High Prevalence of Drug Resistance Mutations Among Patients Failing First-Line Antiretroviral Therapy and Predictors of Virological Response 24 Weeks After Switch to Second-Line Therapy in São Paulo State,Brazil

Abstract

Universal antiretroviral treatment with sustained viral suppression benefits patients and reduces HIV transmission. Effectiveness of therapy may be limited by antiretroviral drug resistance. Information on the resistance profile at treatment failure and its impact on antiretroviral drugs may subsidize subsequent treatment strategies. Partial pol sequences from 319 patients failing first-line therapy were analyzed for resistance associated mutations (RAMs) and HIV subtype. Demographic data, CD4 T cell count, viral load, and antiretroviral regimens and mutational profile at first-line failure were also investigated for associations to the response to second-line regimens. RAMs at the reverse transcriptase gene were frequent. Most sequences (88%) showed at least one mutation. A higher number of reverse transcriptase RAMs were associated to lower CD4 T cell counts and the use of tenofovir/lamivudine in first line. Among 205 with follow-up data, 76.6% were virally suppressed (below 200 copies/ml) after 24 weeks of second-line therapy. Most cases initiated second line with a regimen genotypic susceptibility score ≥2, but it did not predict viral suppression, that was independently associated with higher CD4 T cell counts and with the presence of nucleos(t)ide analog reverse transcriptase inhibitor (NRTI) RAMs. This study documented extensive resistance at first-line failure in this area in Brazil, highlights the risks of low CD4 T cell counts to second-line therapy, and supports the notion that recycled NRTIs may contribute to viral suppression even when genotypic resistance is present.

Introduction

Universal antiretroviral therapy (ART), along with preventive measures, may control, and eventually end, the HIV epidemic. Modern ART suppresses viremia in most adherent patients, rendering them virtually incapable of transmitting the virus.^1,2 Drug resistance may be an obstacle to viral suppression, compromising ART effectiveness and may negatively impact epidemic control.

Highly active antiretroviral combinations have been freely available through the Brazilian health system since the late 90s. Currently, about half a million patients are on ART in the country. Genotypic resistance tests and viral load are provided by the public health system to guide second-line and subsequent salvage regimens.³

From 2004 to March 2017, the Brazilian guidelines have recommended an efavirenz (EFV)-based regimen as first-line treatment. In 2013, tenofovir (TDF) replaced zidovudine (AZT) as the preferred nucleos(t)ide analog reverse transcriptase inhibitor (NRTI) to be combined with lamivudine (3TC).³ Nevirapine (NVP), ritonavir boosted atazanavir (ATV/r), and lopinavir (LPV/r) were available in special situations, including pregnancy, intolerance, or contraindications to the preferred regimen. Recently the integrase inhibitor dolutegravir became the recommended drug to be used with TDF/3TC in first-line therapy, with darunavir/r added as an alternative protease inhibitor (PI).

Data on HIV resistance profile in patients failing first-line regimens in Brazil are limited. In a study evaluating genotype reports from five cities, the authors found a high prevalence of resistance associated mutations (RAMs) after first-line failures, including M184V/I in 70% and K103N in 42% of cases.⁴ Proportions of NRTI RAMs as high as 80% for M184V/I were observed in northern Brazil, but in this analysis patients who had failed two or more regimens (76% of the cases) were included.⁵ Similarly, studies in Asia and Africa found a high prevalence of RAMs after first-line failure.^6,7 A French nationwide study, including patients with two consecutive viral load measures over 50 copies/ml, showed a somewhat lower resistance prevalence (56%), but information on the number of previous regimens was not available.⁸ A comparable proportion of resistance (52%) was observed in patients failing the first-line regimen in China.⁹ In both studies, patients were likely to be at early stages of failure when genotyping tests were performed. Since the mid-2000s, a decrease in resistance has been recorded in the United States¹⁰ and France.⁸ In Brazil, the assessment of 2.115 HIV-1 sequences from patients failing ART from 2002 to 2012 showed a decrease in the prevalence of PI and NRTI resistance, in contrast to an increase in resistance to non-nucleoside reverse transcriptase inhibitor (NNRTI), mainly due to K103N.¹¹

Evidence on the impact of RAMs on the efficacy of second-line regimens was also limited until recently, when three large randomized studies consistently showed that NRTI mutations do not compromise NRTI-based second line success.^12
–14 In contrast, lack of adherence, high viral load,^12,13 and low CD4 T cell counts¹⁵ were strong predictors of second-line failure in randomized, as well as in observational, studies.

Data on the HIV mutational profile after first-line ART failure and correlates of second-line outcomes in real-life context in Brazil are scarce. In this study we evaluated viral drug resistance in 319 patients failing first-line ART in the state of São Paulo and their 24-week response to the second-line regimen.

Materials and Methods

The retrovirus laboratory of Adolfo Lutz in São Paulo, Brazil generated 458 pol sequences from patients failing their first antiretroviral regimen between 2013 and 2015. Sequences were analyzed for RAMs using the HIVdb algorithm v7.0.1 (Stanford University HIV Drug Resistance Database). A report was sent to the clinical services to subsidize treatment decision. We reviewed the national antiretroviral dispensing database (SICLOM) for each patient to confirm treatment history at sample collection. As this system was introduced and became consistently used by different clinical units at different time points, cases with an HIV diagnosis preceding ART initiation in more than 1 year were considered as having potential previous treatment exposure and were retained in the dataset but excluded in some analyses. Efforts were made to carefully review available documents and electronic systems to confirm information on treatment history provided by the assistant physicians at request forms and to obtain data on the virological and immunological outcomes. Patients with inconsistent incomplete information or with evidence of previous exposure to ART were excluded. We also reviewed viral load and CD4 T cell databases to identify cases with possible undocumented ART (e.g., changes over 0.5 log₁₀ in viral load along with consistent CD4 T cell gains). These cases were excluded, as well as those with conflicting clinical information, with 319 cases with viral failure (viral load above 200 copies/ml after 6 months on treatment) included in this study. We then recorded the second-line regimen and dispensation date and searched the laboratory database for the first available CD4 T cell count and viral load measurements after 24 weeks of the recorded date. Cases were henceforth anonymized.

All sequences were generated from retrotranscribed plasma RNA, extracted with QIAamp Viral RNA Mini Kit (Qiagen, Germany) or a M2000 extractor (Abbott). A reverse transcriptase (RT)-polymerase chain reaction (PCR) was performed using commercial kit (TRUGENE; Siemens) or with a previously described “in-house” protocol.¹⁶ Briefly, a one-step RT-PCR, using High Fidelity Taq platinum and SuperScript III, was followed by nested PCR and Big Dye (Life) incorporation, resolved in an ABI 3130XL.

Sequences were edited with Sequencer 4.7 software (Gene Codes) and/or with RECall (http://pssm.cfenet.ubc.ca/), subsequently analyzed at NCBI (www.ncbi.nlm.nih.gov/projects/genotyping/help.html) and REGA Genotyping tools (http://regatools.med.kuleuven.be/typing/v3/hiv/typingtool) to evaluate HIV subtype and at Stanford HIV database Genotypic Resistance Interpretation (Stanford db, https://hivdb.stanford.edu/hivdb/by-sequences/) for Drug Resistance Mutations. Genotypic information was expressed as individual IAS RAMs list,¹⁷ categorized as mutations to NRTI or NNRTI or major PI mutations. In addition, the Stanford db individual drug genotypic susceptibility score (GSS) was used to predict the collective impact of RAMs on each drug; score 1 when full drug activity was predicted (no resistance), 0 if no drug activity was predicted, 0.25 for intermediate resistance, 0.5 for low resistance, and 0.75 for potential low resistance. We expressed the Stanford db score as (1) total GSS, the sum of predicted resistance to all the 19 drugs acting at the pol (RT or PT) analyzed at Stanford db and as (2) regimen GSS, considering only the drugs used in each patient's second-line therapy. Integrase region was not analyzed, and raltegravir, when part of the regimen, was assumed as fully active (GSS = 1).

CD4 T cell counts were performed by flow cytometry (BD) and HIV viral load by real time quantitative PCR (Abbott) after July 2013. Versant HIV-1 RNA 3.0 Assay (bDNA; Siemens) was used before July 2013. Viral load determination on samples collected for genotyping was performed at our reference laboratory. CD4 T cell counts and additional viral load determinations were performed at one of the national viral load/CD4 laboratories.

Virological response was dichotomized according to Department of Health and Human Services (DHHS) metric as suppressed (below 200 copies/ml) or viremic.¹⁸ To evaluate the significance of dichotomous variable, we used two-tailed Fisher or Yates corrected, as appropriate (Epi6, CDC). Mann–Whitney was used to compare variables from two groups, Kruskal–Wallis for three or more groups, and Spearman correlation to evaluate two continuous variables (Stata 8; Stata Corporation). Logistic regression (Stata 8) was used to evaluate independence of associations, with variables included in the regression when the unadjusted significance was below 0.2. The study was approved by Institutional Committee (IAL84d2010) and Ethics Committee (CEP-IAL/0032010).

Results

Our laboratory generated 2,629 partial pol sequences from July 2013 to December 2015. Test request forms allowed for selection of 458 patients referred for testing due to first-line treatment failure. After exclusion of 139 cases with incomplete or conflicting information, 319 patients failing first-line therapy were available for analyses.

Most participants (71%) were men, and the median age was 40 years old (interquartile range [IQR] 32–48), with 12 cases below 12 years of age. Most patients were asymptomatic at collection. Fifty-nine cases (17%), although included as failing first line, started follow-up before electronic database could rule out previous nondocumented ART exposure and were excluded in some analyses (see Mutations according to first-line regimen section). Subtype B (74%) predominated and followed subtype F (11%), BF recombinants (8%), and subtypes C (6%). Two BC recombinants and one AG recombinant were also identified. Demographic and laboratory data available at genotype test collection are shown in Table 1.

Table 1.

Demographic and Laboratory Characteristics at the Time of Genotype Collection According to First-Line Antiretroviral Regimen

	All	AZT based	EFV/AZT/3TC	NVP/AZT/3TC	PI/AZT/3TC	TDF based	EFV/TDF/3TC	NVP/TDF/3TC	PI/TDF/3TC	Other ^a
No. of patients	319	214	137	27	50	100	87	2	11	5
Male sex (%)	71	66	75	44	56	82	81	100	91	40
Age,^b years	40 (32–48)	42 (34–48)	42 (35–49)	41 (26–47)	41 (33–49)	39 (30–46)	38 (30–44)	55 (49–60)	49 (31–57)	18 (14–35)
CD4 count,^b cells/mm³	295 (150–502)	327 (186–558)	307 (173–504)	534 (250–896)	337 (211–581)	206 (77–360)	203 (61–356)	273 (18–527)	222 (144–521)	261 (195–408)
HIV RNA,^b log₁₀	4.00 (3.26–4.71)	3.81 (3.15–4.47)	3.96 (3.40–4.57)	3.26 (2.85–4.17)	3.59 (2.98–4.64)	4.51 (3.67–5.03)	4.59 (3.66–5.07)	5.14 (4.74–5.54)	4.09 (2.74–4.47)	3.08 (1.97–3.11)
Time of diagnosis,^b months	51 (19–99)	73 (38–108)	77 (46–110)	53 (29–101)	72 (24–105)	18 (9–43)	16 (9–44)	13	30 (23–40)	9
Time on ART,^b months	31 (12–72)	48 (27–81)	48 (30–82)	40 (26–101)	49 (21–76)	10 (7–22)	10 (7–20)	9 (9–11)	23 (20–38)	21 (8–107)
Subtypes (%)
HIV-1 B	74	72	79	70	54	77	79	100	64	60
HIV-1 C	6	3	3	0	4	11	12	0	9	40
HIV-1 F	11	15	12	15	20	4	3	0	0	0
Recombinants	9	10	6	15	22	8	6	0	27	0

PI used included LPV or ATV, with ritonavir boost.

Other regimens: EFV/AZT/DDI; AZT/DDI; EFV/ABC/3TC; LPV/ABC/3TC; ATVr/ABC/3TC.

Results expressed as median and IQR.

3TC, lamivudine; ABC, abacavir; ART, antiretroviral therapy; ATV, atazanavir; AZT, zidovudine; DDI, didanosine; EFV, efavirenz; IQR, interquartile range; LPV, lopinavir; NNRTI, non-nucleoside reverse transcriptase inhibitor; NRTI, nucleos(t)ide analog reverse transcriptase inhibitor; NVP, nevirapine; PI, protease inhibitor; TDF, tenofovir.

Mutations according to first-line regimen

Most sequences (88%) showed at least one RAM, 79.3% with one to seven NRTI mutations. Among the 256 sequences from patients using NNRTI, 92% had one to six NNRTI mutations. Mutations at codon K103 were observed in 56.7% of them.

Most patients used either AZT/3TC (67%, n = 214) or TDF/3TC (31%, n = 100) as NRTI backbone. Duration on therapy was longer among AZT than TDF users (median 48 months, IQR 27–81 vs. 11 months, IQR 7–22; p < .0001). The proportion of patients using AZT or TDF with at least one NRTI RAM (79.4% vs. 80%) or NNRTI RAM (78.5% vs. 79%, respectively) was similarly high, but the number of mutations differs. Although the number of TAM1 and TAM2 was higher among AZT users, the total number of NRTI mutations were significantly higher among TDF users (1.78 ± 1.4 vs. 1.39 ± 1.3 for AZT, p = .003), mostly due to discriminatory mutations. The number of NNRTI mutations were also higher among TDF users (2.01 ± 1.4 vs. 1.46 ± 1.0 for AZT users; p = .0003). These differences remained significant even when patients with a long period from HIV diagnosis were excluded, as well as when analysis was restricted to patients on NNRTI or on EFV-based schemes, but not if only cases with higher CD4 (>500 cells/mm³) were considered (data not shown).

M184V/I were the most common mutations, observed in 74.3% of all cases. K65R was observed in 32% of patients exposed to TDF/3TC. Table 2 describes NRTI RAMs according to first-line treatment. The proportions of samples with K65R were higher among subtype C (46%) and F (75%) than in subtype B (30%) or recombinant (20%) genomes, but the differences were not significant (p = .5).

Table 2.

Resistance Mutations: Percent Occurrence According to First-Line Antiretroviral Regimen

ART regimen			Discriminatory mutations					Thymidine analog mutations
NRTIs	NNRTI/PI	No.	184VI	65R	70EQG	74VI	115F	41L	67NG	70R	210W	215YF	219QE	69ins
AZT/3TC	EFV	137	81	0	0	0	0	9	6	13	7	15	7	0
	NVP	27	93	0	0	0	0	15	0	11	0	22	7	0
	PI	50	58	0	0	0	0.5	6	4	4	6	6	2	0
Total		214
TDF/3TC	EFV	87	76	37	9	14	9	3	6	5	0	3	5	0
	NVP	2	50	0	0	50	50	0	0	0	0	0	0	0
	PI	11	45	0	0	0	0	18	0	0	18	18	0	9
Total		100
Other regimens^a		5	0	0	0	0	0	20	40	0	20	20	0	0

PI used included LPV (lopinavir) or ATV (atazanavir), with ritonavir boost. Mutation Q151M and codon 69 deletion were not observed.

Other regimens: EFV/AZT/DDI; AZT/DDI; EFV/ABC/3TC; LPV/ABC/3TC; ATVr/ABC/3TC.

Considering the 63 patients failing PI-based regimens, one to five major PI mutations were observed in 25%. In this subgroup, a NRTI mutation was detected in 60% of the cases; this proportion was significantly higher when PI mutations were present (94% vs. 49%, p = .004). Among these patients, 9% had nonpolymorphic (98AG, 101E, K103N, 190A, or 225HP) and 10% polymorphic NNRTI mutations (V90I, 106I, or 138A).

Mutations and CD4 T cell counts

Patients with virus harboring at least one mutation at genotyping tended to have lower CD4 T cell counts than those without RAMs (277 cells/mm³ vs. 336 cells/mm³, p = .046). A similar difference was observed for the presence of NNRTI mutations (271 vs. 335, p = .03). A CD4 T cell count lower than 200 cells/mm³ was strongly associated with a higher number of NRTI (mean 2.07 vs. 1.39, p < .0001) and NNRTI RAMs (2 vs. 1.6, p = .008). However, if M184V/I was the only NRTI mutation detected, CD4 T cell values were significantly higher in comparison with cases with M184V/I plus additional NRTI mutations (348, IQR 222–582 vs. 82, IQR 55–246 cells/mm³, p < .0001).

Overall, patients failing TDF tended to have a lower CD4 count at genotype test than those failing AZT (206 cells/mm³ vs. 327 cells/mm³, p < .001), but the CD4 T cell counts from cases failing TDF with or without K65R were comparable (182 cells/mm³ vs. 211 cells/mm³, p = .45).

Second-line therapy characteristics

A second-line therapy was documented for 259 patients (81%). The regimen was initiated before the availability of the genotype test results in 9% of these cases. The median time from genotype collection to initiation of the second-line regimen was 2.2 (IQR 1.4–4.2) months.

The NRTI backbone combination most frequently selected was TDF/3TC (75% of patients), with AZT/3TC used by 16% of the patients. Regimens with three NRTI (6%), single NRTI (1%), and no NRTI (2%) were also used.

Most regimens (93.8%) included a boosted PI (46.7% ATV/r, 40.5% LPV/r, 4.6% darunavir/r, and 1.9% fosamprenavir/r). Nonanalogous nucleosides were used by 18 (8.5%) patients (etravirine in four cases), and the integrase inhibitor raltegravir was used by 31 (12%) patients.

Considering all 19 drugs evaluated at HIV db, the median global GSS was 13.5 (IQR 11–16), while the median GSS for the NRTI class was 4.25 (IQR 2.25–4.25). Full activity to all drugs was detected in 19 sequences (7.3% of patients). The median regimen GSS predicted for the 259 second-line antiretroviral combinations was 2 (IQR 2–2.5, range 1–4), whereas the median regimen GSS for the NRTI backbone was 1 (IQR 1–1, range 0–3).

Predictors of viremia control after week 24

Of the 259 patients with a documented second-line prescription, 205 patients (79%) had an available viral load measurement after 24 weeks after initiating the regimen. The median time for the first measurement after this period was 34 weeks (IQR 27–39). Most patients (76.6%) had a viral load less than 200 copies/ml at this measurement. In four cases, viral suppression was documented before completing 24 weeks of second-line therapy, but confirmed at a later collection. For the 149 patients with available CD4 T cell counts after week 24, the median gain was 80 cells/mm³ (IQR 11–195). After week 24, viremic cases had a net loss of 9 cells/mm³ (IQR −49 to −64), whereas those suppressed, a net gain of 111 cells/mm³ (IQR 10–217). There was a strong correlation between CD4 T cell counts at genotype sampling and after 24 weeks of second-line therapy (Spearman's rho = 0.75, p < .0001).

Compared to patients with a CD4 T cell count above 200 cells/mm³, those with CD4 T cell counts below 200 cells/mm³ at genotyping were less likely to achieve viral suppression at week 24 (64% vs. 83%, p = .004).

Most patients with viral load less than 1,000 copies/ml at genotype sampling were virally suppressed at week 24, but there was no significant difference between them and those with viral load above 1,000 copies/ml at sampling (89% vs. 75%, respectively, p = .14). Likewise, higher viral load at collection was not associated to lack of response to the second-line regimen. There was no noticeable association between viral subtype or other variables with viral suppression after week 24 (Table 3).

Table 3.

Demographic and Laboratory Variables and Viral Suppression (Viral Load <200 Copies/ml) After Week 24 of Second-Line Therapy

	Unadjusted			Adjusted
	Odds ratio	p	95% CI	Odds ratio	p	95% CI
Age above 40 years	1.06	.860	0.55–2.02
CD4 T count >200 cells/mm³	2.82	.002	1.45–5.47	2.56	.010	1.25–5.22
Male sex	1.43	.339	0.69–2.98
Time on treatment >3 years	1.18	.67	0.55–2.52
PI treatment	0.72	.481	0.30–1.77
Recombinant genome	2.77	.183	0.62–12.46	2.65	.225	0.55–12.84
AZT treatment	0.91	.790	0.45–1.84
Viral load <1,000 copies/ml	2.82	.103	0.81–9.78	2.04	.296	0.54–7.81
Viral load >100,000 copies/ml	0.90	.805	0.37–2.15
Second line GSS ≥2	1.40	.420	0.618–3.17
NRTI mutation^a	4.15	.001	1.89–9.07
NNRTI mutation^a	1.17	.721	0.49–2.83
PI mutation^a	2.51	.392	0.30–20.56
M184I/V mutation	4.05	.000	1.94–8.43	4.37	.000	2.01–9.48

Logistic regression of variables at the genotyping collection associated to virological suppression after week 24. AZT use in first-line NRTI compared to TDF. Adjusted analysis using any NRTI mutation, instead of only M184I/V, also shows significant independent association, along with low CD4T cell count, to viral suppression.

Mutations according to IAS list.

GSS, genotypic susceptibility score.

The predicted GSS for the second-line regimen did not correlate with virological success. On the contrary, greater predicted regimen activity was associated with worse viral load suppression. Indeed, while failures were documented in 38% of the 47 patients using three or more active drugs, it was observed in only 19% of the 158 patients with less than three active drugs (p = .01). Of the 35 cases with a regimen GSS >2, 71.4% achieved suppression, with 91% of the 11 cases suppressing with a regimen GSS = 1.

The presence of at least one NRTI RAM, detected in 69% of the viremic and in 89% of the viral suppressed patients, was associated to virological success (p = .001). The total number of NRTI mutations among cases with virological success (mean 1.75 ± 1.39) was higher than in those with failure (1.17 ± 1.14, p = .0063). As nonadherent patients could be overrepresented in the group with no mutations, we restricted the analysis to those cases with at least one NNRTI or PI mutation, and the association between the presence of NRTI mutations and virological success at 24 weeks persisted. On the same line, the significant difference remained after excluding from the analysis cases with a GSS = 1 to all antiretroviral drugs (78% of suppression for cases with no NRTI RAMs vs. 95% for cases with NRTI RAMs, p = .002).

When each NRTI RAM was evaluated individually, the M184V/I, present in 82.4% of virally suppressed cases and in 54% of viremic cases, showed a strong association with virological success (p = .0002). Except for mutations at codons K65 and K70, which were somewhat more prevalent in failure cases, most NRTI RAMs were more prevalent in the group of virological success. Figure 1 depicts NRTI RAMs according to virological outcome.

FIG. 1.

Proportion of mutations (IAS list) to nucleos(t)ide analog reverse-transcriptase inhibitors detected at genotype test performed during first-line failure according to virological outcome to second-line treatment, considering suppression a plasma HIV RNA determination <200 copies/ml at the first test after 24 weeks of second-line therapy.

Suppression was not related to the presence of either NNRTI (77% with vs. 74% without mutations, p = .9) or PI RAMs (89% with vs. 76% without PI mutation, p = .69).

Logistic regression analysis of the associations between variables and virological success is shown in Table 3. CD4 cell counts above 200 cells/mm³ (OR 2.8, 1.5–5.5, p = .002) and the presence of NRTI RAM were independently associated to virological success (OR 4.2, 1.9–9.1, p = .001). However, when only cases with an NRTI mutation other than M184V/I were considered, this association was not significant (OR 1.38, 0.7–2.7, p = .35). In contrast, the detection of M184V/I at first-line failure, independent of the presence of other RAMs, showed a strong association with virological success after the switch to a second-line regimen (OR 4.1, 1.9–8.4, p < .001). CD4 T cell counts above 200 cells/mm³ remained independently associated to viral suppression either if any NRTI RAM or only M184V/I was used in the regression.

Discussion

Resistance to antiretroviral drugs may cause treatment failure or emerge due to lack of viral suppression. In any case resistance profile may impact subsequent therapy. Our aim was to describe RAMs in patients failing first-line therapy, CD4 T cell counts and viral load at the time of genotype test, and, most importantly, its impact on short term (week 24), second-line therapy response. We evaluated in this study HIV sequences from 319 individuals failing first-line therapy. Overall, presence of RAMs at RT, both to NRTI and NNRTI drug classes, was common, with at least one IAS list mutation in 88% cases. These high rates have been observed in studies in Africa and other resource limited settings^19
–21 and could reflect accumulation of mutations during failure. However, in a EuroSIDA study,²² a high proportion of TAM1 was observed in Europe at 1 year of failure, with a lower than anticipated rate of accumulation among those maintaining the failing regimen. This might suggest that patients acquire most resistance in the first months of failure, with a slow accumulation thereafter.

We were able to compare two commonly used nucleos(t)ide combinations, AZT/3TC and TDF/T3C. Patients using TDF/3TC tended to have more RAMs, but different factors might have confounded our findings. We made subgroup analyses as evaluating only cases with similar time on treatment, cases with treatment initiation at more recent years, and excluding those using PI (as more AZT users had a PI as a second class), but the number of RT RAMs remained consistently higher among TDF users. These differences lose significance at higher CD4 cell counts. Differences in adherence to the regimens could influence, but at least one surrogate for adherence, the proportion of cases with any mutation, was similarly high (79% vs. 80% for NRTI RAM) for both combinations.

The prevalence of TDF resistance observed in our study (32%) was higher than that observed at the TenoRes study for patients in high-income settings (present in 25% of virological failures), but was in the range of that observed in sub-Saharan Africa (28%–50%).¹⁹ The literature is not conclusive whether virological failure on TDF or AZT containing regimens leads to more resistance mutations. A Systematic Cochrane review in 2010 found more resistance in TDF users, but the conclusions were based in limited information, and the authors warn for the fact that the studies reviewed evaluated different drug combinations.²³ Some articles point to more mutations among AZT users,²⁴ while others document a higher prevalence among TDF users.²⁵ This latter study was done in South Africa where subtype C predominates, whereas our study population was mostly subtype B infected. The small sample size of non-B subtypes in our study hinders a proper evaluation. However, our data tended to be consistent with clinical and in vitro data suggesting that subtype C viruses are more susceptible to developing the K65R.^26,27

We found a relatively high proportion of NNRTI in patients failing PI at first line. However, many mutations are polymorphisms that were already detected among naive cases in the regions,¹⁶ and nonpolymorphic residues, as at codon K103, show similar prevalence to that observed in different transmitted resistance studies in the area.^16,28 Although we tried to exclude all cases with undocumented antiretroviral exposure, we cannot totally rule out this possibility.

Another objective of our study was to evaluate the response to a second-line therapy. To access the short-term (week 24) response to second-line therapy, we used public laboratory electronic database to search for viral load results. Cases using public antiretroviral dispensary but followed at private laboratories are not to be part of these databases, so the actual denominator to calculate the response to second line could not be determined. Considering this limitation, we observe that most patients (76.7%) in the study were able to suppress viremia after 24 weeks on second-line regimen. NRTI drug recycling was common, with many patients switching one of the NRTI, but keeping 3TC. The extend of NRTI recycling did not show any impediment to suppression to a second line, nor the NRTI combination used at first line, and comparable rates of viral suppression after week 24 were documented in cases failing a regimen with AZT/3TC (76.4%) or TDF/3TC (78.1%). Rates of suppression were also similar across second-line regimens (data not shown), but the study did not have power to discriminate individual regimen success rate.

The predicted second-line regimen GSS did not correlate to virological success and it was actually higher in nonsuppressing cases, but 96% of the cases started second line with a fully active PI or integrase inhibitor. Even when we exclude cases where the genotype test did not detect any mutation (potentially cases with inadequate adherence), still the GSS tended to be comparable, or even higher, among nonsuppressing cases. This finding is similar to the study of Jiamsakul, with a median GSS for the second-line regimen of 2 and no significant association with virological outcome.²⁹ This study also documented a relation of low CD4 T cell counts to the presence of RAMs.

Multivariable logistic regression shows that both low CD4 T cell counts and the presence of NRTI RAMs were independently associated to virological success after week 24. However, when we evaluated individual mutations, the presence of M184V/I was necessary to observe this association and cases with only other NRTI mutations did not show the association of NRTI RAM to suppression. However, the number of cases with only NRTI RAMs other than M184V/I was small, so the study cannot evaluate properly the role of other individual RAM.

Mutations at codon M184 (valine or isoleucine) confer high cytosine analog resistance, reducing susceptibility to these drugs >100-fold and low-level resistance to abacavir. However, it increases susceptibility and slows the emergence of AZT and TDF resistance,^29
–31 a fact that may favor these drugs in second-line therapy. M184V/I is associated with reduced viral replication in vitro and in vivo.^32,33 In the TenoRes study the prevalence of M184V/I was highest in sub-Saharan Africa and Latin America and lowest in western Europe.¹⁹ It is important to note that all our cases used generic 3TC, and results might differ for emtricitabine, but recent meta-analyses suggest comparable clinical outcome with either drug.³⁴ Due to its negative impact in viral fitness, M184 RAM may reverse in the absence of therapy more often than other mutations. Therefore its presence could be only a good marker of adherence. However, one cannot rule out both its effect in TDF and AZT activity, as well as its direct impact in viral fitness. The significantly higher CD4 T cell counts among cases with only M184 RAM, compared to cases with only another NRTI RAMs, may support this assumption. Recent data on the evolution of dolutegravir resistance suggest that it may be limited by the effect of M184 mutations on viral fitness.³⁵ If RAM at codon M184 is proven to be associated to salvage therapy success, for any reason, it may be part of simpler RAM detection point of care devices, which could identify key mutations at positions such as M184 and K65, subsidizing salvage regimens.

The retrospective nature of the study, different exposure time to the first-line regimens, longer among AZT users, and lack of adherence information are clear limitations to the study. We could however evaluate this issue in a rather homogeneous population, with similar access to antiretroviral drugs and monitoring resources. Our findings are in line with large trials^12
–14 that show that NRTI resistance or class recycling is not an impediment to virological success of second-line therapy. However, long-term monitoring is needed to guarantee sustained success of ART in limited resource settings.

Conclusions

Patients failing first-line regimen in public services in São Paulo have high rates of mutations to first-line therapy, but the use of second-line PI based regimens with partially active or inactive NRTI was able to suppress most patients. A high CD4 T cell count at genotype and the presence of NRTI RAMs, especially M184V/I, were independently associated to virological success. Evaluation of long-term suppression is necessary to further support the usefulness of partially active analogs in second and subsequent therapeutic regimens.

Footnotes

Acknowledgments

The authors are grateful to study participants, clinical staff, and other caregivers from the public services involved in the study and to State and Federal AIDS program personnel responsible for the laboratory and drug dispensary data systems. Partially supported by grant no. 2013/194417 São Paulo Research Foundation. Salvage Therapy Working Group: Joao Leandro de Paula Ferreira, Norberto Camilo Campos, Cilene Proenca, and Laís Prado from the Centro de Virologia, Instituto Adolfo Lutz, São Paulo, Brazil; Denise Lotufo and Simone Queiroz from the Centro de Referência e Treinamento em AIDS, São Paulo, Brazil; Luiz Carlos Pereira Junior and Margareth Eira from the Instituto de Infectologia Emílio Ribas, São Paulo, Brazil; Mariana Carvalho e Silva de Carvalho from the Departamento de Clínica Médica, Faculdade de Ciências Médicas da UNICAMP, Campinas, Brazil; Valeria Almeida from the Centro de Referência de AIDS, Campinas, Brazil; and Rafaella S. Gomes and Flavia Gennari Pinheiro from the Ambulatório de Doenças Infecciosas, Jundiaí, Brazil.

Sequence Data

GenBank accession numbers: KY639846–KY640165.

Author Disclosure Statement

No competing financial interests exist.

References

Grinsztejn

, Hosseinipour

, Ribaudo

, et al.; HPTN 052-ACTG Study Team: Effects of early versus delayed initiation of antiretroviral treatment on clinical outcomes of HIV-1 infection: Results from the phase 3 HPTN 052 randomized controlled trial. Lancet Infect Dis, 2014; 14:281–290.

Rodger

, Cambiano

, Bruun

, et al.; PARTNER Study Group: Sexual activity without condoms and risk of HIV transmission in serodifferent couples when the HIV-positive partner is using suppressive antiretroviral therapy. JAMA, 2016; 316:171–181. (Erratum in: JAMA 2016;316:667. Erratum in: JAMA 2016;316:2048).

Protocolo Clínico e Diretrizes Terapêuticas para Manejo da Infecção pelo HIV em Adultos: Available at www.aids.gov.br/tags/publicacoes/protocolo-clinico-e-diretrizes-terapeuticas (2013), accessed September 20, 2016 .

Brites

, Pinto-Neto

, Medeiros

, Nunes

, Sprinz

, Carvalho

: Extensive variation in drug-resistance mutational profile of Brazilian patients failing antiretroviral therapy in five large Brazilian cities. Braz J Infect Dis, 2016; 20:323–329.

Lopes

, Soares

, Falci

, Sprinz

: The evolving genotypic profile of HIV-1 mutations related to antiretroviral treatment in the north region of Brazil. Biomed Res Int, 2015; 2015:738528.

Hamers

, Sigaloff

, Wensing

, et al.; PharmAccess African Studies to Evaluate Resistance (PASER): Patterns of HIV-1 drug resistance after first-line antiretroviral therapy (ART) failure in 6 sub-Saharan African countries: Implications for second-line ART strategies. Clin Infect Dis, 2012; 54:1660–1669.

Sivamalar

, Dinesha

, Gomathi

, et al.: Accumulation of HIV-1 drug resistance mutations after first-line immunological failure to evaluate the options of recycling NRTI drugs in second-line treatment: A study from South India. AIDS Res Hum Retroviruses, 2017; 33:271–274.

Assoumou

, Charpentier

, Recordon-Pinson

, et al.; ANRS AC-11 Resistance Study Group: Prevalence of HIV-1 drug resistance in treated patients with viral load >50 copies/mL: A 2014 French nationwide study. J Antimicrob Chemother, 2017; 72:1769–1773.

Zuo

, Liang

, Sun

, Bussell

, et al.: Drug resistance and virological failure among HIV-infected patients after a decade of antiretroviral treatment expansion in eight provinces of China. PLoS One, 2016; 11:e0166661.

10.

Miller

, Haddad

, Su

, Gibbs

, McColl

, Guyer

: Trends in HIV-1reverse transcriptase resistance-associated mutations and antiretroviral prescription data from 2003–2010. Antivir Ther, 2012; 17:993–999.

11.

Duani

, Aleixo

, Tupinambás

: Trends and predictors of HIV-1 acquired drug resistance in Minas Gerais, Brazil: 2002–2012. Braz J Infect Dis, 2017; 21:148–154.

12.

Paton

, Kityo

, Hoppe

, et al.; EARNEST Trial Team: Assessment of second-line antiretroviral regimens for HIV therapy in Africa. N Engl J Med, 2014; 371:234–247.

13.

Boyd

, Moore

, Molina

, et al.; SECOND-LINE study group: Baseline HIV-1 resistance, virological outcomes, and emergent resistance in the second-line trial: An exploratory analysis. Lancet HIV, 2015; 2:e42–e51.

14.

La Rosa

, Harrison

, Taiwo

, et al.; ACTG A5273 Study Group: Raltegravir in second-line antiretroviral therapy in resource-limited settings (SELECT): A randomised, phase 3, non-inferiority study. Lancet HIV, 2016; 3:e247–e258.

15.

Shearer

, Evans

, Moyo

, et al.: Treatment outcomes of over 1000 patients on second-line, protease inhibitor-based antiretroviral therapy from four public-sector HIV treatment facilities across Johannesburg, South Africa. Trop Med Int Health, 2017; 22:221–231

16.

Guimarães

, Ferreira

, Coelho

, et al.: Transmitted drug resistance among recently diagnosed adults and children in São Paulo, Brazil. AIDS Res Hum Retroviruses, 2015; 31:1219–1224.

17.

2017 Resistance Mutations Update: Volume 24, Issue 4, December 2016/January 2017. Available at www.iasusa.org/sites/default/files/uploads/2017hiv-muta-article.pdf, accessed January 20, 2017 .

18.

Panel on Antiretroviral Guidelines for Adults and Adolescents: Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents. Department of Health and Human Services. Available at http://aidsinfo.nih.gov/contentfiles/lvguidelines/AdultandAdolescentGL.pdf, Section C5, accessed January 20, 2017 .

19.

TenoRes Study Group: Global epidemiology of drug resistance after failure of WHO recommended first-line regimens for adult HIV-1 infection: A multicentre retrospective cohort study. Lancet Infect Dis, 2016; 65:1–11.

20.

Marconi

, Sunpath

, Lu

, Gordon

, et al.; South Africa Resistance Cohort Study Team: Prevalence of HIV-1 drug resistance after failure of a first highly active antiretroviral therapy regimen in KwaZulu Natal, South Africa. Clin Infect Dis, 2008; 46:1589–1597.

21.

Trotter

, Hong

, Srikantiah

, et al.: Systematic review of HIV drug resistance in Southeast Asia. AIDS Rev, 2013; 15:162–170.

22.

Cozzi-Lepri

, Phillips

, Martinez-Picado

, et al.; EuroSIDA Study Group: Rate of accumulation of thymidine analogue mutations in patients continuing to receive virologically failing regimens containing zidovudine or stavudine: Implications for antiretroviral therapy programs in resource-limited settings. J Infect Dis, 2009; 200:687–697.

23.

Spaulding

, Rutherford

, Siegfried

: Tenofovir or zidovudine in three-drug combination therapy with one nucleoside reverse transcriptase inhibitor and one non-nucleoside reverse transcriptase inhibitor for initial treatment of HIV infection in antiretroviral-naïve individuals. Cochrane Database Syst Rev, 2010; CD008740.

24.

Von Wyl

, Yerly

, Böni

, et al.; Swiss HIV Cohort Study: Incidence of HIV-1 drug resistance among antiretroviral treatment-naive individuals starting modern therapy combinations. Clin Infect Dis, 2012; 54:131–140.

25.

Van Zyl

, Liu

, Claassen

, Engelbrecht

, et al.: Trends in genotypic HIV-1 antiretroviral resistance between 2006 and 2012 in South African patients receiving first- and second-line antiretroviral treatment regimens. PLoS One, 2013; 8:e67188.

26.

Theys

, Vercauteren

, Snoeck

, et al.: HIV-1 subtype is an independent predictor of reverse transcriptase mutation K65R in HIV-1 patients treated with combination antiretroviral therapy including tenofovir. Antimicrob Agents Chemother, 2013; 57:1053–1056.

27.

Brenner

, Oliveira

, Doualla-Bell

, et al.: HIV-1 subtype C viruses rapidly develop K65R resistance to tenofovir in cell culture. AIDS, 2006; 20:F9–F13.

28.

Avila-Rios

, Sued

, Rhee

, et al.: Surveillance of HIV transmitted drug resistance in Latin America and the Caribbean: A systematic review and meta-analysis. PLoS One, 2016; 11:e0158560.

29.

Jiamsakul

, Sungkanuparph

, Law

, et al.; TREAT Asia Studies to Evaluate Resistance–Monitoring Study (TASER-M): HIV multi-drug resistance at first-line antiretroviral failure and subsequent virological response in Asia. J Int AIDS Soc, 2014; 17:19053.

30.

Whitcomb

, Parkin

, Chappey

, et al.: Broad nucleoside reverse-transcriptase inhibitor cross-resistance in human immunodeficiency virus type 1 clinical isolates. J Infect Dis, 2003; 188:992–1000.

31.

Miller

, Margot

, Lu

, et al.: Genotypic and phenotypic predictors of the magnitude of response to tenofovir disoproxil fumarate treatment in antiretroviral-experienced patients. J Infect Dis, 2004; 189:837–846.

32.

Averbuch

, Schapiro

, Lanier

, et al.: Diminished selection for thymidine-analog mutations associated with the presence of M184V in Ethiopian children infected with HIV subtype C receiving lamivudine-containing therapy. Pediatr Infect Dis J, 2006; 25:1049–1056.

33.

Campbell

, Shulman

, Johnson

, et al.: Antiviral activity of lamivudine in salvage therapy for multidrug-resistant HIV-1 infection. Clin Infect Dis, 2005; 41:236–421.

34.

Ford

, Shubber

, Hill

, et al.: Comparative efficacy of lamivudine and emtricitabine: A systematic review and meta-analysis of randomized trials. PLoS One, 2013; 8:e79981.

35.

Oliveira

, Ibanescu

, Pham

, et al.: The M184I/V and K65R nucleoside resistance mutations in HIV-1 prevent the emergence of resistance mutations against dolutegravir. AIDS, 2016; 30:2267–2273.