In this double-blind clinical trial, we randomly assigned households to one of 2 intervention groups: those who used handwashing and household cleaning products with antibacterial ingredients and those who used products without such ingredients. The interventions lasted for 48 weeks.
Sample and Setting
We conducted the study in an inner-city neighborhood in northern Manhattan, New York, with a predominantly immigrant population in multigenerational households. Almost 30% of residents spoke little or no English, and about 90% of the households had telephones (13
). To qualify for the study, a household unit had to include 3 or more persons with at least one preschool-age child and had to have access to a telephone. In addition, household members had to speak English or Spanish. In a preliminary survey (14
) conducted in this neighborhood, 78.5% of 398 households reported infectious disease symptoms within the previous month, and in 37.9% of these households, at least one person sought medical attention and received specific treatment or antibiotics for an infectious disease symptom or symptoms. On the basis of this pilot work, we concluded that a randomized clinical trial with sufficient statistical power was feasible.
Household cleaning products containing antibacterial ingredients are widely available and popular. Although manufacturers use claims of health benefits to market these products, evidence linking the use of antibacterial products to health outcomes has been lacking.
This innovative trial found no difference in episodes of infectious disease symptoms over one year in 228 inner-city households randomly assigned to use antibacterial household cleaning products or identically packaged products without antibacterial ingredients.
These findings highlight the need to better educate consumers about the use and limitations of household antibacterial cleaning products.
Recruitment was by word of mouth, referral, and English- and Spanish-language flyers (preapproved by the institutional review board) posted throughout the community. Participants were recruited by an experienced, trained interviewer who resided in the community and who was a native Spanish speaker.
We determined sample size by power analysis. With 100 households for each intervention group and a household incidence of infectious disease symptoms of about 35% per month, on the basis of the pilot study, it would be possible to detect an absolute difference between the 2 intervention groups of 20 percentage points or more (for example, from 35% to 15%) with a power of 80% and an α
value of less than 0.05 (15
). We recruited an additional 19% above this desired sample size to account for potential loss to follow-up and dropouts. A total of 238 households were randomly assigned, and 224 (94.1%) completed the entire 48 weeks of data collection. Fourteen households (5.9%) did not complete the entire study period, 9 (64.3%) because the household moved out of the study area, 3 (21.4%) because the household did not continue to use the products, and 2 (14.3%) because the household was inadvertently supplied with the wrong product ().
Profile of randomized, clinical trial.
Criteria for selecting products to be tested were as follows: The products had to be readily available over the counter; have the same or similar formulation, except for the presence or absence of an antibacterial ingredient; be representative of a particular category of product so that results could be generalized to other similar products; and be developed by reputable companies known for good manufacturing practices. Antibacterial was defined as the presence of triclosan, quaternary ammonium compounds, hypochlorite, or another recognized microbicidal agent in amounts greater than preservative levels. Also, the product label had to include the term antibacterial or disinfectant.
Households randomly assigned to the antibacterial group were provided with the following: a liquid kitchen spray and “all-purpose” hard-surface cleaner containing a quaternary ammonium compound, liquid handwashing soap containing triclosan, and a laundry detergent containing oxygenated bleach. The nonantibacterial group received parallel products with similar compositions that did not contain antibacterial ingredients. Both intervention groups were provided with the same liquid dishwashing detergent and bar soap, neither of which contained antibacterial ingredients.
The study was approved by the Columbia Presbyterian Medical Center Institutional Review Board. After we obtained written informed consent, households were randomly assigned to one of the intervention groups; the master key code was retained by the biostatistician. All products were provided without cost, were packaged identically with a generic label indicating their use, and were delivered to the household monthly. On the initial home visit, we collected baseline data on home hygiene practices and the presence of infectious disease symptoms within the previous month for each household member by using our Home Hygiene Assessment Form.
We made a weekly telephone call and a monthly home visit to each household. During the monthly visit, we assessed adherence to the product regimens by weighing the remainder of products with a postal scale and inspecting the home for the presence of other products. Every 3 months, we assessed symptoms in individual household members, and the Home Hygiene Assessment Form was readministered to determine whether any hygienic practices had changed. We conducted an average of 226 interviews each week. For most of the households (98.8%), at least 20 weekly interviews were completed, and for 89.0% of households, 45 or more weekly interviews were completed.
Data were collected by 3 interviewers who received extensive training using a written orientation manual, practice sessions with return demonstrations, and inter-rater reliability assessments. The interviewers and project director were native Spanish speakers; 3 were physicians, and the fourth was a trained community health worker. Initially and on a random monthly basis, each interviewer was accompanied by the project director on 10% of the home visits for ongoing quality control.
Because cleaning and hygiene practices within the home would probably affect the dependent variable, infectious disease symptoms, we collected extensive data on cleaning and hygiene practices at baseline and at quarterly intervals. The Home Hygiene Assessment Form is a 31-page interview booklet that includes questions about demographic characteristics and illness (age, sex, ethnicity, country of birth, hours per week spent outside the home, type of work for adults, school or child-care arrangements for children, state of health, and presence of chronic diseases), home hygiene practices (54 items), and other relevant household factors (numbers and ages of household members, size of living space, presence of pets and visitors, type of building, and heating and cooling systems). We also asked participants about their attitudes and beliefs about how “germs” are spread and what they did to prevent infections in their home. The components of the instrument were originally derived from a literature search, focus groups of consumers, and a panel of environmental sanitation experts. The Home Hygiene Assessment Form was tested extensively for validity and reliability (16
). Whenever possible, direct observations were made to confirm self-reports.
Measurement of Dependent Variable (Infectious Disease Symptoms)
The presence of infections was assessed symptomatically. We instructed participants to call their interviewer if any member of the household had vomiting, diarrhea, fever, sore throat, cough, runny nose, skin infection, or conjunctivitis (“pinkeye”). We provided each household with a supply of single-use thermometers (Tempa-DOT, 3M Health Care, St. Paul, Minnesota) and gave instructions for their use.
In the weekly telephone call from the interviewer, we also solicited information on symptoms. If participants reported a cough, they were queried by a physician about whether it could be due to allergies or asthma. The cough was recorded only when other noninfectious causes were ruled out. Sore throat was not recorded for children younger than 3 years of age. If one or more symptoms were present, the informant was asked whether medical attention was sought, whether any treatment and antibiotics were administered, and whether the symptom or symptoms resulted in missed work or school.
We assessed the reliability and validity of the self-reports of symptoms as follows. For the first 100 reports of illness, an on-call physician verified the presence of the symptom or symptoms by direct observation in a home visit. In 93 of 100 reports, the physician was able to directly confirm the presence of the reported symptom. In addition, in 3 of these 100 home visits, a symptom that had not been reported by a participant was identified by the interviewer. Thus, the sensitivity and specificity of the first 100 self-reports of symptoms were 0.93 and 0.97, respectively. We considered this an acceptable level of reliability, and for the remainder of the study, we did not continue to have on-call staff verify every symptom report; however, symptoms present at the monthly home visits were confirmed by direct observation. No treatment was provided during the study; if symptoms were judged to warrant attention, participants were referred to their primary care provider or to a local urgent-care clinic.
Data were entered by an independent data entry firm (CDPS, Inc., Milford, Ohio). Members of the research team remained blinded to the household study group until after analyses were completed. The primary outcome of the study was the presence of at least one infectious disease symptom within the household for each one-month period. For the design of intention-to-treat analysis, all households randomly assigned to the 2 groups were included.
We performed chi-square or Student t
-tests to compare the characteristics of the antibacterial and nonantibacterial groups for each hygienic practice and demographic variable, as well as the health status of household members at baseline. Unadjusted relative risks and 95% CIs were calculated for each symptom. We then performed logistic regression analysis by using the SUDAAN software program (Research Triangle Institute, Research Triangle Park, North Carolina) (17
). The generalized estimating equations approach was used because it accounts for intracorrelated binary data at each time point (18
). The estimated relative risk was generated by adjustment for potential confounders, including the number of children younger than 6 years of age, the number of people who rated their health as poor or fair or who had chronic conditions, the number of people who spent 40 or more hours outside the household per week, the size of the household, and any factors that differed substantially between the intervention groups in the univariate analyses.
To determine whether the use of antibacterial products had a cumulative effect, we also tested in separate regression models the interaction of treatment with the number of months that each group used the assigned products. In addition, we used Poisson regression models to examine the number of different symptoms reported by each household, and the incidence density ratio comparing the number of symptoms in the antibacterial and nonantibacterial groups was estimated. Finally, using chisquare analyses, we examined the effect of the intervention among persons (not at the household level) who might be at particular risk for infection—those with poor health or chronic disease, children who were 5 years of age or younger or were attending daycare, and adults working outside the household for 40 or more hours per week.
Role of the Funding Source
The funding source had no role in the design, conduct, or reporting of the study or in the decision to submit the manuscript for publication.