The Centers for Disease Control and Prevention (CDC) estimates that each year there are more than 9 million episodes of foodborne illness, over 55,000 hospitalizations, and at least 1,351 deaths that can be attributed to foods consumed in the United States (1
). The CDC regularly publishes reports that summarize data on surveillance for foodborne disease outbreaks in the United States (2–6
). Those reports list more than 30 contributing factors linked to foodborne disease outbreaks in the year or years summarized in the reporting period. Factors are grouped into three categories related to contamination with and proliferation and survival of foodborne pathogens. Food handlers or others suspected to be infectious are linked to several contamination factors. One factor is specifically related to cross-contamination from surfaces and not ill individuals. When those surface cross-contamination data are summarized from 1998 to the present, about 12% of all outbreaks reported to the CDC are linked in some way to this type of surface cross-contamination. This is the 6th most common contributing factor (out of 32) (2–6
Household and other surface types have been the focus of numerous cross-contamination studies; the surfaces studied include ceramic tile (7–9
), stainless steel (7
), wood (8
), glass (7
), plastic (7
), and carpet (8
). Stainless steel has often been considered the optimal material choice for kitchen sinks and commercial food preparation surfaces because of its resistance to corrosion, mechanical strength, ease of cleaning, and resistance to chemical degradation (17
), although stainless steel may have higher bacterial transfer rates than other surfaces (19–21
). Tile is also a common surface found in homes; the variations of tile (unglazed versus glazed) may have an effect on the bacterial transfer rate because of varying surface topography (22
). Wood surfaces are commonly found in households, either as flooring or as cutting board surfaces. The sanitary properties of wood cutting boards have been compared to those of plastic cutting boards (23
), and the studies have come to contradictory conclusions, in part because of differences in the methods used. The United States Department of Agriculture recommends one cutting board for produce and bread and a separate cutting board for raw meat, poultry, and seafood (25
). Carpet is a likely site of contamination in the household, and inactivating or removing bacteria by conventional cleaning methods is difficult once the carpet is contaminated (16
). Microorganisms on carpet can be controlled by specific chemical treatments of the fibers or the materials used in constructing the carpet (26
The popular-culture notion of the “five-second rule” is that food dropped on the floor for <5 s is “safe” because bacteria need time to transfer. The rule has been explored to a limited degree in the published literature and popular culture. Previous studies on the five-second rule used different surfaces, foods, organisms, contact times, and numbers of replicates, making comparisons and conclusions difficult. The first known research recorded on this topic was performed at the University of Illinois but was never published in the peer-reviewed literature (27
). These researchers used tile inoculated with Escherichia coli
and studied transfer to cookies and gummy bears and found that bacterial transfer was observed in <5 s (27
). The popular television show MythBusters aired an episode on the five-second rule in 2005 and found no conclusive difference between contact times of 2 and 6 s (28
). In the only peer-reviewed research on the topic, researchers at Clemson University concluded that longer contact times (5, 30, and 60 s) did increase the transfer of Salmonella enterica
serovar Typhimurium from wood, tile, or carpet to bologna or bread, but only ≥8 h after the surface was inoculated (8
). Researchers at Aston University in the United Kingdom published a press release in 2014 reporting that contact time significantly affected the transfer of both E. coli
- and Staphylococcus aureus
-contaminated surfaces (carpet, laminate, and tile) to food (toast, pasta, biscuit, and a sticky sweet) (29
). Discovery Science Channel's The Quick and the Curious television show aired a short segment offering cookies to strangers in a park—after dropping them onto the ground. The show's narrator stated that “moist foods left longer than 30 seconds collect 10 times the bacteria than [sic] those snapped up after only three” but offered no data in support of this statement (30
This research sought to quantify cross-contamination between a variety of foods and common kitchen surfaces while varying the contact time and bacterial matrix and to do so in an extensive and comprehensive manner. The results described here advance our understanding of cross-contamination and the factors that influence it. This research informs the popular culture and enhances our scientific understanding of cross-contamination and the factors that influence it.
MATERIALS AND METHODS
Bacterial strain and preparation of culture.
A nonpathogenic, food-grade microorganism, Enterobacter aerogenes
B199A, with attachment characteristics similar to those of Salmonella
, was used for all experiments (Vivolac Cultures, Indianapolis, IN) (14
). The E. aerogenes
strain used here is resistant to nalidixic acid, which allows it to be enumerated in the presence of other microorganisms on food samples or surfaces. Control experiments (done by sampling and plating onto tryptic soy agar [Difco, BD, Sparks, MD] with 50 μg/ml nalidixic acid [Sigma Chemical Co., St. Louis, MO] [TSA-na]) showed that nalidixic acid-resistant E. aerogenes
cells were not initially present on any of the foods or surfaces at levels of >2 log CFU/surface or food.
Cultures were prepared on the basis of prior work done in our lab (13
) and by others (14
). A frozen stock of E. aerogenes
in 80% sterile glycerol was streaked onto TSA-na. One colony from each plate was transferred to 10 ml of tryptic soy broth (Bacto, BD, Sparks, MD) with 50 μg/ml nalidixic acid (TSB-na) and incubated at 37°C for 24 h. Inoculum matrices were of two types, one using cells harvested by centrifugation at 5,000 × g
for 10 min and washed twice in 10 ml of 0.1% peptone (Difco, BD) and one using cells taken directly from an inoculated overnight TSB-na culture. A final concentration of ∼108
CFU/ml was verified by enumeration on TSA-na.
Preparation of domestic surfaces.
Four different surfaces typical of those found in domestic environments were used: stainless steel (type 304, 0.018-in. thickness, 16 gauge; OnlineMetals, Seattle, WA), ceramic glazed tile (Brancacci Windrift Beige; Daltile, Dallas, TX), maple laminate wood (Northern Maple; Mohawk, Calhoun, GA), and indoor-outdoor carpet (Morella; Foss Manufacturing, Hampton, NH). They were ordered online or purchased from a local home improvement store. Surface materials were cut into coupons (5 by 5 cm). The stainless steel and ceramic tile coupons were disinfected prior to inoculation by soaking in 70% ethanol for 1 h, removed, air dried, and autoclaved. Disinfection of wood and carpet coupons caused structural changes, so these were discarded after autoclaving following a single use.
Four foods (watermelon, white bread [ShopRite, Wakefern Food Corp., Elizabeth, NJ], unsalted butter [ShopRite, Wakefern Food Corp., Elizabeth, NJ], and gummy candy [Haribo Strawberries]) were purchased online or from a local supermarket. Whole watermelon was stored at 4°C prior to use. The watermelon (flesh only) and bread (excluding the crust) were cut into pieces (approximately 4 by 4 cm). Unsalted butter was brought to ambient temperature (∼24°C) prior to being spread onto bread. All of the foods had equivalent contact areas (∼16 cm2). The pH and water activity (aW) of samples were measured in triplicate with a surface pH probe (Accumet Basic AB15 pH meter; Fisher Scientific) and an aW meter (Rotronic Instrument Corp., Hauppauge, NY), respectively.
Transfer between food and surfaces.
Transfer scenarios were evaluated for each contact surface type (4
), each food type (4
), four contact times, and two inoculum matrices, totaling 128 scenarios. Each scenario was replicated 20 times, totaling 2,560 measurements. Each contact surface type was spot inoculated with 1 ml of inoculum by using 8 to 10 drops spread over the 5- by 5-cm surface. The surfaces were placed in a biosafety cabinet (SterilGARD Hood; The Baker Company, Inc., Sanford, ME) for 5 h, after which the surfaces were visibly dry. Prior to 5 h, surfaces were still wet and at times longer than 5 h, the difference in recovery rate between the inoculum matrices increased. Both the peptone buffer and TSB-na inoculum matrices yielded an approximate concentration of 107
CFU/surface after drying. Foods were dropped onto the respective surfaces by using gloved hands from a height of 12.5 cm and left to rest for four different times (<1, 5, 30, and 300 s). A height of 12.5 cm was selected because it was the greatest height possible that still ensured that the entire food sample would reliably contact the entire surface.
Surfaces were placed into a sterile Whirl-Pak filter bag (Nasco, Fort Atkinson, WI), 20 ml of peptone buffer was added, and the mixture was hand massaged for 2 min. Foods were placed into a sterile filter bag (Fisherbrand Lab Blender Bags) with 50 ml of peptone buffer, and the samples were homogenized (Stomacher; Cooke Laboratory Products, Alexandria, VA) for 3 min. Surfaces and food samples were serially diluted in 0.1% peptone buffer and surface plated (0.1 ml) onto TSA-na for enumeration of E. aerogenes colonies. Plates were incubated at 37°C for 24 h. Colonies were counted, and population levels were expressed in numbers of CFU per food or surface sample.
Percent transfer was calculated as follows: [total CFU food/(total CFU food + total CFU surface)] × 100. Percent rates of transfer from surface to food were log transformed with Microsoft Excel (Microsoft, Redmond, WA) and SigmaPlot (Systat Software Inc., San Jose, CA), as prior research has shown that untransformed transfer rates are highly skewed and log-transformed transfer rates are approximately normally distributed (13
). When foods contained less than the detection limit (2 log CFU), transfer rates were calculated as if the concentration on the foods was at the detection limit. Variables and the interactions between variables were considered significant at a P
value of <0.05. Multiple linear regression analysis was performed with StatPlus for Microsoft Excel (AnalystSoft, Inc., Walnut, CA). Quantitative values were given to the surfaces tile (0), stainless steel (1), wood (2), and carpet (3); the foods bread (0), bread with butter (1), gummy candy (2), and watermelon (3); and the matrices TSB (0) and buffer (1) for regression analysis.
Our study shows that bacterial transfer is dependent on the surface, food type, contact time, and inoculum matrix. Studies involving transfer from similar surfaces to foods have come to various conclusions (7
). These differences may be due to the range of experimental procedures among published studies. Differences include the times of contact between surfaces (7
), the organisms used (7
), and the foods and contact surfaces used (7
), each of which can result in different outcomes. Our research also shows that the nature of the matrix containing the cells inoculated onto the surface can play an important role, even when all other experimental variables are the same, an observation we have seldom seen reported in the literature. Studies of bacterial adhesion to surfaces used a variety of drying times, in comparison to the 5-h drying time used in this study (7
). Additionally, there is a difference in data analysis regarding transfer rates. Some studies determined the transfer rate by calculating the recipient surface/source surface (13
), whereas in our study, the transfer rate was analyzed by calculating the recipient surface/(source surface + recipient surface) (7
), which can lead to slight differences when the number of bacteria transferred to the recipient surface is high. More importantly, some studies used very small numbers of replicates and/or failed to statistically transform the percent transfer rates and may have come to erroneous conclusions (31
). Although not always reported in studies, standard deviation is a good indication of the degree of variability (13
). In our study, the standard deviation varied considerably with the food tested.
Although pressure was not a variable in our study, it may play a role in facilitating bacterial transfer. Kusumaningrum et al. found that more transfer occurred when light pressure (20 g/cm2
) was applied, although differences were slight (∼0.3-log percent transfer difference) (33
). Mbithi et al. used pressures of 200 and 1,000 g/cm2
, with and without friction, and found that differences in transfer rates were also small (a ∼0.5-log percent transfer difference when pressure was applied) (37
). Research by D'Souza et al. showed that pressure changes ranging from ∼1 to 100 g/cm2
had no effect on virus transfer (38
). Later research in the same laboratory showed more transfer at higher pressures (∼100 g/cm2
) than at lower pressures (∼10 g/cm2
), especially when the inoculum was drier (39
Our data clearly show that contact time does influence bacterial transfer, with more bacteria transferred at longer times. Peer-reviewed research by Dawson et al. reported that a longer food contact time (5, 30, or 60 s) did result in greater transfer but only at longer drying times (≥8 h) (8
) roughly equivalent to our drying time of 5 h. Non-peer-reviewed research at the University of Illinois on bacterial transfer from tile inoculated with generic E. coli
to cookies and gummy bears found that bacterial transfer was observed in <5 s (27
) (consistent with our <1-s observations), although other contact times were not studied. The popular television show MythBusters (28
) aired an episode on the five-second rule and found no conclusive difference when pastrami and crackers were exposed to contaminated tile with contact times of 2 and 6 s. It is unclear from viewing the episode what was used to contaminate the tile surface, although the inoculated tile was left for 5 days before the experiment was begun. MythBusters also used <10 replicates per scenario. A press release by Aston University, in the United Kingdom, showed that time significantly affected transfer, depending on the contaminated surface and the food (29
). The Aston University study observed the transfer of E. coli
and S. aureus
from carpet, wood, and tile to toast, pasta, biscuit, and a sticky sweet at 3- and 30-s contact times. Moist foods that contacted contaminated wood and tile showed higher transfer rates, and longer times increased the transfer between these foods and surfaces. The Aston University study shows that transfer from carpet was not affected by the food composition or the contact time (29
Our data show that the rate of bacterial transfer was greatest for tile, stainless steel, and wood surfaces at 300 s. The food with the highest transfer rate was watermelon, regardless of the contact time, which may be due to several factors. When watermelon is cut, it is very moist, and moisture is known to facilitate transfer (40
), regardless of whether the contact surface is dry or wet. Watermelon may also present a flatter, more uniform surface at the microscopic level than bread or gummy candies. Jensen et al. also found that transfer from stainless steel or tile to watermelon was the highest of any produce type used in their study (7
). Kusumaningrum et al. measured the rates of transfer to cut cucumber from stainless steel and observed that almost all of the bacteria (∼100%) transferred to the cucumber, regardless of pressure (33
). Cut cucumbers also have a moist, uniform surface, which may facilitate bacterial transfer. We observed lower transfer rates (∼0.2%) when transfer was from carpet to food. Carpet may promote less bacterial transfer because of bacterial attachment to or infiltration of absorbent carpet fibers. Dawson et al. also found that transfer from carpet to bologna was very low (<0.5%) in comparison to transfer from wood and tile to bologna (5 to 68%) (8
The starting concentration of all of the surfaces in our experiments was ∼7 log CFU/surface. Although this was not a variable explicitly considered, the starting concentration may have an effect on how much bacterial transfer to the recipient surface occurs. Montville and Schaffner reported on the influence of inoculum size on bacterial cross-contamination between surfaces. Their results showed that the effect of inoculum size on the transfer rate was statistically significant (P
< 0.0001) for all transfer rate data and that a greater inoculum size resulted in a lower transfer rate (41
Transfer of bacteria from surfaces to food appears to be affected most by the moisture of the food, as shown by the transfer of E. aerogenes from tile, stainless steel, wood, and carpet to watermelon. Longer food contact times usually resulted in the transfer of more bacteria from each surface to food. Carpet has very low transfer rates, compared with those of tile and stainless steel, whereas transfer from wood is more variable. The topography of the surface and food seems to play an important role in bacterial transfer. The risk of illness resulting from deciding to consume food that has fallen on the floor depends on factors including the prevalence, concentration, and type of organism; the nature of the food (especially moisture); and the nature of the surface topology; as well as the length of time the food is in contact with the surface. Although this research shows that the five-second rule is “real” in the sense that longer contact time resulted in more transfer, it also shows that other factors, including the nature of the food and the surface, are of equal or greater importance. The five-second rule is a significant oversimplification of what actually happens when bacteria transfer from a surface to food.