Introduction

Darunavir (DRV; TMC114; Prezista®) is a second-generation non-peptidomimetic protease inhibitor (PI) that was approved in 2007 in Italy for use in combination with ritonavir booster (DRV/r). DRV is used in combination with other antiretroviral (ARV) drugs for the treatment of human immunodeficiency virus (HIV) type 1 infection at two dosage regimens [800 mg once daily (OD) and 600 mg twice daily (both co-administered with ritonavir)]1,2. These regimens allow treatment of the entire setting of HIV-positive patients, from treatment-naive to highly experienced subjects and even those harboring HIV resistance mutations3.

The efficacy and tolerability of DRV/r have been evaluated in registrative randomized controlled clinical trials (RCT) in treatment-naïve4,5 and treatment-experienced6,7,8,9 patients with HIV-1 infection, with documented long-term efficacy and tolerability7,10,11,12. These results have been confirmed by real world evidence from observational studies13.

A once-daily co-formulation of DRV 800 mg plus a new booster, cobicistat 150 mg (Rezolsta®), is currently available. This fixed-dose combination (FDC) allows replacement of ritonavir as a booster for the treatment of both naïve and treatment-experienced adults14. The safety and efficacy of a single tablet regimen (STR) of darunavir/cobicistat/tenofovir alafenamide/emtricitabine (D/C/F/TAF) is being evaluated in two large phase III trials in treatment-naive and virologically suppressed patients (NCT02431247 and NCT02269917, respectively). The results of studies using cobicistat as a booster for darunavir showed no difference in efficacy from the use of ritonavir as a booster; therefore, the results of the present meta-analysis can be considered of interest even in this changing environment.

Current Italian15 (with some restrictions), European16, British17 and DHHS18 HIV/AIDS guidelines recommend the use of darunavir boosted with ritonavir or cobicistat as the only boosted protease inhibitor (bPI) (alongside other options, including integrase inhibitors and rilpivirine) as one preferred third agent in addition to a nucleoside reverse transcriptase inhibitor backbone, including tenofovir fumarate or tenofovir alafenamide and emtricitabine18.

Hence, the primary purpose of the present meta-analysis was to evaluate the efficacy, safety and tolerability of DRV/r-based regimens for treatment-naive HIV-1-infected patients or ART-experienced patients using reported RCTs.

Results

A search of electronic medical databases retrieved a total of 1055 articles. After title and abstract screening, we excluded 891 articles mainly because the authors did not report original data (i.e., narrative reviews, editorials, guidelines, or case reports) or the studies were designed as pharmaco-economic evaluations. After removal of duplicates using the Endnote X7 software, 134 articles on DRV were considered in-depth, and all full texts were downloaded and screened for final inclusion. After cross-checking for additional potentially missed references, 46 original articles with data on efficacy and safety were included in the present meta-analysis (Fig. 1). We considered three groups of studies based on the features of the enrolled patients: ART-naïve, ART-experienced failing and ART-experienced virologically controlled subjects. From a statistical perspective, we considered only studies with 48 and 96 weeks of follow-up (FU) to obtain sufficient subjects to conduct a meta-analysis. The main characteristics of the design and the baseline characteristics of the enrolled patients in the studies included in this analysis are summarized in Table 1 (ART-naïve adult patients) and Table 2 (ART-experienced adult patients). The results of the individual study quality assessments are reported and summarized in Supplementary Table 1. The study protocols were obtained where available to assess selective outcomes reports. The included studies achieved adequate sequence generation, but allocation concealment was not reported in all studies. All studies reported statistical analyses of the outcomes and addressed any incomplete data, such as loss to follow-up. All RCTs included were open-label; therefore, the two domains of performance bias and attrition bias were deemed to have a high risk of bias (Supplementary Table 1).

Figure 1
figure 1

Flow-chart describing the literature search and study selection processes.

Full size image
Table 1 Main characteristics of trials considering ART-naïve adult patients.
Full size table
Table 2 Main characteristics of trials considering ART treatment-experienced adult patients.
Full size table

Efficacy

Efficacy was defined as the virological response rate (viral load < 50 copies/ml) at 48 and 96 weeks for the ART-naïve adult patients and at 48 weeks for the ART-experienced patients.

For the ART-naïve patients, we included eight studies in this meta-analysis covering a total of 4430 adult patients evaluated (four with 48 weeks of FU and four with 96 weeks of FU). In the intention-to-treat (ITT) analysis, the virological response rate with DRV/r was not significantly different from the comparator at weeks 48 and 96, with risk ratio (RR) values equal to 1.04 (95% confidence interval (CI): 0.92–1.18) and 0.99 (95% CI: 0.90–1.08), respectively. A high degree of heterogeneity emerged between the RR estimates at week 48 (heterogeneity test I2 = 75%, p = 0.007) and week 96 (I2 = 81%, p = 0.001) (Fig. 2).

Figure 2
figure 2

Meta-analysis of viral suppression for ART-naïve adult subjects at 48 (Panel a) and 96 (Panel b) weeks of follow-up.

Full size image

For the ART-experienced patients, data were available from three studies for failing subjects (a total of 1440 adult patients evaluated) and from 11 studies for virologically controlled subjects (a total of 1553 adult patients evaluated). At week 48, the ITT analysis of the treatment-experienced failing subjects showed that the virological response rate was significantly higher for DRV/r than for the comparator group (RR 1.45, 95% CI: 1.01–2.08), but the heterogeneity test showed high variability among the studies (p < 0.0001). Conversely, for the treatment-experienced virologically controlled DRV/r group, no significant difference was found between the DRV/r and comparator groups (RR 1.03, 95% CI: 0.98–1.08), and the variability of the study estimate was low (I2 = 32%, p = 0.14) (Fig. 3).

Figure 3
figure 3

Meta-analysis of viral suppression for ART-experienced adult subjects at 48 weeks of follow-up.

Full size image

In the sensitivity analyses conducted in naïve subjects at week 48, we calculated pooled RRs after excluding the studies one by one. No study had a notable influence on the overall estimate, because the pooled RRs varied between 1.01 (when excluding the IMEA19 study) and 1.09 (when excluding the FLAMINGO20 study). The same result was obtained for the treatment-experienced virologically controlled subjects. No evidence of publication bias was detected.

Safety

We evaluated the discontinuation rate due to adverse events (AEs) related to DRV/r for 13 studies and pooled the results for weeks 48 and 96. The DRV/r safety profile was not significantly different from that of the comparator (RR 0.84, 95% CI: 0.59–1.19); this result was supported by the low variability between studies (I2 = 34%, p = 0.11), as shown in Fig. 4.

Figure 4
figure 4

Meta-analysis of studies reporting data on treatment discontinuation due to adverse events and any serious adverse event related to the administered treatment.

Full size image

Regarding the discontinuation rate due to serious adverse events (SAEs) related to DRV/r, we evaluated 10 studies and pooled the results for weeks 48 and 96. In this analysis, the difference between the DRV/r and the comparator was also not significant (RR 0.78, 95% CI: 0.57–1.05), and low-to-moderate variability was found between the study RRs (I2 = 41%, p = 0.08) (Fig. 5).

Figure 5
figure 5

Meta-analysis of studies reporting data on any serious adverse events (SAEs).

Full size image

Notably, cardiovascular (CV) events were analyzed for all of the studies included in this meta-analysis. In the 19 trials (including a total of 4992 subjects), seven non-specified CV events were reported in the MONOI21 trial, one stroke was reported in the DRV/r arm and one myocardial infarction (MI) in the lopinavir/ritonavir (LPV/r) arm in the ARTEMIS10 trial, one MI was reported in the DRV/r arm and one cardiomyopathy in the dolutegravir (DTG) arm in the FLAMINGO20 trial, and one case of pericarditis was reported in the atazanavir (ATV) arm in the IMEA19 trial. When publications were available, CV events were also evaluated at the longest follow-up time point (Table 3). The proportion of CV events in the DRV/r-treated patients was 0.18% (9/4992). For DRV/r, the incidence rate (IR) was 1.44 per 1000 person-years.

Table 3 Cardiovascular events reported in clinical trials containing darunavir.
Full size table

Mono vs triple therapy

To evaluate the sole impact of DRV/r on safety, we compared the results of monotherapy with those of triple therapy in the studies reporting on DRV/r in treatment-experienced, virologically controlled subjects. The monotherapy arm of the trials was taken as a comparator. We considered four studies reporting endpoints of viral suppression at week 48. DRV/r was significantly better in triple therapy than in monotherapy (RR 0.94, 95% CI: 0.90–0.99). No heterogeneity was found between the estimates (I2 = 0%, p = 0.63) (Fig. 6).

Figure 6
figure 6

Meta-analysis of viral suppression for ART-experienced subjects at 48 weeks of follow-up considering monotherapy vs triple therapy.

Full size image

We evaluated three studies to assess discontinuation due to AEs at week 48. This variable did not significantly differ between DRV/r in monotherapy and DRV/r in triple therapy (RR 1.70, 95% CI: 0.80–3.62) in the absence of heterogeneity between RRs (I2 = 0%, p = 0.37) (Fig. 7).

Figure 7
figure 7

Meta-analysis of studies reporting data on any serious adverse events (AEs) considering monotherapy vs triple therapy* - *Favors triple indicates that a higher number of AEs was reported in the triple therapy arms.

Full size image

Discussion

Nineteen RCTs were included in this meta-analysis. The first RCT was published in 2007 and described treatment-experienced subjects, and the most recent trials were published in 2016 and involved naive subjects.

In the ITT analysis of the ART-naïve subjects, the virological response rate did not differ between the DRV/r and the comparator arms, at both 48 and 96 weeks, despite the wide variability of the studies. Heterogeneity can be explained by the baseline characteristics of the subjects included in studies, such as ARTEMIS10 and IMEA19 compared to FLAMINGO20. The subjects were more advanced in ARTEMIS10 and IMEA19 than in FLAMINGO20, with higher viral loads and lower CD4+ cell counts.

In the ITT analysis at week 48 of the ART-experienced failing subjects, the virological response rate was significantly higher for DRV/r than for the comparator drugs, regardless of the previous clinical and treatment history and despite the wide heterogeneity of the studies. To date, DRV/r is the only antiretroviral drug which have been studied in highly pretreated subjects, and this population has not been enrolled in any subsequent study. In a pooled analysis of POWER studies22, DRV/r provided a sustained virological response in patients with reverse transcriptase and protease resistance-associated mutations at baseline22. This finding shows the high potency and high genetic barrier of DRV/r23 and its efficacy against resistant viruses. These results are in line with the well-known genetic barrier of DRV/r and its proven efficacy against resistant viruses. Furthermore, the DRV genetic barrier is still unequalled with respect to both other PIs and to inhibitors of strand transfer (INSTIs).

In the ITT analysis at week 48 of the ART-experienced virologically controlled subjects, the virological response rate was comparable to that of the comparator group (I2 = 34.6%, p = 0.122). In four of these studies, DRV/r was used as a monotherapy, and its potency in reaching viral undetectability was confirmed, as was its good penetration in HIV reservoirs21,23,24,25. These results were achieved in clinical practice in both naïve and highly experienced patients, the latter of whom had approximately seven years of FU, as reported in an Italian observational cohort (the TMC114HIV4042 study13, registered in ClinicalTrials.gov under the identifier NCT01375881).

The safety profile of DRV/r was similar to that of the comparator irrespective of the dosage and the comparator used. In this analysis, pooling the naïve and experienced subjects could have introduced bias, because the naïve subjects had never taken DRV/r. Notably, in the FLAMINGO trial, significantly more SAEs occurred in the DTG arm than in the DRV/r arm (RR 0.58; 95% CI: 0.35–0.97)20.

The safety data were also confirmed in the TMC114IHIV4042 study13, where the DRV/r-based treatment was well tolerated, with only 3.0% of the treatment discontinuations due to AEs. Notably, no differences were observed in the AE/SAE types and/or frequencies in this study compared to those reported in the DRV/r RCTs1,2,13.

Moreover, following the recently presented D:A:D cohort data on cardiovascular risk in HIV-positive subjects treated with DRV/r-based regimen26, we showed that the cardiovascular events rates in all studies included in this meta-analysis were low, even though the observational period was approximately three years compared to the more than six-year observation period included in the D:A:D26.

Triple therapy proved to be superior in efficacy (defined as viral suppression) to monotherapy. Patient characteristics (i.e., residual viremia and a nadir CD4+ count <100 cells/μL) should be taken into account when establishing a monotherapy regimen, as highlighted by Gianotti et al.27, who reported selection criteria for entry of candidate virologically suppressed HIV-positive individuals into DRV/r monotherapy27. Following this scoring system, DRV/r monotherapy and standard therapy “could be equally effective” with the same virological failure rate as standard triple therapy27.

No mutations associated with DRV resistance were reported for monotherapy based on DRV/r, and sensitivity to DRV was maintained28,29,30. To date, no INSTI drug has shown the same genetic barrier: INSTI resistance-associated mutations have been found in failing monotherapy31. In terms of safety, adverse events leading to therapy discontinuation were relatively rare and were even rarer in the monotherapy studies28,29,30.

Limitations

The limitations of this meta-analysis include the use of different comparators in the studies, inhomogeneity in the study duration, the use of different timepoints for the efficacy/safety assessments, the wide timespan of the studies considered and the inclusion of only English-language publications. All the RCTs included were open-label; therefore, the risk of performance bias was increased. However, the outcomes evaluated were objective measures, which might have decreased the risk of bias. Furthermore, this analysis only included studies using DRV boosted with the pharmaco-enhancer ritonavir. However, the results of two recent registrative studies conducted with naïve and virologically suppressed, experienced patients taking ART based on DRV boosted with the new pharmaco-enhancer cobicistat have been published32,33. Further research including those data are recommended.

Strenghts

The main strength of this meta-analysis is the comprehensive search for published clinical studies from multiple electronic databases using a cross-checking strategy for additional potentially missed articles. The meta-analytic approach allowed us to obtain more precise estimates of the pooled results, which can provide clinicians with suggestions for use in clinical practice, as previous meta-analyses have done34,35. Furthermore, the studies considered here were conducted in different years; therefore, the patient characteristics differed greatly among the studies (in previous years, the patients were more advanced). Nevertheless, the results shown in response to DRV treatment were consistent and confirmed its well-known efficacy and safety profile; thus, this treatment remains an effective option for current patients.

Using this meta-approach, we re-analyzed study-level data; however, additional original studies involving a longer follow-up period and patients enrolled in real-life settings are required to better understand the efficacy, effectiveness and safety of DRV/r.

Conclusion

The evidence shown in this analysis confirms that DRV/r is an effective regimen for ART-naive and ART-experienced subjects, with no differences from the comparator arms detected. DRV/r was safe and well-tolerated in every group of subjects. The good safety profile of DRV when used in monotherapy is highlighted.

Methods

Search strategy

A systematic literature search of clinical trials including DRV use in HIV-positive patients was conducted in September 2016 using the Medline and EMBASE databases. No data were generated in this work, which analyzed publicly available publications. We did not prepare a specific review protocol for this project. We adopted a wide-ranging search strategy using a predefined generic search string with no temporal restrictions and no search filters whenever possible. This strategy was finalized to minimize the probability of excluding relevant papers from the present meta-analysis. The Medline/Pubmed search string was as follows: “(darunavir OR prezista OR tmc114) AND trial”. A similar combination of keywords was used in the EMBASE search; however, that search was restricted to clinical trials using the “study types” filter. A cross-check for additional articles that were potentially missed during the main search process was conducted by exploring the Cochrane Register of Controlled Trials (CENTRAL) and Google Scholar (using the same keywords and reviewing the first 150 papers according to their relevance) and performing thorough searches of the reference lists of relevant reviews and the papers selected for inclusion. Figure 1 provides a flow-chart with detailed information on the search and selection processes.

Inclusion and exclusion criteria

The identified publications were considered for inclusion in the meta-analysis if the following criteria were met: randomized clinical trials with at least 48 weeks of follow-up and with DRV use in at least one study arm. Observational studies, interventions other than DRV use, reviews, meta-analyses, indirect comparisons, commentaries and other articles lacking original data were excluded. Single-arm trials and pooled analyses were also excluded after careful consideration. Conference abstracts were included, whereas unpublished studies and articles in languages other than English were excluded. No studies were excluded a priori for weakness of design or data quality.

Study selection, data extraction and risk of bias assessment

Two researchers independently examined the articles retrieved from the Medline/PubMed and EMBASE databases. Discrepancies between the researchers’ results were discussed and resolved. In the first selection step, the articles were evaluated based on their titles and abstracts. After merging the publications from the PubMed and EMBASE searches, a total of 134 unique publications remained. The second and third selection steps were based on full-text examinations of the retrieved articles. Sixty articles reporting data on the efficacy or safety of DRV in HIV-positive patients from RCTs with at least 48 weeks of follow-up were retained. Fourteen of these studies were included in the tables but were not used in the meta-analyses due to the relatively small number of studies with their specific characteristics (i.e., they reported results for follow-up periods other than 48 or 96 weeks or they reported results from trials on treatment-experienced subjects with switched or mixed treatments).

Two researchers reviewed the selected studies and extracted relevant information. In particular, the extracted data included the trial name, enrollment period, geographic area, number of patients included and treatment regimen in each study arm, the reason for discontinuation of earlier treatments (for studies with treatment-experienced patients), the patient characteristics at baseline, and the follow-up duration. This information was organized in two tables that separated the trials with treatment-naïve and treatment-experienced patients. The latter patients were further divided into subgroups representing trials of (i) treatment-experienced failing subjects treated with a DRV 600 mg BID regimen compared with another regimen, (ii) treatment-experienced virologically controlled subjects treated with a DRV 800 mg regimen compared with another regimen, and (iii) treatment-experienced subjects treated with a mixed/other DRV regimen. The main results for the efficacy (i.e., viral suppression defined as <50 copies/ml) and safety outcomes (i.e., treatment discontinuation due to adverse events or serious adverse events) were also extracted into spreadsheets for subsequent meta-analyses. Whenever available, we extracted the results from the intention-to-treat analysis. Discrepancies between researchers were checked in the original reports and resolved.

The risk of bias in the included studies was assessed by three authors using the Cochrane risk of bias tool36. Discrepancies between the researchers were discussed and resolved through discussion with a senior reviewer.

Statistical analyses

In the efficacy outcome analyses, the results obtained at weeks 48 and 96 and for the treatment-naïve and treatment-experienced patients were always analyzed separately. However, for the safety outcomes, all trials were jointly analyzed using the results for the longest follow-up time when several results were available from the same trial. The risk ratios for each study were pooled. When the risk ratio was not provided but sufficient data were available in the publication to compute this measure, we calculated unadjusted risk ratios and their 95% CIs from the outcome distributions of subjects in the treatment and control arms. When more than one publication reported results from the same study (i.e., with extended follow-up periods), we included the earliest publication in the meta-analysis because the completion rate was higher and the endpoint was more similar to those of the other studies. The ACTG5257 study37 was a three-arm trial. Therefore, we pooled data from the ATV and RAL arms to compute a single risk ratio for each efficacy and safety outcome. These ratios were included in the meta-analysis.

We computed summary risk ratios (RR) for each efficacy and safety outcome for the patients treated with DRV compared to other treatments using random-effects models (i.e., as weighted averages using the inverse of the sum of the variance of the log (risk ratio) and using the moment estimator of the variance between studies as the weight)38,39. Heterogeneity between trials was assessed using the χ2 test (defined as a p-value less than 0.10), and inconsistency was measured using the I2 statistic, which describes the percentage of total variation across studies due to heterogeneity rather than chance40. Values of the I2 statistic of approximately 25%, 50% and 75% are indicative of low, moderate and high heterogeneity, respectively40. The presence of publication bias was assessed based on a visual examination of the funnel plots and by applying the tests proposed by Begg and Mazumdar41 and Egger42. We conducted sensitivity analyses by excluding each study one by one from the meta-analysis. No other sub-group analyses were planned. All statistical analyses were performed using the RevMan software (version 5.3 for Windows).