovascular events [24], and both NLR and platelet/lymphocyte ratio (PLR) have been found to be associated with risk of death in patients with solid cancer, in general population and HIV positive subjects [25].
Lastly, among proposed serum non-invasive biomarkers of liver fibrosis, the Fib-4 score has been validated both in HIV/HCV co-infected and in HIV monoinfected patients and it was a reliable driver of both liver fibrosis progression and clinical events [26,27].
The goals of present study were: (i) to estimate the risk of clinical progression (including both AIDS and non AIDS related morbidities, and death) in patients prescribed uATV; (ii) to explore possible predictors of the aforementioned outcomes, including systemic inflammation scores.

Characteristics of the Italian MASTER cohort
The present study was conducted including patients enrolled in the Italian MASTER cohort (MAnagement Standardizzato di TErapia antiRetrovirale, Standardized Management of Antiretroviral Therapy), a longitudinal multicenter cohort including patients in nine referral centers throughout Italy (http://www.mastercohort.it) [11]. Patients' data are recorded on a common electronic clinical chart software (Health & Notes 3.5W, Healthware S.p.A., Naples, Italy) employed by all participating Centers. The electronic clinical chart is used to record data of patients at each visit and periodically revised and updated. All participant subjects signed written informed consent and each center obtained approval by its Ethics Committee.

Study design
An observational study including HIV infected patients switching to uATV was performed. All patients were 18 years old or more. Patients were included in the study if received uATV from 2000 to 2015. Subjects included in the study were all experienced to antiretroviral therapy. Baseline was defined as the last evaluation before uATV prescription. Patients were followed from baseline (date of the last visit or exam evaluation before uATV prescription) to the time of death, switching from uATV to another regimen, or to the last available visit. The observation was interrupted also when patients were lost to follow up. Patients with any renal, liver, cardiovascular events or non AIDS related cancers before or at switch to uATV were excluded from the present study. A composite end-point (the first occurring of these conditions: liver, cardiovascular, kidney, diabetes, non AIDS related cancer or death events). Renal disease was defined as estimated glomerular filtration rate (eGFR) ≤60 ml/min calculated using Chronic Kidney Disease Epidemiology (CKD-EPI) formula [28], an additional measurement of eGFR within 90 days was acquired to confirm renal impairment. Clinical events of patients were recorded on an electronic clinical chart and reported following a standard classification [11]. Cardiovascular events included diagnosis of acute myocardial infarction, stroke, transient ischemic attack, angina pectoris, coronary bypass, angioplasty, chronic occlusive arterial disease, hypertension. Liver diseases included: hyperbilirubinemia (>2.5 mg/dl), liver fibrosis stage ≥ F3, diagnosis of liver cancer. Diabetes was defined as fasting glucose ≥126 mg/dl or oral anti diabetes therapy or insulin prescription. Discontinuation of uATV was defined as uATV interruption or ATV/r prescription. Causes of uATV discontinuation were categorized as: virological failure, toxicity, simplification, other. A sensitivity analysis limited to patients who reported <4 changes of therapy in the nucleoside/nucleotide (NRTI) backbone (at baseline or in the past) was conducted, in order to exclude potential confounders due to multiple cART switches and multiple virological failures.

Data collection
Baseline data collected included age, gender, nationality, risk factors for HIV acquisition, HBV and/or HCV coinfections, number of AIDS events, HIV RNA, CD4+ T cell count, ATV/r exposure, cART switches occurred for the backbone. The following parameters were also collected during the follow-up if available: CD4+ T-cell count, HIV RNA, bilirubin levels, γ-glutamyl transferase (γGT) levels, alanine transaminase (ALT) and aspartate transaminase (AST) levels, total and fractioned cholesterol, triglycerides, serum glucose, serum creatinine, platelet count, albumin, blood count, C reactive protein and coagulation markers.
Liver fibrosis estimated by Fib-4 formula [29], eGFR, body mass index (BMI) and systemic inflammation scores (GPS-Glasgow Prognostic Score; mGPS-modified Glasgow Prognostic Score; NLR and PLR) were calculated at baseline and during the follow up using data collected at each time point. Clinical events (AIDS, diabetes, non AIDS related cancers, cardiovascular events, renal and liver diseases), discontinuations of uATV and deaths were recorded at baseline and at each time-point.

Statistical analysis
Incidence of AIDS and non-AIDS clinical events and of the composite end-point were calculated. Kaplan-Meier analysis was performed to estimate probability of being event-free during follow up. Cox regression analysis, univariate and multivariable analyses (with bi-directional stepwise selection driven by Akaike Information Criterion, AIC), were used to assess predictors of composite clinical end-point. Data collected at baseline and during the follow-up, were input in the models as time updated variables. All analyses were performed using R, the language for statistical computing (https://www.r-project.org/).

Baseline characteristics of patients
Four hundred and thirty-six patients were selected (see Table 1). The majority of patients were males (61.5%) and Italians (85.3%). Mean age was 42.7 years (IQR: 37.7-42). The most frequent risk factor for HIV transmission was heterosexual intercourse (47%), followed by injection drug use (25%) and homosexual contact (24%). The rate of positive HCV-Ab was 16.3%. All patients were previously exposed to antiretroviral drugs, including protease inhibitors, before switching to uATV.

Clinical outcomes
Patients were observed for a median time of 882 days (IQR: 252-1,769) under uATV. For the composite clinical end-point, 40 and 30 patients contributed after day 200 and 300, respectively. Ninety three clinical events (3 cardiovascular events, 20 kidney diseases, 33 liver diseases, 9 cancers, 21 diabetes, 7 AIDS events), and 19 deaths occurred during the follow up. Eighty nine composite clinical events were observed (Fig. 1). Causes of death were: AIDS for 6/19 patients, cardiovascular event 1/19, non AIDS related cancer 1/19, other reasons 11/19 patients. An incidence of 3.7 events per 100 person-year of followup (PYFU) was estimated. Three hundred eighty-six patients (88.5% of total subject enrolled) discontinued uATV during the follow up. For uATV discontinuations, 48 and 36 patients contributed with data after 2000 and 3000 days, respectively. The cumulative incidence of discontinuations of uATV was 16.6 per 100 PYFU. Forty-seven per cent patients switched to alternative drugs and 53% patients were prescribed ATV/r. Fourteen percent of the patients stopped uATV for toxicity, 19% for simplification, 15% for virological failure, 5% for patient's choice, and 47% for other/unknown reasons including avoidance of off label prescriptions in patients who did not have any other reasons for stopping uATV.

Predictors of clinical outcomes
Incidence of the composite end point according to risk factors for HIV acquisition, HIV RNA, Fib-4 score, and number of NRTI switches in the cART backbone are depicted in Fig. 2. At multivariable analysis (

Discussion
This study evaluated long-term clinical outcomes of HIV infected patients included in the Italian MASTER cohort after switching to a simplification regimen which included uATV. Findings of the present study are original because, for the first time, we explored the possible association of both inflammation and liver fibrosis scores with clinical events. Moreover, this is the first study which evaluated long-term clinical outcomes in patients treated with uATV. High levels of inflammation are predictors of clinical events and mortality in HIV infected patients [30,31]. In particular, studies from our cohort demonstrated that NLR score is an independent predictor of mortality for cancer [25,32,33], and cardiovascular events [24]. Moreover, it was demonstrated that may predict death in patients with liver cirrhosis [34]. Herein we confirm that a high NLR is a predictor of clinical events when a composite clinical outcome is used in patients prescribed uATV. Conversely, we did not observe any significant associations between clinical outcome and GPS or PLR. So, probably, these scores (GPS and PLR) are predictors of progression of disease and mortality but only in specific subgroups of HIV infected patients, such those with cancer [25,32].
The fact that, in a subset of patients with experience of <4 changes in the NRTI backbone, NLR was not significantly associated with the composite clinical outcome may indicate that this score is more reliable in those who were more heavily exposed to cART, with a longer disease history, and with a greater risk of complications. Therefore, our results suggest that NLR warrants more attention and should be measured in HIV infected patients, especially in the most fragile, even when they are treated with "metabolic friendly" regimens (such as cART including uATV).
In the present study, we found that intravenous drug use and higher γGT levels (that may be considered as a proxy for alcohol abuse [35]) are associated with the composite clinical outcome. Also, higher levels of serum glucose were associated with clinical events. These findings are not unexpected because it is well known that life-style factors, such as nutritional habits, alcohol abuse, intravenous drug use or smoking exert an important role for the risk of clinical events even in HIV infected patients [27,[36][37][38][39]. So, we confirm that targeted interventions to correct these modifiable risk factors should be prioritized in HIV infected patients to reduce risk of clinical events even when more "metabolic friendly" regimens are prescribed.
The Fib-4 score is a reliable marker of liver fibrosis progression and clinical events in HIV infected patients [26,27]. Indeed, we also found a statistically significant association of higher Fib-4 scores with the composite clinical outcome. Hence, our results suggest that patients treated with uATV with higher estimated liver fibrosis stage should be considered more at risk for clinical events. This is particularly applied to patients co-infected with HCV, for whom uATV was one of the preferred regimens [7,8], and to those with higher NLR. Overall, these results reinforce the importance of eradicating HCV as it is the main driver of liver fibrosis progression, to reduce both liver complications and non-AIDS related co-morbidities, especially if Fib-4 and NLR remain elevated.
Results of our study are affected by several limitations. First, high incidence rates of uATV discontinuations (in most cases for "administrative" reasons to reduce off-label prescriptions) may have reduced the length of observation and, therefore, the statistical power of the study may have been limited further. However, the impact of NLR was demonstrated on the risk of both clinical complications and discontinuations of uATV.
Second, our study did not include a control group. Clearly, a control group would be of value, however, we believe that comparison with other cART regimens is not recommended in this study. Indeed, the main inclusion Third, albeit alternative regimens are preferred over uATV when metabolic or liver complications are of concern, our results (especially the potential impact of persisting inflammatory state estimated through NLR) may be of general interest, provoking further investigations. Since uATV may have a smaller effect on residual HIV RNA which may increase, in its turn, levels of inflammation, it is important conduct further studies in patients receiving different cART regimens.

Conclusions
This work showed that NLR score is associated with clinical events in patients prescribed uATV, especially if they were intravenous drug users with uncontrolled HIV RNA, higher liver fibrosis stages, high γGT and serum glucose levels, and more heavily pre-treated with cART. In conclusion, the role of NLR in predicting the risk of non AIDS related complications and mortality should be investigated in further studies, particularly in most fragile patients (either for HCV co-infection or metabolic complications) such as those who were usually treated with uATV in clinical practice.

Availability of data and materials
For ethical and legal restriction we cannot upload a minimal data set. Data are available upon request, the interested researchers could contact directly scientific secretariat of the Italian MASTER Cohort (http:// www.mastercohort.it).
Authors' contributions MCP, CT, and MP conceived the study and participated in its design and coordination; MP performed the statistical analysis; MCP, MP and CT interpreted the data and participated in drafting the manuscript; NM participated to data management and extraction from database; EF, EQR, EDF, FR, AB, NL, MDP, AG, LS and AP participated in revision of the manuscript and contributed to patients' enrollment in the study and follow up; EF contributed to manuscript for intellectual content. All authors read and approved the final version of the manuscript.

Competing interests
Prof. Carlo Torti is a member of the editorial board of BMC Infectious Diseases journal. Authors declare that they have no competing interests to declare that may bias results of this work.

Consent for publication
Not applicable.

Ethics approval and consent to participate
The study was conducted in accordance with the guidelines of the Declaration of Helsinki and the principles of Good Clinical Practice. All patients provided written informed consent to include their clinical and biological data in the MASTER database for scientific purposes. Data submitted by the participating clinics to the data center were anonymized.