The data for this study were collected as part of a double-masked and randomized home intervention trial (4); participant enrollment began in October 2000, and follow-up occurred for a 12-month period. The methods and randomization procedures for this study have been reported elsewhere (5). A total of 238 households were recruited at baseline; 224 households completed the entire 1-year follow-up (Figure 1). The study was approved by Columbia University Medical Center Institutional Review Board.

Households were supplied with over-the counter, generically repackaged consumer cleaning and personal hygiene products free of charge on a monthly or as-needed basis. Households randomly assigned to use antibacterial products received the following: 1) liquid handwashing soap containing 0.2% triclosan, 2) liquid kitchen spray and liquid all-purpose cleaner for hard surfaces that contained a quaternary ammonium component, and 3) oxygenated bleach laundry detergent. Households randomly assigned to the nonantibacterial group received the same products but without antibacterial ingredients. Both groups received the same nonantibacterial liquid dishwashing detergent and bars of body soap to control for potential use of other products that might contain antibacterial ingredients. Study participants were required to use only assigned home hygiene products and were asked not to change any of their normal hygiene practices. Participants, interviewers, and study coordinators were blinded to brand names and ingredients in all products. Adherence to product treatment group was assessed monthly, and products were weighed during each visit to monitor compliance. Households were immediately dropped from the study if they did not adhere to randomized treatments.

At baseline, and quarterly during the 1-year period, a trained interviewer collected demographic information from the person self-identified as the primary caregiver in the household. The baseline interview determined the type of handwashing soap, hygiene, and cleaning products that were used before randomization into the study (i.e., the brand and whether or not the ingredients were labeled as antibacterial). The baseline and quarterly assessment forms provided information such as the number and age of household members, childcare attendance, symptoms of infectious illnesses (fever, diarrhea, sore throat, vomiting, conjunctivitis, skin boils, runny nose), antimicrobial drug use, chronic diseases, self-rated health, birthplace, travel outside of the United States, and occupation. In addition, reported number of handwashes per day by the primary caregiver and a timed observation of the handwash before hand culturing were gathered.

The hands of the primary caregiver were cultured during the home visit at baseline and at the end of the 12-month period before and after washing with the assigned liquid handwashing product. The trained data collector used a coin flip to choose the test hand, which was then inserted into a sterile polyethylene bag containing 50 mL culture medium (0.075 mol/L phosphate buffer, pH 7.9, containing 0.1% polysorbate 80). The hand was massaged for 1 min through the wall of the bag containing culture medium. Only postwash samples were used in analyses since they were considered to be representative of normal versus transient flora found on hands.

The laboratory methods for this study have been described previously (5,6). The microbiologic analysis and antimicrobial drug–susceptibility testing were conducted at New York Presbyterian Hospital, Columbia University Medical Center, New York. Selective media were used to isolate gram-positive cocci, gram-negative bacteria (GNB), Staphylococcus aureus, and enterococci.

Only clinically important bacterial species that were prevalent (species with >38 isolates recovered at baseline and end of year combined) on the hands of homemakers were selected for susceptibility analyses (7,8). These included the following GNB: Acinetobacter baumannii, A. lwoffi, Enterobacter agglomerans, E. cloacae, Klebsiella pneumoniae, and Pseudomonas fluorescens/putida; and the following gram-positive staphylococci: S. aureus, S. warneri, S. epidermidis, and S. capitis. Therefore, only persons who were carrying at least 1 of these organisms on their hands were included in the final analyses (N = 164 at baseline and N = 201 at year-end). No significant differences were noted between the measured demographic characteristics (Tables 1 and 2 for listing of demographics) among those included in the final analyses versus those excluded (all p>0.10).

Bacterial isolates were tested against a panel of antimicrobial agents by using MicroScan WalkAway 96 SI (Dade Behring, Deerfield, IL, USA). Using the recommendations of the Clinical and Laboratory Standards Institute (formerly NCCLS), we classified antimicrobial drug susceptibility as resistant, intermediate, or susceptible to a particular antimicrobial agent (9). Organisms that tested as either resistant or intermediately resistant to antimicrobial agents were classified as "antibiotic resistant" (10). The selection of antimicrobial agents to be tested for each organism was based on clinical applicability of the antimicrobial drug and consistency with earlier studies that examined a link between triclosan and antimicrobial drug resistance (11–14). GNB were tested against several antimicrobial agents, and staphylococci were tested against oxacillin to indicate methicillin resistance. For analytic purposes, GNB species were classified as resistant if a given isolate was resistant to >1 antimicrobial agent(s).

Triclosan susceptibility was examined at Tufts University School of Medicine, Boston, Massachusetts, by using a modified NCCLS agar dilution method (10). Minimum inhibitory concentration (MIC) was defined as the lowest dilution of triclosan that inhibited visible growth. A detailed description of antimicrobial drug and triclosan testing, including controls used and MIC distribution for each organism, has been described previously (6). Since data from the literature regarding triclosan susceptibility testing are sparse and provide no standardized breakpoints (6), we dichotomized triclosan MIC values for each isolate by using the median MIC as a cutoff; low MIC represents less than or equal to the median value and high MIC indicates greater than the median value.

First, chi-square and Student t tests were used to compare demographic characteristics of antibacterial and nonantibacterial users. Next, chi-square tests were used to compare the overall proportion of antimicrobial drug–resistant isolates found on the hands of the antibacterial and nonantibacterial groups. Finally, multivariate logistic regression analyses were conducted to examine the relationship between antibacterial product use and 2 separate outcome variables: antimicrobial drug resistance (measured by the presence of >1 antimicrobial drug–resistant species on the hand) and increased triclosan MICs (measured by the presence of >1 species exhibiting a triclosan MIC above the median value).

Each potential covariate (i.e., characteristics of the household and primary caregiver) and our 2 outcome variables were examined in univariate analyses to establish criteria for inclusion in final multivariate models by using a p value <0.05 as the cutoff. Covariates meeting the cutoff criteria were included in multivariate models along with the main effect of the randomized treatment (i.e., antibacterial versus nonantibacterial product use). Analyses were conducted separately for baseline and after 1 year of study participation. Unadjusted and adjusted odds ratios (OR) and 95% confidence intervals (CIs) were generated from logistic regression analyses by using SPSS V.10 (SPSS Inc., Chicago, IL, USA).

At baseline, primary caregivers with higher than average CFU on their hands were twice as likely to carry antimicrobial drug–resistant organisms (Table 3). A slightly increased risk of carrying antimicrobial drug–resistant organisms occurred among those who washed their hands for a longer duration before the culture sample at baseline (Table 3). However, longer duration of handwashing was not associated with reduced bacterial CFU on hands (OR 1.02, 95% CI 0.99–1.06).

At year-end, both the number of times hands were washed per day and the presence of any household member(s) with a healthcare or daycare occupation were significantly associated with reduced carriage of antimicrobial drug–resistant organisms on hands of the primary caregiver (Table 3). Primary caregivers residing in households with members working in healthcare or daycare were significantly more likely to report above-average number of handwashes per day (OR 3.05, 95% CI 1.71–5.44). None of the other characteristics, such as health conditions or antimicrobial drug use, were significantly associated with carriage at baseline or after 1 year (all p>0.05).

Several hygiene-related factors were significantly associated with carriage, regardless of antibacterial product use. Longer handwashes were slightly associated with increased risk for carriage of antimicrobial drug–resistant species at baseline; as reported previously, these findings may be an artifact of sampling technique (5). The culture was taken directly after the handwash; an increased duration of the wash may have allowed greater dispersal of bacteria into the culture bag.

Primary caregivers residing in households with healthcare or daycare workers had significantly fewer antimicrobial drug–resistant organisms on their hands. This association appears to be influenced by above-average number of handwashes per day by the primary caregiver and indicates that hygiene, regardless of antibacterial ingredients, may reduce household transmission of antimicrobial drug–resistant bacteria.

A factor that might have attenuated the associations found in this study is a higher baseline level of antimicrobial drug resistance in this community. Higher baseline levels would make detecting small changes in susceptibility attributed solely to use of antibacterial cleaning and hygiene products more difficult. Most persons from our study population were from the Dominican Republic, a country that provides over-the-counter access to antimicrobial agents. In an earlier study within this same community, antimicrobial agents were taken by 354 (39%) of 911 persons reporting infectious disease symptoms within the previous 30 days, which suggests high levels of use (21). In addition, this study was conducted for a 1-year period and therefore may not adequately reflect the time-course for development of resistance attributable to use of antibacterial products. Changes in antimicrobial drug resistance during the 1-year period might have been lower than the level of detection that this study was statistically powered to identify. This study was designed to detect an OR >2.11 after 1 year of use, given a power of 80% and a 2-sided α level of 0.05.

Although triclosan susceptibility was examined among various species, we were not able to evaluate potential mechanisms for cross-resistance, such as overexpression of efflux pumps. In addition, when we examined the association between use of antibacterial cleaning and hygiene products and antimicrobial drug resistance, the definition of resistance (>1 organism[s] with antimicrobial drug resistance) did not allow exploration of the potential association with each separate species or antimicrobial drug tested. However, the purpose of our study was to examine overall trends and shifts in antimicrobial drug resistance attributed to the use of antibacterial cleaning and hygiene products, given that the effects of these products in the community are relatively unexplored.