The HIV/AIDS surveillance systems of the New York State Department of Health (NYSDOH) and the New York City Department of Health and Mental Hygiene (NYC DOHMH) have been described previously [13]–[15]. Nucleotide sequences from HIV genotypes, along with other HIV-related tests and conditions, are reportable by law [7]–[12]. Laboratory and provider reports are transmitted to NYSDOH where they are matched to the New York State HIV registry; data relating to cases within New York City are forwarded to NYC DOHMH where they are matched to the NYC HIV registry. Incoming data at the state or city level that do not match an existing registry record initiate a field investigation to confirm the case, date and disposition of diagnosis and collect other data required by surveillance. An analysis dataset was created based on diagnoses and laboratory results dated January 1, 2006, through December 31, 2008, reported by April 30, 2010, and added to the NYS HIV registry as of May 31, 2010. A total of 14,046 persons aged 13 and older and not perinatally infected had an initial diagnosis date between January 1, 2006, and December 31, 2008; 937 (6.7%) were excluded because of missing or discrepant data on date of initial diagnosis or genotype, leaving 13,109 for analysis.

Diagnosis refers to a new diagnosis of HIV with or without a concurrent diagnosis of AIDS. Concurrent diagnosis was defined as AIDS diagnosis within 31 days of initial diagnosis of HIV. Region at diagnosis was categorized as New York City or New York State excluding New York City. Poverty area was defined as residence at diagnosis in a ZIP code tabulation area in which at least 20% of residents per US Census 2000 met the federal definition of poverty. Poverty area was not calculated for homeless or sheltered persons or for persons residing in zip codes created after 2000. Cases with missing risk factor were assigned to the category, “no identified risk.” Initial resistance test was defined as the first HIV genotype (if any) within 3 months of diagnosis. The 3 month interval was chosen to limit the number of persons that may have started ART before resistance testing and to allow comparison with results from the Centers for Disease Control's (CDC) Variant, Atypical, and Resistant HIV Surveillance (VARHS) system [16]–[17]. In addition to nucleotide sequences, laboratory data included the first CD4 count and viral load drawn within 3 months of diagnosis. Persons with a viral load, CD4 count or resistance test within 3 months of diagnosis were considered in care because these tests must be ordered by a physician [14]. CD4 counts were dichotomized as >350 cells/ml or <350 cells/ml because Department of Health and Human Services (DHHS) guidelines in place during the reporting period recommended initiation of ART at this threshold [18]. Viral loads were grouped into three intervals, <10,000 copies/ml, 10,000–100,000 copies/ml, and >100,000 copies/ml.

HIV genotype testing was performed by commercial laboratories using various test kits, including Viroseq™, GenoSure™, TRUGENE™ and in-house kits. Only protease and reverse transcriptase sequences of the pol gene were reported. Nucleotide sequences were analyzed using the Resistance Analysis System (RAS), version 2.0 (Frontier Science & Technology Research Foundation, Amherst, New York), a program built specifically to facilitate NYS resistance surveillance. Mutations were ascertained by a comparison of aligned sequences with the Los Alamos National Laboratory subtype B consensus sequence [19]. ARV-specific predicted resistance was calculated using code developed by Frontier Science and scores from the Stanford HIVDB algorithm, version 6.0.9; [20], [21] this algorithm was also used to determine HIV-1 subtype. Sequences that did not meet the minimum processing requirement of the HIVDB algorithm could not be analyzed [20].

Transmitted drug resistance was defined as the presence of 1 or more mutations in the surveillance drug resistance mutation list (SDRM) [22]. ARVs were categorized by class. Single, double or triple class resistance was defined as 1 or more surveillance drug mutation within one, two or three antiretroviral drug classes respectively. Predicted resistance to specific antiretroviral drugs was defined as sequences with a score of ≥4 on the Stanford HIVDB 5-point resistance scale [20].

Multivariate logistic regression was used to assess the likelihood of an initial resistance test and the likelihood of TDR as a function of demographic and clinical characteristics. Unadjusted and adjusted odd ratios with 95% confidence intervals (CIs) were calculated. Concurrent diagnosis of HIV/AIDS was excluded from the regression analysis of testing patterns because it is partially defined by CD4 count. Variables significant (p<0.05) on bivariate analysis were entered into multivariate logistic regression models for the two outcome variables, testing and TDR. Confidence limits for proportions were calculated using exact CIs for the binomial proportion. Trends were examined using the Cochran-Armitage test and are reported with two-sided p-values. All statistical tests were performed using SAS, version 9.1.3 (SAS Institute, Cary, North Carolina).

Of the 13,109 persons included in the analysis, 4,155 (31.7%) received their first resistance test within 3 months of diagnosis (“initial resistance test”); 1,311 (10.0%) were first genotyped >3–12 months after diagnosis, 7,643 (13.3%) were first genotyped >12 months after diagnosis, and 44.9% were never genotyped. Of all persons ever genotyped, three-quarters were genotyped within three months of initial diagnosis. Patients never genotyped differed significantly from patients ever genotyped by age, race, risk factor, and disease stage at diagnosis (data not shown). Patients with CD4<350, VL>100,000 and concurrent HIV/AIDS at diagnosis, i.e., patients meeting DHHS guidelines for ART, were more likely to have ever been genotyped.

Initial resistance testing among newly diagnosed persons differed significantly by sex, race/ethnicity, age, risk factor, region of diagnosis, poverty area, year of diagnosis, and disease stage at diagnosis (Table 1). Among persons with new diagnoses, 9,785 (74.6%) showed evidence of care (i.e., saw a physician) within 3 months of diagnosis. Of persons in care, 4,155 (43%) had initial resistance tests. Among all newly diagnosed, the proportion with an initial resistance test increased from 25% in 2006 to 38% in 2008 (p for trend <0.0001). Subsequent analyses were conducted among persons in care within three months (N = 9,785 or 74.6% of the total number of newly diagnosed) because these would be the only persons in the database who would have had the opportunity for initial resistance testing.

In the multivariate analysis of initial resistance testing among newly diagnosed persons in care, blacks and Hispanics were less likely to be tested than whites (AOR 0.70 (0.61–0.79), AOR 0.85 (0.74–0.97)) (Table 1). Persons aged 13–24 or 40–59 at diagnosis were slightly less likely to be tested than those 25–39 (AOR 0.77 (0.67–0.89); AOR 0.87 (0.79–0.97)), while persons 60 and older were no more likely to be tested (AOR 0.96 (0.79–1.19)). Compared with men who have sex with men (including men who have sex with men and use injection drugs (MSM + MSM/IDU)), persons with heterosexual transmission risk were less likely to be tested (AOR 0.77 (0.66–0.90)).

Persons diagnosed in New York State excluding New York City were more likely to have a resistance test than persons diagnosed in New York City (AOR 1.25 (1.11–1.40)), as were persons diagnosed in 2008 in comparison to those diagnosed in 2006 (AOR 1.85 (1.66–2.06)). Persons living in a non-poverty area were not significantly more likely to have a resistance test than those living in a poverty area (AOR 0.93 (0.85–1.03)). Persons with initial CD4 count ≥350 cells/ml were less likely to have a resistance test than persons with CD4 count <350 cells/ml (AOR 0.57 (0.51–0.62)), and persons with viral loads of <10,000 copies/mL or 10,000–100,000 copies/mL were less likely to have a resistance test than persons with >100,000 copies/mL (AOR 0.14 (0.10–0.19), AOR 0.85 (0.77–0.95)).

Of the 4,155 initial resistance tests, 123 were reported with partial nucleotide sequences, and 4,032 (97.0%) had analyzable sequences. Among these, 450 (11.2% (10.2%–12.1%)) had evidence of TDR (Table 2). TDR did not significantly increase over time (p for trend = 0.204). In the multivariate analysis, risk, year of diagnosis and viral load remained significantly associated with TDR., However persons with a heterosexual risk factor were no more likely to have resistance than MSM (AOR 1.03 (0.79–1.35)). In addition, black and Hispanic MSM were no more likely to have TDR in comparison with other race and risk groups (OR 1.10 (0.82–1.49) data not shown). Persons with no identified risk (NIR) were significantly less likely to have TDR than MSM (AOR 0.68 (0.53–0.87)).

TDR varied by drug class. Resistance was highest to NNRTIs (6.3% (5.5%–7.0%)) and was significantly higher than resistance to NRTIs (4.3% (3.6%–4.9%)) and PIs (2.9% (2.4%–3.4%)) (Table 3). Over time, there was no significant increase in resistance in the NNRTI, NRTI or PI classes (p = 0.144, p = 0.686, p = 0.851, respectively). Resistance in two classes was highest in the NRTI-NNRTI (0.8%) combination; resistance in three classes was minimal (<0.5%). The most frequently observed polymorphisms by drug class are shown in Table 4.

We also examined predicted resistance to 1 or more components of selected starting regimens recommended by DHHS [18]. Resistance to 1 or more ARVs within the NNRTI-inclusive regimen EFV+TDF+FTC was observed in 7.1% (6.3%–7.9%) of cases (data not shown). In contrast, 1.4% (1.0%–1.8%) were resistant to 1 or more drugs in the PI-inclusive regimen LPV/r+TDF+FTC. Among individual NRTIs, resistance was highest to AZT and D4T (1.8% (1.4%–2.2%) and 1.6% (1.2%–2.0%)) and lower to TDF (0.3% (0.1%–0.5%) and FTC (0.9% (0.6%–1.2%), the two agents with recently demonstrated efficacy in pre-exposure prophylaxis [23]. 3TC, which in the treatment setting can be used interchangeably with FTC, showed similarly low resistance (0.9% (0.6%–1.2%)).

Most analyzed sequences (92.8%) were subtype B; 118 (2.9%) were CRF02_AG; and 83 (2.1%) were subtype C. Persons residing in NYC at diagnosis were no more likely to have non-B subtypes than persons residing in New York State excluding New York City (7.7% vs. 6.0%, p = 0.0611).

Within the U. S., this analysis represents the first use of routinely reported surveillance data to estimate TDR and to describe resistance testing patterns as well as the largest number of sequences used for resistance surveillance to date [16], [24]. More than half of newly diagnosed persons who entered care within three months did not receive an initial resistance test per DHHS guidelines, although the proportion receiving initial resistance tests increased between 2006 and 2008. The observed increase is consistent with the adoption of the 2007 DHHS guidelines recommending resistance testing for all newly diagnosed persons [18], [25]. Previous guidelines recommended resistance testing for acute infection and patients initiating or failing ART [23]. Significant differences in resistance testing by demographic characteristics, including race, age and transmission risk, are concordant with literature on initiation, source, and utilization of care [26]–[28]. Potential candidates for initiation of ART per DHHS guidelines (CD4<350) were more likely to be tested than others, likely reflecting the decision by some providers to postpone resistance testing until initiation of ART. Similarly, persons with low viral loads (<10,000) were less likely to be tested. While this could be evidence of the impact of viral load on a provider's decision to genotype, it may also be affected by the failure of amplification and genotyping at low viral loads (failed genotypes are not reported). Resistance testing was less common among NYC residents than residents in the rest of the state, an unexpected finding given the concentration of training hospitals and designated AIDS centers in the city. Further analysis is needed to elucidate the relationship between resistance testing, provider type, and utilization of care.

The prevalence of TDR among persons with new diagnoses in NYS in 2006–2008 was 11.2% (10.2%–12.1%). There was no significant change in TDR over time. Worldwide estimates of TDR range from 8%–24%, though comparison between these results is difficult due to differences in the mutations used to define TDR [29]–[38]. Our estimate, based on the SDRM list [22], is higher than the national prevalence estimate (8.3%) for the time period 1997–2001 [39] but is substantially lower than a previous report of resistance in a NYC sample of MSM in 2003–2004 (24.1%) [3]. Both of these studies used modified IAS-USA mutation lists. Wheeler et al. estimated the national prevalence of transmitted drug resistant mutations (TDRM) in 2006 to be 14.6% using a modified SDRM list [16]. We estimated the New York State TDR to be 24.2% using the same mutation list (results not shown). Further analysis is needed to test the utility of the SDRM and TDRM lists in the U.S. epidemic.

In contrast to previous findings of increasing TDR and high levels of TDR among MSM, we found stable resistance evenly distributed between MSM and heterosexual risk groups [3], [40]. Better risk factor ascertainment would allow us to measure the TDR by risk factor more accurately and/or to understand the unexpected findings of this analysis. Our data show that 1 in 9 persons newly diagnosed with HIV in NYS has TDR and 1 in 50 is predicted to have a suboptimal response to a standard ART regimen. Key populations considered to be on the leading edge of the epidemic, e.g., young black and Hispanic MSM, showed no more TDR than others. Ongoing surveillance will confirm the significance and durability of these observations.

Our analysis has important limitations. HIV surveillance data contain limited person-level information; duration of infection and ART history are not available. Newly-diagnosed persons are assumed to be ARV-naïve but may not be. Despite the CDC-sponsored routine interstate duplication review (RIDR) and comprehensive field investigation, persons may be incorrectly identified as newly diagnosed because there is incomplete date information or because they were diagnosed out of state and subsequently received HIV care in New York State. In such cases acquired resistance may be incorrectly classified as TDR. The number of resistance tests reported is an underestimate of the number ordered by providers because resistance tests in which viral RNA amplification fails are not reportable.

Integrating resistance data into the existing surveillance system was logistically and technically challenging. Laboratories certified by NYSDOH to perform resistance testing were required to report nucleotide sequences beginning on June 1, 2005. However, laboratories acquired full capacity to report resistance data at different times after the regulations were enacted, which meant that much of the data was reported retrospectively. Laboratories were required to resubmit when incomplete data were identified; however, some laboratories were not able to do so. Completeness of laboratory reporting was estimated by comparing self-reported laboratory testing logs to received data transmissions. Completeness was estimated to be 82% in 2006, 89% in 2007, and 98% in 2008. Adjusting for completeness, the proportion of persons with new diagnoses with initial resistance tests increased from 29% in 2006 to 39% in 2008 (p<.0001) (Figure 1). Incomplete data in key fields (e.g. name and date of birth) affected the matching of some reports to the surveillance registry. However, the proportion of resistance tests that could be matched was similar to other reportable tests.

The completeness and accuracy of risk ascertainment is an ongoing challenge for surveillance. Misclassification of heterosexual transmission as NIR and misclassification of MSM as heterosexual may account for our observations of reduced risk associated with NIR and equivalent risk in MSM and heterosexuals [41].

Our TDR estimate may be biased because it is based on genotypes for less than one-third of persons with new diagnoses and less than one-half of those initiating care within 3 months. It is possible that TDR in persons not genotyped is significantly different from the TDR patterns reported here. In addition, our estimate may under-represent clinically important resistance; minority quasispecies, not detectable by genotypes reported to NYSDOH, have been shown to be prevalent in untreated persons and to reduce treatment efficacy [42]. Finally, current reporting does yet not allow the monitoring of resistance certain classes of ARVs including integrase strand-transfer inhibitors, entry inhibitors or CCR5 receptor antagonists.

Using the New York State HIV resistance surveillance system, we have taken an important step in addressing transmitted drug resistance as a public health concern. In contrast to earlier local reports, our data suggest that TDR is not increasing and that multiclass TDR is not prevalent. Furthermore, TDR is not isolated to a specific subgroup, and common starting regimens are still effective for most new diagnoses in New York State. This information will help shape our response to the epidemic in both the public health and medical communities.

This analysis suggests that continuing routine resistance surveillance is appropriate for three reasons. First, more data are needed to verify the trend in TDR. Resistance surveillance systems such as the one describe here are uniquely qualified to provide consistent, long-term monitoring. Methodological differences between short-term studies make it difficult to evaluate trends in TDR. Second, treatment-intensive community strategies such as ‘Test and Treat’ and PrEP may increase TDR. Third, in contrast to surveillance based on specimen salvage, which is costly and logistically difficult, resistance surveillance through routine electronic reporting is relatively low cost and scalable. If improved TDR estimation is found to be necessary, routine reporting could be supplemented with specimen salvage from new diagnoses without routine genotype results.

This work illustrates the power of surveillance to establish baselines and monitor progress toward goals established to achieve epidemic mitigation and control [43]. However, broader provider uptake of genotype testing is needed to better estimate population TDR and to understand the TDR prevalence at which routine genotyping and surveillance of new diagnoses provide clinically and epidemiologically significant information.