refining predictions for water pathogens

1
Refining predictions for water pathogens When Patti Fowler decides wheth- er to close or open a commercial shellfish bed because of contami- nants, she must use one of two tests. These tests, which were agreed upon for interstate use by the National Shellfish Sanitation Program (NSSP), measure fecal coliform levels and have been used for decades. “It’s the gold stan- dard,” says Fowler, manager of shellfish beds for North Caro- lina’s Department of Environ- ment and Natural Resources. Because shellfish are sold across state boundaries, re- source managers in various parts of the country must be sure to use the same tests for bacteria that serve as prox- ies for the presence of human pathogens, such as E. coli and other bugs. The consequences of an inaccurate result could be needlessly closed shellfish beds and loss of profit or, on the other hand, sick people if the predictions fail to show that pathogens are present. Unfortunately, the tests some- times are misleading, accord- ing to new research published in ES&T (pp 4676–4682). The researchers suggest a statisti- cal technique that could help bring predictions closer to reality. Although it may seem that di- rect measurements taken at a shellfish bed would be the best way to tell if the oysters or oth- er creatures there contain con- taminants, such testing would be incredibly expensive and time- consuming. “We will never know what the concentration in the bed really is, but that’s what we’re af- ter,” says Andrew Gronewold, a graduate student at Duke Universi- ty and lead author of the research. “If I could submerge myself in that body of water and count every mi- crobe,” the results still wouldn’t be accurate because of the vari- ability in water conditions, which fluctuate within a water column throughout a day, over a season, and over a year. Yet the uncertain- ties inherent in the lab tests and models can be ameliorated by us- ing Bayesian statistics, Gronewold and his colleagues assert. For current fecal coliform tests, water samples taken from recre- ational beaches or near shellfish- ing sites are tested for what are known as most probable num- ber (MPN) and colony-forming units (CFU). These tests indicate the number of microbes present or colonies formed, respectively. Shellfish beds must meet the stan- dards set by NSSP and are judged to be safe (or not) for human con- sumption on the basis of 30 mea- surements. Hypothetically, those measurements could be taken dai- ly or intermittently over a year or several years. (Recreational beach managers assessing water cleanli- ness for human swimmers gener- ally rely on one spot sample that can take 24 hours to evaluate.) To clarify the accuracy of the tests, Gronewold and colleagues ran statistical models on hundreds of data points spread out over time and space for about 30 shellfish sites in North Carolina. They then back-calculated, in a sense, to de- termine how the methods and the ensuing predictions matched the probable reality for the concentra- tions in the shellfish beds. These new statistical methods trans- late MPN and CFU tests into “real numbers,” Fowler says. Rachel Noble of the University of North Carolina Chapel Hill, a marine microbial ecologist who advised the team, says the probability of pathogens pres- ent in water can be thought of in terms of a mountain slope. Each method has its own mountain profile. The less un- certainty a method has, the narrower the cone, so that the mountain slopes are in effect steeper. The new Bayesian analysis allows resource man- agers to get a better picture of their “mountain” profiles. That reduction in uncertainty increases the confidence with which managers might make their decisions, no matter which test they are using. The uncertainty decreases for both scenarios—closing contami- nated sites or opening uncon- taminated ones. The bottom line, however, may be that fecal coliform tests—whether MPN or CFU— are no longer the best indicators of pathogens present in shellfish beds or recreational waters. It would be better to test for actual pathogens with molecular probes, suggests David Parkhurst, professor emeri- tus of environmental sciences at Indiana University. Researchers at the U.S. EPA and elsewhere are working on this option, and quan- titative polymerase chain reaction and other methods are under way for getting direct results within hours. In the meantime, because of federal and state guidelines and the need to validate such relatively new techniques, Fowler and other managers must continue to use the accepted indicators—fecal coli- form tests—to make decisions. —NAOMI LUBICK Managers might needlessly close shellfish beds, like this one in North Carolina. COURTESY OF ANDREW GRONEWOLD JULY 1, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 4621

Upload: naomi

Post on 21-Feb-2017

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Refining predictions for water pathogens

Refining predictions for water pathogens

When Patti Fowler decides wheth-er to close or open a commercial shellfish bed because of contami-nants, she must use one of two tests. These tests, which were agreed upon for interstate use by the National Shellfish Sanitation Program (NSSP), measure fecal coliform levels and have been used for decades. “It’s the gold stan-dard,” says Fowler, manager of shellfish beds for North Caro-lina’s Department of Environ-ment and Natural Resources.

Because shellfish are sold across state boundaries, re-source managers in various parts of the country must be sure to use the same tests for bacteria that serve as prox-ies for the presence of human pathogens, such as E. coli and other bugs. The consequences of an inaccurate result could be needlessly closed shellfish beds and loss of profit or, on the other hand, sick people if the predictions fail to show that pathogens are present. Unfortunately, the tests some-times are misleading, accord-ing to new research published in ES&T (pp 4676–4682). The researchers suggest a statisti-cal technique that could help bring predictions closer to reality.

Although it may seem that di-rect measurements taken at a shellfish bed would be the best way to tell if the oysters or oth-er creatures there contain con-taminants, such testing would be incredibly expensive and time-consuming. “We will never know what the concentration in the bed really is, but that’s what we’re af-ter,” says Andrew Gronewold, a graduate student at Duke Universi-ty and lead author of the research. “If I could submerge myself in that body of water and count every mi-crobe,” the results still wouldn’t be accurate because of the vari-ability in water conditions, which fluctuate within a water column throughout a day, over a season,

and over a year. Yet the uncertain-ties inherent in the lab tests and models can be ameliorated by us-ing Bayesian statistics, Gronewold and his colleagues assert.

For current fecal coliform tests, water samples taken from recre-ational beaches or near shellfish-ing sites are tested for what are

known as most probable num-ber (MPN) and colony-forming units (CFU). These tests indicate the number of microbes present or colonies formed, respectively. Shellfish beds must meet the stan-dards set by NSSP and are judged to be safe (or not) for human con-sumption on the basis of 30 mea-surements. Hypothetically, those measurements could be taken dai-ly or intermittently over a year or several years. (Recreational beach managers assessing water cleanli-ness for human swimmers gener-ally rely on one spot sample that can take 24 hours to evaluate.)

To clarify the accuracy of the tests, Gronewold and colleagues ran statistical models on hundreds of data points spread out over time and space for about 30 shellfish

sites in North Carolina. They then back-calculated, in a sense, to de-termine how the methods and the ensuing predictions matched the probable reality for the concentra-tions in the shellfish beds. These new statistical methods trans-late MPN and CFU tests into “real numbers,” Fowler says.

Rachel Noble of the University of North Carolina Chapel Hill, a marine microbial ecologist who

advised the team, says the probability of pathogens pres-ent in water can be thought of in terms of a mountain slope. Each method has its own mountain profile. The less un-certainty a method has, the narrower the cone, so that the mountain slopes are in effect steeper. The new Bayesian analysis allows resource man-agers to get a better picture of their “mountain” profiles. That reduction in uncertainty increases the confidence with which managers might make their decisions, no matter which test they are using. The uncertainty decreases for both scenarios—closing contami-nated sites or opening uncon-taminated ones.

The bottom line, however, may be that fecal coliform tests—whether MPN or CFU—

are no longer the best indicators of pathogens present in shellfish beds or recreational waters. It would be better to test for actual pathogens with molecular probes, suggests David Parkhurst, professor emeri-tus of environmental sciences at Indiana University. Researchers at the U.S. EPA and elsewhere are working on this option, and quan-titative polymerase chain reaction and other methods are under way for getting direct results within hours.

In the meantime, because of federal and state guidelines and the need to validate such relatively new techniques, Fowler and other managers must continue to use the accepted indicators—fecal coli-form tests—to make decisions.

—NAOMI LUBICK

Managers might needlessly close shellfish beds, like this one in North Carolina.

CO

UR

TES

Y O

F A

ND

REW

GR

ON

EWO

LD

July 1, 2008 / EnvironmEntal SciEncE & tEchnology ■ 4621