*2.1. Studies*

Overall, 1763 protease sequences from 1497 patients reported in 30 studies who received either boosted or unboosted atazanavir as their first PI were available for the analysis (Table 1). These sequences included 773 sequences from 740 patients in 27 studies from Stanford HIV Drug Resistance Database (HIVDB) [13], and previously unpublished sequences, including (i) 741 sequences from 562 patients from the EuResist Integrated Database (EIDB) [14]; (ii) 206 sequences from 152 patients from the Stanford University Hospital (SUH); and (iii) 43 sequences from 43 patients from the RHIVDB [15], a freely accessible database of HIV-1 sequences and clinical data of infected patients. Of the 184 patients with more than 1 sequence, 17 had sequences that differed from one another by one or more DRMs. For these patients, we selected the sequence containing the largest number of PI-associated DRMs. The complete set of 1497 one-per-person HIV-1 group M sequences from persons receiving atazanavir was provided in Table S1.




**Table 1.** *Cont.*


**Table1.***Cont.*

Footnotes: 1 DRMs were defined as those with a Stanford HIV drug resistance program penalty score for ≥1 PI. 2 Subtypes with ≥10% sequences were listed. 3 Colonno04 contained sequences from previously PI-naïve patients with virological failure with resistance on ATV-containing regimens in three clinical trial, AI424-007/041, AI424- 008/044, and AI424-034. Additional notes: All studies used the Sanger dideoxynucleoside sequencing method, except for Alves19 in which next-generation sequencing was used; samples from peripheral blood mononuclear cells (PBMCs) were used in Alves19, Makwaga20, and Mziray20, and from both PBMC and plasma in Kim13. In the remaining studies, plasma was used. Abbreviation: b-ATV—boosted atazanavir.

The 30 studies were published between 2004 and 2021. The median number of patients per study was 12 (IQR: 1–39). The distribution of studies and patients by region included Africa (10 studies; 119 patients), North America (5 studies; 383 patients), Europe (4 studies; 565 patients), Latin America (4 studies; 199 patients), Eastern Europe (3 studies; 55 patients), and Asia (2 studies; 38 patients). Two studies included 138 patients from 1 or more regions.

The median sample year was 2011 (IQR: 2007–2015). Approximately 99% of sequences were obtained from plasma and 1% from peripheral blood mononuclear cells (PBMCs). Next-generation sequencing (NGS) was performed in 1 of the 30 studies. The most common subtypes were B (61.9%), C (13.6%), A (6.7%), G (6.1%), 02\_AG (4.9%), F (3.1%) and D (1.1%). Of 1497 patients, 62.7% (n = 939) received boosted atazanavir and 37.3% (n = 558) received unboosted atazanavir. A higher proportion of patients with subtype B (70.4% of 927) compared with non-subtype B (50.2% of 570) viruses received boosted (*p* < 0.001). Table 2 summarizes the numbers of patients according to the administration of atazanavir (boosted vs. unboosted), subtype (B vs. non-subtype B), previous antiretroviral therapy (ART) (naïve vs. experienced), and year of ART initiation.

**Table 2.** Proportion of patients with PI-associated drug resistance mutations (DRMs) and median number of DRMs per patient according to ART history and HIV-1 subtype.


Footnotes: 1 DRMs were defined as those with an HIVDB drug resistance program penalty score for ≥1 PI. 2 Patients with available year of ART initiation (n = 1127) were grouped into four time periods containing approximately equal numbers of patients.

## *2.2. Mutation Prevalence*

Of the 1497 patients, 264 (17.6%) had 1 or more PI-associated DRMs. Of the 57 HIVDB PI-associated DRMs, 48 occurred in ≥1 patient, 38 in ≥2 patients, and 24 in ≥5 patients. The most commonly occurring major DRMs were I50L (34.1%), M46I (32.6%), V82A (22.3%), L90M (19.3%), I54V (16.3%), N88S (10.2%), M46L (7.6%), V32I (6.4%), and I84V (6.1%) (Table 3). The most common accessory DRMs were L33F (20.8%), Q58E (15.9%), K20T (14.4%), G73S (11.7%), L10F (9.8%), F53L (9.8%), K43T (8.7%), and L24I (6.1%).

**Table 3.** Drug resistance mutations (DRMs) occurring in ≥1 sequences from patients receiving boosted or unboosted atazanavir as their first PI.



#### **Table 3.** *Cont.*

1 DRMs were defined as those with a Stanford HIVDB drug resistance program penalty score for ≥1 PI. 2 See the method for DRM classification.

Of the 264 sequences with 1 or more PI-associated DRMs, the proportions of the sequences containing 1 DRM, 2–3 DRMs and ≥4 DRMs were 33.7%, 31.4% and 34.9%, respectively. The distribution of DRMs differed according to the total number of DRMs per sequence (Figure 1). Among sequences with a single DRM, the most common major DRMs were I50L, M46I/L, L90M, and N88S, while the most common accessory DRMs were Q58E, K20T, G73S, and L33F. In contrast, among sequences with ≥4 DRMs, the most common major DRMs were M46I/L, V82A, L90M, I54V, I50L, and N88S, while the most common accessory DRMs were unchanged. The major DRMs V32I and I84V occurred in approximately 5% to 6% of sequences regardless of the total number of DRMs.

An additional 197 mutations, previously classified as nonpolymorphic treatment selected mutations (NP-TSMs), occurred in 149 sequences, including in 109 of the 264 sequences containing a PI-associated DRM and 40 of the 1215 sequences without a PIassociated DRM. There were 33 different NP-TSMs of which the most common were L89T (34.9% of 149 sequences), K55R (15.4%), I85V (11.4%), A71I (9.4%), and E34Q (7.4%) (Table S2). These mutations were not classified as DRMs because they do not receive an HIVDB mutation penalty score.

#### *2.3. Unboosted versus Boosted Atazanavir*

PI-associated DRMs occurred in 21.0% of 558 patients receiving unboosted atazanavir and 15.7% of 939 patients receiving boosted atazanavir (*p* = 0.01) (Table 2). However, among patients with a DRM, the median number of DRMs was not significantly greater in those receiving unboosted atazanavir (3 DRMs; IQR: 1–4) compared with boosted atazanavir (2 DRMs; IQR: 1–4; *p* = 0.1). Of the 48 reported DRMs, I50L was the only DRM that occurred more commonly in patients receiving unboosted as compared with boosted atazanavir (10.4% vs. 3.4%; adjusted *p* < 0.001).

Sequences from patients receiving unboosted atazanavir were also slightly more likely to have one or more NP-TSMs compared with sequences from patients receiving boosted atazanavir (12.2% of 558 vs. 8.6% of 939; *p* = 0.03). Each of the 33 reported NP-TSMs occurred in similar proportions in patients receiving unboosted and boosted atazanavir.

**Figure 1.** Prevalence of PI-associated drug-resistance mutations (DRMs) in 264 sequences containing 1 or more DRMs from previously PI-naïve patients receiving a boosted or unboosted atazanavircontaining regimen. The distribution of DRMs is plotted separately according to the number of PI-associated DRMs in the sequence: ( **A**) 1 DRM, (**B**) 2 to 3 DRMs, and ( **C**) ≥4 DRMs. The DRMs shown are those occurring in ≥5% of the sequences, including 9 major DRMs indicated in red and 8 accessory DRMs indicated in yellow.
