Role of HPV Genotype, Multiple Infections, and Viral Load on the Risk of High-Grade Cervical Neoplasia

This article is featured in Highlights of This Issue, p. 1769

Cervical cancer is caused by infection from 1 or more of at least 13 high-risk human papillomavirus (hrHPV) genotypes (1, 2). HPV testing provides a more sensitive method of detection for high-grade lesions than cytology (3, 4). Different HPV genotypes have different natural histories, and it has become increasingly important to identify which genotypes are most indicative of an increased risk of developing a high-grade cervical lesion. Advances in HPV-based testing have enabled infections with individual HPV genotypes to be routinely determined, but currently only types 16 and 18 are widely reported and used to guide clinical management (5, 6). Previous reports have indicated differing risks associated with different genotypes (7–9). Other HPV-related factors associated with high-grade disease include viral load (10–13), multiplicity of types (14–16), and methylation status (17–20) as well as cytology and p16^Ink4a which both require intact cellular preparations.

High viral load has been shown to be important for HPV16 (11, 12, 21–24), but more recently, Xi and colleagues (25) reported an association between cervical intraepithelial neoplasia grades 2 and 3 (CIN2/3) and high viral loads for other alpha-9 HPV species. Women with high viral loads have also been found to have more persistent infections, with longer clearance times (26, 27). However, Sherman and colleagues (28) showed that although viral loads were higher in women with CIN diagnoses than women with negative histology, there was no trend correlating viral load and severity of CIN grade. Current evidence about the effect of multiple HPV infections is conflicting; some studies have shown that coinfections increase a woman's risk of cervical precancer and cancer (29), whereas others show no impact (14, 30). Further, it has been shown that HPV-positive women with elevated methylation levels of both human and viral genes have an increased risk of precancerous lesions and cancer (17, 19, 31, 32), but this, and cytology and p16 which require cellular preparations, will not be explored in the study we report on here.

We examined the risk of CIN2+ and CIN3+ associated with different HPV genotypes and assess whether there is additional risk associated with genotype-specific high viral load and multiple HPV infections.

A population-based stratified sample of all women who underwent cervical screening in the state of New Mexico between December 2007 and April 2009 was used for this study. Data were obtained from the New Mexico HPV Pap Registry (NMHPVPR). The NMHPVPR is a statewide public health surveillance program established in 2006 to assess all aspects of cervical cancer preventive care. It includes records of all cervical cytology and HPV tests, and all cervical, vaginal, and vulvar pathology. Laboratories performing cervical cytology, pathology, and HPV tests on individuals residing in New Mexico are required to report all results to the NMHPVPR under NMAC 7.4.3 (http://164.64.110.134/parts/title07/07.004.0003. html). Specimens were collected from selected laboratories under research protocols approved by the University of New Mexico Human Research Review Committee.

Residual material from liquid-based cytology (LBC) samples in 7 in-state laboratories were collected, stratified by age (≤30 years vs. >30 years) and cytology outcome (negative or abnormal). The sampling plan targeted all specimens with abnormal cytology, along with 45% of specimens from women aged ≤30 years with negative cytology, and 8% of specimens with negative cytology from women aged >30 years. A total of 59,644 specimens were included for genotyping. The sample was further restricted to “screening cytology” defined as LBC samples from women with no previous cytology in the past 300 days, and women <15 years or >75 years were excluded. This resulted in a sample of 47,120 women (Supplementary Fig. S1). Although samples were chosen based on proportions of available samples, sampling weights that were applied in this report were based on the first screening sample per woman (Supplementary Table S1). Full details of this cohort have been described previously (33–35). Follow-up was for 3 years after the collection of their screening specimen, and the worst histopathologic diagnosis identified in this period was used as the endpoint.

Broad-spectrum HPV genotyping was performed using the Roche LINEAR ARRAY (LA) HPV GENOTYPING test (Roche Diagnostics) on residual LBC specimens. The Roche LA genotyping test, which has been described in detail previously (34), identifies 37 genotypes, of which 13 are hrHPV types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68), and the remaining 24 are low-risk (lr) types.

Viral load was visually determined as low, intermediate, or high by two independent reviewers for each genotype detected based on intensity of staining on the test strips. Before undertaking this task, readers calibrated their semiquantitative interpretation of viral load on a sample of readings to densitometry values and reference standards. Any discrepant results between the two readers for either HPV genotype or viral load were reviewed by a third independent reader, and the determination of the third reader was final.

Prevalence and positive predictive values (PPV) were calculated for hrHPV types, both overall and for each genotype in single and multiple infections. All analyses were weighted to reflect the number of first screening samples in the statewide population (Supplementary Table S1). Hierarchical rankings of HPV genotypes for CIN2+ and CIN3+ were formed based on sequentially maximizing the PPV for the new genotype when infections which also contained HPV types higher in the hierarchy were omitted. Hierarchies were created both overall and within two age strata (≤30 and >30 years). Cumulative sensitivity and specificity for increasing numbers of genotypes ordered by the hierarchy were plotted as an ROC curve. PPVs were also calculated for each of the 13 hrHPV types stratified by viral load (low, intermediate, high). Viral load was determined separately for each genotype when more than one type was present, and hierarchies were created using genotype only and also using both genotype and viral load. Weighted multiple logistic regression models were also fit to estimate the ORs of CIN2+ and CIN3+ for the joint effects of HPV genotype and viral load.

All analyses were conducted in STATA 13.1.

Cervical screening data and residual LBC samples were used for 47,120 women and estimated to represent 328,427 women in New Mexico who had a screening test between December 2007 and April 2009. The mean age of the weighted population was 40.3 years. Note that 1,893 (0.6%) women had high-grade cytology [high-grade squamous intraepithelial lesion (HSIL), atypical squamous cells—cannot exclude HSIL, adenocarcinoma in situ, or cancer], and a further 19,946 (6.1%) had low-grade cytological abnormalities (low-grade squamous intraepithelial lesion or atypical squamous cells of undetermined significance). We estimated 27.4% of women had at least one hr- or lrHPV infection, and 15.4% had at least one hrHPV infection. Of those with at least one hrHPV infection, 25.1% had multiple hrHPV infections. In women aged ≤30 years, the prevalence of hrHPV infections was 28.1% compared with 9.2% in women aged >30 years (P < 0.001). The prevalence of HPV16 was 3.5% overall, and 1.4% (39.6% of all HPV16 infections) were single infections (Table 1). HPV16 was present in 22.7% of all hrHPV infections. On a population basis, 1.1% of women were diagnosed with CIN2+ and 0.5% with CIN3+. The sensitivity for CIN2+ and CIN3+ in the 3 years after being positive for any hrHPV type was 79.6% and 88.1%, respectively. The specificity for <CIN2 was 85.3%.

HPV genotypes were selected sequentially to maximize the PPV for CIN3+ endpoints and separately for CIN2+ endpoints among women who did not have HPV infections from higher-risk types. This resulted in similar hierarchies for both endpoints. The resulting overall rank orders were HPV16> 33>31>18>35>58>45>52>59>51>39>56>68 for CIN3+ and HPV16>33>31>35>18>58>51>45>39>52>59>56>68 for CIN2+ (Table 2 and Supplementary Table S2). Genotypes 16, 33, and 31 had the highest PPVs for CIN3+ and provided a cumulative sensitivity of 66.5% and specificity of 94.8%. Cumulative sensitivities and specificities for CIN2+ and CIN3+ as the number of HPV types included from the hierarchy were sequentially increased are plotted in Fig. 1. Types 56 and 68 had the lowest PPVs (<0.2% for CIN3+) and should probably not be considered “high-risk” types, but either omitted or called “intermediate-risk” types. HPV positivity was higher for women aged ≤30 years versus those aged >30 years, and the PPV for CIN3+ was also higher in younger women (2.9% vs. 2.1%, P < 0.001). The CIN3+ hierarchy for women aged ≤30 years was similar to the overall rank order, except HPV18 ranked lower (8th vs. 4th place; Supplementary Table S3A and S3B). For woman aged >30 years, HPV18 was ranked third, and HPV35 moved down in the hierarchy, although there were no significant differences in the PPV values. The PPV for CIN2+ for all hrHPV types combined was also greater for women aged ≤30 years compared with women aged >30 years (6.6% vs. 3.7%, P < 0.001).

The PPV for CIN3+ was greatest for HPV16, being 7.0% [95% confidence interval (CI), 6.5–7.5]. The PPV for HPV33 was slightly lower at 4.9% (95% CI, 3.9–6.0) univariately, and when multiple infections with HPV16 were excluded, it was 3.3% (95% CI, 2.4–4.4; Table 2). Similar results were seen for CIN2+ (Supplementary Table S2).

Using the highest ranking HPV type within the hierarchy for each woman when multiple infections were present, weighted logistic regression models were fit. After excluding individuals with multiple infections with types higher in the hierarchy, the odds of having CIN3+ were statistically significant for all hrHPV types except HPV56 and 68 (OR 2.5; 95% CI, 0.6–10.8 and OR 1.7; 95% CI, 0.2–12.9, respectively; Table 3).

Viral load was scored as high, intermediate, or low as detailed in the Materials and Methods. Overall 36.7% of hrHPV infections were scored as high, 26.5% as intermediate, and 36.8% as low. When considering the highest ranked HPV type per women, the odds of CIN3+ were 4.9 (95% CI, 3.4–7.3) times higher for high compared with low levels. The PPV for CIN3+ was above 5% for high-level infections of HPV16 and intermediate-level infections for HPV33, and the PPV was <2% for low-level infections of all genotypes except HPV16 (Table 4). When considering only the highest rank ordered genotype in the hierarchy if there were multiple HPV types detected, there was significant heterogeneity in viral load level for all HPV types (P < 0.001; Table 5). For HPV16, 53.9% of HPV infections were scored as high viral load, whereas only 41.5%, 15.2%, and 14.9% were high for HPV33, 31, and 35, respectively. For HPV16, the odds of a woman having CIN3+ was 5.5 (95% CI, 2.6–11.7) times greater for high versus low viral load (Table 5). After omitting the lower level infections in the hierarchy when there were multiple types, the odds of a woman having CIN3+ was greater for high versus low viral loads for HPV16 and 33. For CIN2+, the odds were significant for the first six HPV types (HPV16, 33, 31, 35, 18, and 58; HPV58 OR 5.5; 95% CI, 1.8–17.4; Supplementary Table S4). However, for most genotypes, the odds of having CIN3+ or CIN2+ was still significantly higher for low-level infections compared with women who were negative (Table 5 and Supplementary Table S4). When allowing for an intensity interaction with age, overall there was no statistical evidence for a difference in risk by viral load level between women aged ≤30 years versus >30 years. However, there was evidence that high viral load of HPV16 increased the risk significantly more for women >30 years (CIN3+ OR 30.7 vs. 86.7, P < 10⁻⁵). Further, there was some suggestion that high viral load of HPV18 and 58 also increased the risk significantly more in older women [hierarchical CIN3+ increase of 5.5 for HPV18 (P = 0.001), and 5.1 for HPV58 (P = 0.004)], but no other types showed a clear age interaction with intensity. Hierarchically, HPV16, 18, and 45 all had over 50% of HPV infections classified as high viral load. There was no difference between low and intermediate viral load by age overall or for any individual HPV type.

PPVs based on a bivariate model for individual HPV genotypes and viral load are shown in Table 4. As anticipated by the logistic regression analyses, there was a trend for increasing PPV with viral load for most HPV types, but especially for HPV16 (CIN2+ PPVs increased from 3.9% to 6.6% to 17.6% for low, intermediate, and high viral loads, respectively; trend χ² = 401.7, P < 10⁻⁷²). The trend in PPVs for CIN2+ with increasing viral load level was significant for all hrHPV types except HPV39 and 68. High viral load infections with HPV16, 33, 31, and 35 posed the greatest risk for CIN2+. However, high viral loads of HPV18 and 58 posed the same risk as intermediate viral loads of HPV16, 33, 31, and 35.

Similarly, for detection of CIN3+, PPVs increased from 2.1% to 3.6% to 10.6% for low, intermediate, and high viral loads of HPV16, respectively (trend χ² = 247.5, P < 10⁻⁵⁶). For CIN3+, the increase was significant for the first six HPV types in the hierarchy (HPV16, 33, 31, 18, 35, and 58), but the number of cases was too limited further down the hierarchy to make reliable inferences.

The inclusion of multiple infections with HPV types lower in the hierarchy added little to the risk prediction for CIN2+ or CIN3+ overall or when restricted to hrHPV types. There was a borderline significant increased risk of CIN3+ for women when adding any other lower-risk HPV types to HPV35 (OR 2.7; 95% CI, 0.9–8.0; P = 0.07), but no other specific type had a significant increase (Table 3).

The ability to identify women at higher risk of CIN2+ and CIN3+ based on both HPV genotype and viral load will be important for individualizing triage plans, particularly when HPV is the primary screening test. The NMHPVPR provides a unique opportunity to investigate the effect of HPV genotyping, multiple HPV infections, and viral load on the risk of high-grade CIN and cancer in a large population-based screening cohort. Use of the Roche LA HPV genotyping test enabled analyses of different genotypes, both in individual and multiple HPV infections, an area where current research has produced conflicting findings. Consistent with previous research, we found HPV16 to be the most prevalent HPV type. Overall 11.4% of women had multiple HPV infections, consistent with findings from Monsonego and colleagues (36) who showed prevalence of multiple HPV infections in a U.S. screening population after using a hierarchical ranking to be 13.4%.

Ranking of HPV types by PPV provided similar hierarchies for CIN2+ and CIN3+, with HPV16, 33, and 31 posing the greatest risk of precancerous disease. Notably, HPV33 and 31 were both ranked above HPV18. Recent research has emphasized the importance of HPV genotypes that are phylogenetically similar to HPV16 (37). The 13 hrHPV genotypes can be clustered into species with more similar DNA sequences. Notably species alpha-9, which includes HPV16, 31, 33, 35, 52, and 58, is most associated with high disease risk, and the top three ranking HPV types we observed were all within the alpha-9 species. Although HPV35 was not common in this population, as there are few African Americans in New Mexico (2.5%), there are populations with high levels of African lineage, thus its prominent position in the hierarchy indicates its importance more broadly. When only considering the top 3 HPV types (HPV16, 33, and 31), their cumulative sensitivity was 66.5% for CIN3+, indicating they are not sensitive enough to be the sole screening test. However, their combined PPV was 5.5% versus 0.2% for the remaining high-risk types, so they can be useful for deciding upon the need for immediate colposcopy versus repeat testing at a 6- or 12-month interval.

Our previous studies that have used hierarchical ranking methods have also found similar rank orders; Cuzick and colleagues (7) reported a ranking based on PPVs for CIN3+ in a referral population with HPV16 and 33 having the highest ranks. In a further sample of HPV-positive women aged ≥30 years, Schiffman and colleagues (38) found the HPV types with the greatest 3-year risk of CIN2+ were HPV16, 52, and 31. However, their study was based on disease prevalence and did not adjust for genotype prevalence and therefore used a different measure than PPV as used here. PPV is particularly important for HPV33. This type has a low prevalence but high PPV, and when present should be managed similarly to HPV16. Several HPV tests offer individual genotyping (7, 38–40), and an HPV hierarchy helps to identify specific genotypes that pose the greatest risk of high-grade CIN, and thus can assist in improving the triage process for clinical management of HPV-positive women. Currently, HPV tests approved by the FDA only offer individual genotyping for HPV16, 18, or 45, but findings from our data show the importance of HPV31 and 33 as high-risk genotypes, and the value of downgrading HPV types 39, 51, and 59 to “intermediate risk” types, although HPV51 was considerably higher in the CIN2+ hierarchy. This emphasizes the need for more complete hrHPV genotyping assays if the principle of equal management for equal risk is to be applied. Of note, HPV52 did not exhibit substantial risk here, but has been seen to do so in other populations (41). When using the LA HPV Genotyping test, HPV52 is only inferred if coinfections with HPV33, 35, or 58 are not detected, so it will be underestimated in our study. This effect will be small, and under the assumption that the prevalence of these types is independent, HPV52 prevalence would only increase from 1.92% to an estimated 1.96% (Table 1). We did not supplement the HPV genotyping in this large population-based evaluation using HPV52 type-specific PCR, although this could be an area of future research as suggested by others (42, 43).

In our study, coinfections with HPV types lower in the hierarchy did not significantly increase the risk of CIN2+ or CIN3+ beyond that for the highest-risk type found. Similar findings have previously been reported. Schmitt and colleagues (15) found that the occurrence of multiple HPV infections did not affect the risk of a lesion being high- or low-grade, and Wentzensen and colleagues (16) found no association between disease status and the number of genotypes detected in a woman. Previous studies showing increased risk of CIN with multiple HPV infections had few CIN2+ cases, and were restricted to younger women (44), a subgroup known to harbor a larger number of HPV infections (14).

HPV16 is the genotype with the highest PPV for high-grade precursor lesions, especially when the viral load is high. Its PPV of 17.6% overall is well above the 10% threshold for CIN2+ (PPV 10.6% for CIN3+) suggested for determining immediate referral for colposcopy (5, 45, 46). However, much of this previous research has not found an association between viral load and CIN2+ for other genotypes, possibly due to small sample sizes. In our study, after adjusting for multiple infections with types higher in the hierarchy, we found the risk for CIN2+ was above 10% only for high viral load infections for HPV types 16, 33, 31, or 35 and above 5% only for high- or intermediate-level infections with these types or high-level infections with HPV18 or 58 (Table 4). Our finding that viral load was relevant for HPV18 is not in agreement with previous studies which found quantification of HPV18 had little predictive power and thus the clinical utility of this finding requires confirmation (41, 47). Notably, even a high viral load of HPV18 only carried a CIN2+ PPV of 7.9% (PPV 2.5% for CIN3+), which is still below the conventional 10% threshold for immediate colposcopy.

Noticeably, the top ranked HPV types for CIN risk are the seven high-risk types in the nonavalent vaccine (HPV16, 18, 31, 33, 45, 52, and 58) and HPV35. However, a key finding from this article is that the risk can be substantially modified by viral load and thus a full-risk stratification policy needs to include both HPV genotype and viral load. High levels of HPV types lower in the hierarchy (e.g., HPV18 and 58) pose a similar risk of disease to intermediate levels of higher-risk types (HPV16, 33, 31, and 35). The importance of viral load compared with HPV genotype has not been widely appreciated, but when considered together can improve assessments of the risk of a high-grade CIN lesion.

A full triage strategy will require consideration of other measures beyond the scope of this investigation, including cytology and potentially p16 status and HPV methylation in addition to HPV genotype and viral load information to guide management. Of note as well are HPV18 and 45 which are not strong predictors of CIN2+ or CIN3+, but are more common in cancer. They are associated with endocervical cancers, and their precursor lesions are not so easily seen at colposcopy. If HPV18 and 45 infections remain persistent, more careful exploration of the endocervical canal by new methods may be needed, especially in older women (>30 years) where HPV18 poses a greater risk. Although these results were not used for clinical management, if age categories were modified to put 30 year olds in the older age group to align with current screening recommendations, the conclusions from this study remain unchanged.

Further work on viral load is warranted before it can be used routinely, especially to standardize viral load measurements. Although semiquantitative estimates of viral load were important in our analysis, the biology behind viral load is complex. Low-grade lesions with koilocytes can have thousands of copies of HPV per cell and one cell can contribute more DNA than a hundred CIN3 cells with 10 copies per cell. Thus, viral load is a complex correlate of the interplay of grade, lesion size, sampling of lesions, etc. In addition, none of the platforms currently approved by the FDA allow for routine reporting of viral load.

Little variation was seen in the hierarchy for different genotypes when the PPVs were based on both genotype and viral load (Supplementary Table S5) compared with using only genotype (Table 2). In particular, the top ranking HPV types (HPV16, 33, and 31) did not change. However, viral load was important, e.g., low viral loads of HPV16, 31, and 33 were lower in ranking than high levels of HPV18, 35, 52, and 58. This emphasizes the management benefits which could be gained if HPV genotyping and viral load were both used.

One of the strengths of this study was the access to a large population-based stratified screening sample. However, the study had some limitations; histology outcomes were only available for women who were referred to colposcopy because of cytological abnormalities. Thus, CIN2+ lesions arising from infections not producing a detectable cytological abnormality would have been missed. CIN determination was based on routine clinical practice, and although all cases diagnosed within 3 years were included, a closeout visit at 3 years, as would be common for a clinical trial, would result in higher disease detection rates. Analyses of viral load were based on a visual three-level cutoff criteria—high, intermediate, and low— and further work is needed to determine if its assessment would benefit from more precise quantitation. In addition, different HPV tests assess viral load in different ways, e.g., many PCR-based methods use cycle threshold values, Hybrid Capture 2 uses signal-amplified luminescence levels (relative light units) to establish cutoff values, and our method based on LA genotyping uses visual assessment based on colorimetric precipitate observed as lines or “bands” on a solid-phase matrix. However, semiquantitation via the LA genotyping test has been supported by its correlation with a gold standard of quantitative PCR (48). As noted above, the importance of viral load appears to be genotype-specific, so an overall combined result for any hrHPV type may be less informative than the type-specific viral load, as provided here.

To conclude, in a large population-based stratified screening sample of women, we found the risk of high-grade CIN was dependent on both the HPV genotype and viral load, with no added risk associated with coinfections from other HPV types lower in the hierarchy. Algorithms based on both HPV genotype and viral load in combination show promise for refining clinical management of hrHPV-positive women, and reducing the number of women who are currently recommended to have immediate colposcopy.

J. Cuzick reports receiving research grants from Qiagen, Becton Dickinson, Genera Biosystems, Hologic, Gene First, and Trovagene and is a consultant/advisory board member for Qiagen, Becton Dickinson, and Genera Biosystems. J. Cuzick and C.M. Wheeler have received funds from grants, cooperative agreements, or subcontracts related to cervical screening and triage through their institutions. C.M. Wheeler reports receiving reagents and equipment from Roche for this work. Outside of the work reported here, C.M. Wheeler reports a contract with Genera Biosystems through her institution and personal fees from Becton Dickinson.

Conception and design: J. Cuzick, C.M. Wheeler

Development of methodology: J. Cuzick, C.M. Wheeler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.C. Hunt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Adcock, J. Cuzick, C.M. Wheeler

Writing, review, and/or revision of the manuscript: R. Adcock, J. Cuzick, W.C. Hunt, C.M. Wheeler

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.M. McDonald

Study supervision: C.M. Wheeler

The authors thank George Montoya, Amanda Pearse, and Erika Ingersoll, who performed the HPV genotyping for this study using the Roche Linear Array Genotyping Assay, and Michael Robertson and Cathy Sherman, who created the initial informatics systems to enable the overall research evaluations of the New Mexico HPV Pap Registry (NMHPVPR).

The work was funded by a cooperative agreement awarded by the NIH to C.M. Wheeler (U19AI113187). Roche Molecular Systems provided HPV genotyping reagents and equipment.

The sponsors had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; or approval of the article. R. Adcock, J. Cuzick, W.C. Hunt, R.M. McDonald, and C.M. Wheeler had full access to all the data in the study. All authors reviewed and approved the final article and had final responsibility for the decision to submit for publication.

Members of the NMHPVPR Steering Committee reviewed and gave input to the manuscript and supported the concept and directions of the NMHPVPR including the evaluations presented in this manuscript. No compensation was received for contributions to this article by any named authors or by the NMHPVPR Steering Committee members. The NMHPVPR Steering Committee members participating in this effort are as follows: Nancy E. Joste, MD, University of New Mexico Health Sciences Center and Tricore Reference Laboratories; Walter Kinney, MD, retired, Kaiser Permanente Northern California; Cosette M. Wheeler, PhD, University of New Mexico Health Sciences Center; William C. Hunt, MS, University of New Mexico Health Sciences Center; Ruth M. McDonald, MS, University of New Mexico Health Sciences Center; Michael Robertson, BS, University of New Mexico Health Sciences Center; Alan Waxman, MD, MPH, University of New Mexico Health Sciences Center; Steven Jenison, MD, community member; Julia C. Gage, PhD, MPH, U.S. National Cancer Institute; Philip E. Castle, PhD, MPH, Albert Einstein School of Medicine; Vicki Benard, PhD, U.S. Centers for Disease Control and Prevention; Debbie Saslow, PhD, American Cancer Society; Jane J. Kim, PhD, Harvard TH Chan School of Public Health; Mark H. Stoler, MD, University of Virginia; Jack Cuzick, PhD, Wolfson Institute of Preventive Medicine; Giovanna Rossi Pressley, MSc, Collective Action Strategies; and Kevin English, DrPh, MPH, Albuquerque Area Southwest Tribal Epidemiology Center (AASTEC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.