Dyeing to Know

Abstract

This issue contains the report of a study in which vaginal applicators containing either gel or powder were stained to detect evidence of vaginal insertion. The study showed good results for both types of products with 100% sensitivity and specificity among known positive and negative controls and at least 86% positivity on applicators returned by study participants and reported as having been inserted. This report follows reports of similar studies starting in 2004 in which either low-density polyethylene Microlax-type applicators or polypropylene applicators from HTI Plastics (Lincoln, NE) designed to dispense microbicide gel or placebo gel were stained using either trypan blue or FD&C Blue No. 1. The goal of these studies is to find a method to objectively measure adherence to product use that is more accurate than subject report. Several recent studies demonstrate the importance of adherence to product use in microbicide trials for HIV prevention. In the Centre for the AIDS Programme of Research in South Africa (CAPRISA) 004 trial of pericoitally applied vaginal tenofovir (TFV) gel, TFV gel users were, overall, 39% less likely to acquire HIV than placebo users. However, among high adherers, defined as those who reported using the gel in at least 80% of coital acts, TFV gel users were 54% less likely to acquire HIV than placebo users.1 In the Microbicide Trials Network study entitled Vaginal and Oral Interventions to Control the Epidemic, the once-daily vaginal TFV gel arm was stopped early because there was no difference in HIV incidence between the TFV gel and placebo gel arms. It is not yet known why this study did not show efficacy of TFV gel; possible reasons include the drug regimen which was different from that of CAPRISA (daily vs. pericoital), unreported risky behaviors such as anal sex, and also nonadherence. Finally, the Methods for Improving Reproductive Health in Africa (MIRA) trial found that HIV infection rates were the same in the control arm (in which women received condoms only) and the intervention arm (condom plus diaphragm plus lubricant).2 Condom use was, however, lower in the intervention arm, and it was speculated that diaphragm use may have compensated for this difference, but that “uncertainty about self-reported condom use makes other explanations equally plausible.” Self-report is known to be an inaccurate measure of adherence, with participants almost certainly reporting better adherence than what they actually achieved.3,4 The result of this in an HIV prevention trial is that actual product use is less than that reported, which can lead to an underestimate of product effectiveness and an overestimate of product safety. “Triangulation” or combining case report form interview data with coital diaries and in-depth interviews is more informative than using a single method, but still relies on self-report.5 Other commonly used methods of assessing adherence also rely on self-reported data (e.g., computer-assisted interviews rather than face-to-face interviews)6 or are not feasible in prevention studies of vaginally applied products (e.g., directly observed therapy). Thus, if a simple inexpensive method of staining applicators were found, it could be valuable to the field, although with certain limitations. Eight previous publications on vaginal applicator staining were reviewed and are summarized in the Table 1.7–14 It is important to consider the results of testing known positive and negative controls, from which sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy can be calculated, separately from the results of testing applicators that were reportedly used by participants. Not all of the studies reported sensitivity and specificity, but from the ones that did, trypan blue appears to give better results than FD&C Blue No.1, with all figures for sensitivity, specificity, positive predictive value, negative predictive value, and accuracy >96%. The Hemmerling article in this issue is valuable because a new type of plastic was stained with trypan blue, and both positive and negative controls were used. Its acknowledged weakness is that the single reviewer was not masked. Unfortunately, trypan blue is a possible human carcinogen requiring skin, eye, and respiratory protection when used, even at a concentration of 0.4% (available at: http://www.sigmaaldrich.com/catalog/product/σ/t8154?lang = en&region = US). Over half a million applicators were stained (with FD&C Blue No. 1 dye) in the Population Council Carraguard trial;10 using trypan blue in such a situation does not seem manageable or cost-effective.Table 1: Results of Studies in Which Vaginal Applicators Were Stained Using Trypan Blue or FD&C Blue No. 1The safer FD&C Blue No. 1 dye, when used with Microlax applicators, gives sensitivities and specificities of at least 85%. However, less promising results were seen when the FD&C Blue No. 1 dye was used with HTI applicators, such as those being used in the Follow-on African Consortium for Tenofovir Studies (FACTS) 001 study (http://www.facts-consortium.co.za/?page_id=227), which aims to confirm the CAPRISA results. In the Austin trial, sensitivity was fairly low (81%) by one of the 3 observers. In addition, only 47% to 77% of reportedly used applicators tested positive, whereas 99.4% of the same applicators showed evidence of vaginal insertion by Gram stain.11 It may be speculated that swabbing the applicators to obtain samples for Gram staining interfered with staining them or reading the stained results, but there is no stated evidence of this. In the other trial of HTI applicators using FD&C Blue No. 1 dye carried out by Keller et al., specificity was low: 41% overall, with 27% by one observer and 54% by the other observer.12 Also, only 75% of applicators that were returned unopened by participants were judged to be negative in the dye test. This low specificity (high false-positive rate) is troubling, as it is the only trial that involved TFV, and similar results could lead to an overestimate of product use and an underestimate of product efficacy in an HIV prevention trial. It does not appear that there is a dye technique for HTI applicators that gives acceptable results and can be used without extensive personal protection. Less commonly used applicator tests include the Gram staining mentioned earlier in the text, which requires microscopic evaluation of a specimen from each applicator and is labor-intensive,11 and testing for lactobacilli, which is limited by the fact that the organism cannot be cultured from all women, especially those with vaginitis.7 Moench et al evaluated ultraviolet (UV) light, Alcian blue, potassium iodide, and iodine for detecting insertion of HTI applicators and found low sensitivities (52%–67% of known positives tested positive) and fairly low specificities (63%–83%).15 The UV light method has subsequently been improved by changing the UV lamp source and excitation filter, substantially improving the signal. It has been tested on FACTS 001 applicators on which cervicovaginal fluid (CVF), TFV gel, the Universal Hydroxyethyl Cellulose (HEC) placebo gel,16 TFV gel combined with CVF, and placebo gel combined with CVF have been applied. The gels fluoresce with about equal intensity but CVF is about 20 times more fluorescent than the gels. The presence of either gel did not interfere with detection of CVF. Detection using the original UV light system was more sensitive when applied to applicators that had been used within 24 hours after a previous applicator, as would occur during a daily dosing regimen, than without prior applicator insertion, as would often occur during a pericoital regimen.15 However, the improved system may be suitable for use during pericoital dosing as in FACTS 001 (Moench, MD, Personal Communication, June 1, 2012). Another innovative non–dye-based biomarker is a DNA multiplex polymerase chain reaction test that has been developed, which can detect 2 Y-chromosome-linked genes, SRY and TSPY4, as well as amelogenin, a gene found on both the X and Y chromosomes. The polymerase chain reaction test can distinguish between female-only DNA (a marker for vaginal insertion) and the additional presence of male DNA as a result of semen exposure, which could indicate unprotected intercourse before applicator insertion in pericoital and postexposure regimens (Jacot TA. Unpublished data, May 2012). One concern that has been raised is whether informing study participants that their adherence is being measured by applicator staining (or any other method) will, in fact, influence their adherence. Recent research among sex workers in India suggests that this may not be the case for applicator staining: prior knowledge that dye staining of applicators was going to be used did not improve self-reports nor did knowing the purpose of the dye stain assay encourage gel use in this population, indicating that in some settings, these assays may be used in the context of clinical trials without changing behavior.17 Nevertheless, at their best, applicator tests of any type can neither indicate whether the gel was expelled into the vagina or anywhere else nor can they tell by whom the applicator was used. They cannot tell whether the applicator was used before sex or if sex took place at all, although the DNA test can indicate if semen was present at the time of applicator insertion. They allow calculation of the maximum number of applicators that could possibly have been used, but cannot correct for lost, shared, or discarded applicators. One alternative to applicator testing is electronic monitoring of drug dispensing. The Medical Event Monitoring System has been studied since the early 1990s.18 An application of this type of technology is the Wisebag (Wisepill Technologies Adherence Management Solutions, South Africa): a portable lunch box–like container that can be used to store unused applicators before use. When the bag is opened, an electronic signal can be stored in the device for later downloading in the clinic, or transmitted in real time to the clinical site. In the case of the latter, a reminder can also be automatically sent to the participant if it appears that she has forgotten a dose, thereby not only measuring adherence but also (it is hoped) improving it. The Wisebag was piloted among 10 participants in the CAPRISA 004 trial19 and was studied in a Microbicide Trials Network trial entitled, “The Wisebag Observational Pilot Study: Feasibility, Acceptability, and Performance of an Electronic Event-based Monitoring System (Wisebag).” In the latter trial, Wisebag data indicated 26% adherence to daily dosing for 13 days, whereas diary and self-report showed 46% to 48%, respectively.20 Electronic device monitoring can most reliably identify periods during which a product was not used (unless applicators were removed and stored elsewhere for later use) but has many of the same limitations of an applicator test, e.g., not indicating whether the gel was expelled, where, by whom, and when in relation to sex. Levels of drug in vivo can serve as an indicator of adherence. In CAPRISA 004, it was found that the HIV incidence among women with a TFV concentration >1000 ng/mL in undiluted aspirated CVF was significantly lower than that of placebo users, whereas the incidence among women with TFV concentrations at or below this level was not different from placebo users.21 A recently completed trial of daily oral emtricitabine/tenofovir disoproxil fumarate (FTC/TDF) for preexposure prophylaxis did not demonstrate a reduction in risk of HIV acquisition in participants randomized to FTC/TDF. Adherence by self-report and pill count data were higher than by plasma drug concentration analysis. Among women assigned to FTC/TDF, drug was detectable in plasma in fewer than 50% of infected cases and uninfected controls matched on time of infection.22 Thus, objective measures of adherence, such as drug concentrations, can be of critical value in explaining trial results. Getting a precise picture of adherence has been called the Achilles' heel of microbicide research. Understanding when a drug was taken is essential to interpreting trial results. The authors of the MIRA article wrote, “For all prevention methods that are user dependent, continuing research on ways to increase adherence (and measurement thereof) is as important as the need to develop and test new prevention methods.”2 The need for an objective marker of adherence remains unmet, and efforts to develop such markers deserve substantial support.