Methods for primary concentration of viruses from water samples: a review and meta-analysis of recent studies



Jennifer L. Cashdollar, U.S. EPA, 26 W. Martin Luther King Dr., MS-587, Cincinnati, OH 45268, USA. E-mail:


Since the beginning of environmental virology in the mid-twentieth century, a key challenge to scientists in the environmental field has been how to collect, isolate and detect pathogenic viruses from water that is used for drinking and/or recreational purposes. Early studies investigated different types of membrane filters, with more sophisticated technologies being developed more recently. The purpose of this study was to look at the current state of the science of methods for the concentration of viruses from water. Several technologies were reviewed, and associated data were included in a meta-analysis which showed that electronegative filters, electropositive filters and ultrafilters are comparable in performance and that significant differences in recovery are due to virus type rather than filter type, water matrix or sample volume. This information is useful, as it will help to determine which method(s) should be used, particularly if there is a specific viral type being targeted for a particular study. In addition, it will be helpful when sampling different environmental water matrices and/or when budget allowance must be taken into consideration. Taken together, this will be useful in performing viral occurrence studies, which ultimately can help ensure safer water for both humans and the environment.


In the mid-nineteenth century, Adolf Meyer and Dmitri Ivanovski demonstrated that extracts from diseased plants could produce tobacco mosaic disease when inoculated onto healthy plants (Ivanovski 1892; Lechevalier 1972). These discoveries of the first virus, tobacco mosaic virus, opened the door to the field of virology (Lechevalier 1972) (Fig. 1). It was several years later, in 1908, that Austrian physicians Karl Landsteiner and E. Popper identified the poliomyelitis virus as the causative agent of acute paralysis that for so long had been crippling and debilitating people worldwide (Landsteiner and Popper 1909; Eggers 1999; Kristensson 1999). As the polio epidemic continued to be a scourge both in the United States and abroad, scientists focused on determining how the disease was transmitted. In the early 1940s, it was thought that poliovirus could only be transmitted via the person to person route, although several studies showed that the virus could remain infectious in domestic sewage effluent (Melnick et al. 1978). John Toomey and colleagues, after receiving a tip from the local chief of police in a small Ohio town, investigated an outbreak of gastrointestinal illness associated with wading and playing in a local creek. The wading incident occurred just prior to a large outbreak of polio in the community (Toomey 1945). Toomey directly fed creek water to rats in the laboratory, and subsequently, the rats became infected with poliovirus. This was the first demonstration of viral transmission via the water route; however, waterborne viral transmission did not gain considerable attention until an outbreak of hepatitis E occurred in New Delhi, India, between 1955 and 1956 (Dhopeshwarkar et al. 1957; Naidu and Viswanathan 1957; Viswanathan and Sidhu 1957). During this outbreak, it was demonstrated that the viral agent was being transmitted by water, resulting in nearly 30 000 infections and 73 deaths.

Figure 1.

History of virus discovery and environmental virology.

As more studies were conducted, it became clear that viruses including poliovirus, coxsackievirus and hepatitis virus could be found in and transmitted by water, thus posing a risk to human health. It was also obvious that better methods were needed to measure virus occurrence in water, virus fate and transport in different water environments and to quantify the risk from viruses in water. Some of the first studies that investigated the filtration of viruses from samples were based on the work of Elford (1931) in which viral adsorption to collodion membranes was investigated. The purpose of these early studies was to determine the size of virion particles (Ver et al. 1968); however, such studies were instrumental in providing the techniques that led to virus isolation from samples. As early as 1952, ion-exchange resins were being used to concentrate poliovirus (Lo and Berger 1952), and in 1953, Kelly was able to show that ion-exchange resins were successful at concentrating viruses from sewage samples. Chang was able to remove viruses from river water using flocculation procedures (Chang et al. 1958), which ultimately led to second-step concentration techniques for large-volume water sampling.

During the 1960s and 1970s, there was a greater focus on the development of methods for isolating and detecting viruses from water samples, thus expanding the field of environmental virology. In 1965, a symposium entitled ‘Transmission of Viruses by the Water Route’ was convened to discuss such issues as the significance of viruses in water, available methods for detecting viruses in water and how to protect public health from waterborne viral transmission (Berg 1967). In 1975, the World Health Organization formed a working group on Bacteriological and Virological Examination of Water, which promulgated methods for virus isolation from water samples (WHO 1979). A few years later in 1981, the 15th edition of Standard Methods for the Examination of Water and Wastewater was published and included methods for virus concentration and detection from water samples (Greenberg and Taras 1981). Concurrent with methods development, viruses such as Norwalk virus (Adler and Zickl 1969) and astrovirus (Madeley and Cosgrove 1975) were identified and linked to waterborne gastrointestinal disease outbreaks, thus intensifying the research efforts in environmental virology.

In July of 1997, the U.S. Environmental Protection Agency began an 18-month study as part of the Information Collection Requirement (ICR), a ruling which required the agency to collect data as part of a national research project to support development of national drinking water standards to protect human health (USEPA 1996). There were numerous water quality parameters collected during the ICR study, including data on virus occurrence in source and drinking waters (Fout et al. 1996). For viral monitoring, the ICR was the largest national occurrence study ever undertaken in the United States.

This review first describes the state of the science of viral concentration methods from water samples and then incorporates a meta-analysis of available performance data from these different methods. The methods included in this review have been used to concentrate several enteric viruses, including enteroviruses, noroviruses, astroviruses, rotaviruses, reoviruses, adenoviruses, hepatitis A virus and/or hepatitis E virus from volumes of water ranging from 100 ml up to 1900 l. By reviewing the performance methods currently available in the literature, and incorporating a meta-analysis of the available data, it suggests that more so than water type or method used, efficiency of virus recovery depends on the viral target. This will be an important consideration when collecting occurrence data from water sources.

Adsorption/Elution Methods

Some of the earliest methods in environmental virology involved adsorption of the viral particle to a surface and elution from that surface. The development of VIRADEL, or virus adsorption/elution, methods, is credited to Wallis and Melnick (1967a) and involves the adsorption of viral particles to the filter media by charge interaction and subsequent elution of the virus by a pH-adjusted solution. The water matrix, type of filter used to adsorb the virus and elution solutions vary. However, typically these methods are used for water samples (as opposed to sewage samples, which can clog the filters). The most common elution solution typically contains beef extract; however, solutions containing amino acids and/or salts have also been used to remove adsorbed virus from the filter (Farrah and Bitton 1979; Chang et al. 1981). Enteric viruses have considerable variation in the number of proteins present in their capsid, which then affects size and charge of the proteins that make up the capsid. Enteric viruses range from about 30 nm (enterovirus) to 100 nm (adenovirus) in diameter and have isoelectric points that vary from 2·8 for hepatitis A to about 8·0 for rotavirus (Michen and Graule 2010). Because viruses in water typically have a net negative surface charge, depending on the type of filter used, either the filters or the water sample has to be conditioned prior to filtration of the sample to allow adsorption. There are two basic filter types used to adsorb virus: electronegative filters and electropositive filters.

Electronegative filters

The first electronegative filters to be used for virus concentration were built upon studies using membrane filters to separate viruses from crude cell extracts (Wallis and Melnick 1967a,b,c). By adjusting salt concentration and pH, membrane filters were used to concentrate viruses from crude cell lysates (Wallis and Melnick 1967a; Ver et al. 1968). This approach was then applied to water samples to concentrate viruses using cartridge or flat membrane filters to allow for larger sampling volume. Using poliovirus, Farrah et al. (1976) showed that >90% of seeded virus were adsorbed to Filterite filters when the sample pH was 4 or less and that recoveries varied from 40 to 67% in seeded tap water samples. Gerba et al. (1978) showed that poliovirus could be recovered from seeded tap water samples at an average of 52%, while sea water recoveries averaged 53%. In later work, Shields et al. (1985) showed that poliovirus 1, coxsackievirus B3, coxsackievirus B4 and echovirus 1 averaged 43–54% recovery using negatively charged Filterite filters. In an effort to determine the effect of cationic polymers, Preston et al. (1988) used poliovirus 1, coxsackievirus B5, echovirus 1 and echovirus 5 and showed that 96% of viruses were adsorbed and 99% were recovered when cationic polymers were added to the filter membrane versus 20% adsorption and 18% recovery when filters were not treated.

Recently, Haramoto et al. (2009) have developed a method for using an electronegative filter in conjunction with aluminium or magnesium and have had success in concentrating and recovering human norovirus from 250 to 500 ml of MilliQ water (186%), tap water (80%), bottled water (167%), river water (15%) and pond water (39%). Victoria et al. (2009) evaluated negatively charged membranes and found that noroviruses and human astroviruses could be recovered at rates of 18 and 64% from mineral and river water, respectively, with recovery from tap and sea water from 3 to 14% for both virus types.

The benefits to using electronegative filters are relatively high recoveries for commonly tested enteroviruses, the apparent potential, as shown in the above-cited work, with some strains of norovirus, and its low price and wide availability (Table 1). The major disadvantage, however, is that conditioning a large-volume water sample may be cumbersome. Either the water sample must be pH-adjusted by injecting acid into the water flow prior to the sample going through the filter, which could be performed in the field but may be difficult to control, or the sample must be collected in the field and returned to the laboratory for filtration, which limits the volume of sample that can be taken, as most laboratories are not equipped to handle hundreds of litres of water. Preconditioning of electronegative filters, rather than the water sample, can be performed prior to sample collection. While a reasonable approach, the major disadvantage to this approach is the potential for more variability in filter surface charge, which adds another variable when measuring filter performance, and does not lend itself well to routine monitoring unless strict quality assurance measures are in place.

Table 1. Filter type pros and cons
Filter typeProsCons


Can filter large volumes even in more turbid waters before clogging occurs

Has been tested with an array of enteric viruses

High recoveries for commonly tested enteroviruses

Requires preconditioning of water sample or filter prior to filtration
Electropositive 1 MDS

No preconditioning of water sample required

Has been tested with an array of enteric viruses

Can filter large volumes even in more turbid waters before clogging occurs

Extremely high cost per filter
Electropositive NanoCeram®


Comparable recoveries to 1 MDS for viruses tested

No preconditioning of water sample required

Clogs in more turbid waters

Limited data available at this time to determine its effectiveness with multiple viral pathogens

Electropositive Glass Wool


No preconditioning of water sample required

Easy to use


Requires each laboratory to put filter apparatus together, which could cause interlaboratory variation

Turbid water may cause clogging

Limited data available at this time to determine its effectiveness with multiple viral pathogens

Electropositive ViroCap


No preconditioning of water sample required

Easy to use


Turbid water may cause clogging

May be limit to volume that can be filtered due to filter's size

Limited data available at this time to determine its effectiveness with multiple viral pathogens

Hollow-Fibre Ultrafiltration

Multi-pathogen concentration


No preconditioning of water sample required

Turbid water may cause clogging

Not easily field-deployable

Slow filtration rate

Limited data available at this time to determine its effectiveness with multiple viral pathogens

Electropositive filters

Whereas electronegative filters rely on the manipulation of the water sample to cause a net positive surface charge of a viral particle, electropositive filters work by relying on the innate negative charge of the viral particle in water and do not typically require preconditioning of a water sample prior to filtration. Designed in a cartridge-type format, the 1 MDS filter (Cuno, Meriden, CT, USA) has been a popular electropositive filter choice for many years in the environmental virology field. Early studies on poliovirus with the 1 MDS filter showed recoveries similar to that of the electronegative Filterite filter in spiked tap water samples (Sobsey and Glass 1980). Later studies demonstrated recoveries of 33% for echovirus 1 in tap water (Hill et al. 2009), up to 95% with seeded poliovirus in river water (Dahling 2002), and 67 and 36% for poliovirus seeded into 100 l of tap and river water, respectively (Karim et al. 2009). In an effort to offset the cost of the 1 MDS filter, it was shown that the 1 MDS filter could be washed and re-used (Cashdollar and Dahling 2006). For 1 MDS filters used 1–3 times, average recoveries ranged from 30 to 38% for tap water samples, and for those filters used 1–2 times, average recoveries ranged from 53 to 72% for river water samples (Cashdollar and Dahling 2006). In addition, an alternative positively charged filter was investigated, the N66 Posidyne, which was in the cartridge filter format. It was shown to average 87–100% recovery in river water when used 1–3 times. No cross-contamination of poliovirus from one use to the next was detectable by culture or RT-PCR following filter treatment between uses (Cashdollar and Dahling 2006).

The advantages to most electropositive filters are that they are easy to use, with no preconditioning of the water samples required, and because of the cartridge format, are able to filter large volumes of water (>1000 l) at high filtration rates without clogging in most cases (Table 1). The major disadvantage to the 1 MDS filter is the cost, at approximately $200–300 per filter, depending on the quantity ordered. The net result is that both routine monitoring and some research projects are too costly.

The NanoCeram® filter (Argonide, Sanford, FL, USA) is another electropositive filter that has recently come into the market. This filter comes in cartridge format or is also available in a self-contained cartridge format under the name Virocap (Scientific Methods, Inc., Granger, IN, USA). Unlike the 1 MDS filter, the NanoCeram® filter is much more economical, at approximately $40 per filter (Table 1). A recent report by Karim et al. (2009) reported that seeded poliovirus could be retained on the NanoCeram® filter at an average of 84%. Using optimized elution protocols, recoveries of 54% for poliovirus, 27% for coxsackievirus B5 and 32% for echovirus 7 were obtained from seeded tap water samples. Recoveries for enteroviruses and noroviruses from tap water were comparable with and in some cases better than the 1 MDS filter (Karim et al. 2009). Subsequently, Gibbons et al. (2010)reported recoveries of >96% for male-specific coliphages and noroviruses but <3% for adenoviruses when using the NanoCeram® filter to concentrate from sea water. Bennett et al. (2010) demonstrated recoveries of 65% from deionised water and 63% from sea water for MS2 using the ViroCap filter. In addition, poliovirus recoveries were demonstrated to be 37% from deionised water and 44% from sea water when using the Virocap filter. Most recently, Ikner et al. (2011) reported average recoveries of 45–56% for MS2 coliphage, 66% for poliovirus 1, 84% for echovirus 1, 77% for coxsackievirus B5 and 14% for adenovirus 2 from 20 l spiked tap water samples when using the NanoCeram® filter.

Further research into the ability of the NanoCeram® filter to concentrate other virus types in other water matrices is warranted. Its ease of use makes it an attractive choice for large-scale virus concentration as it requires no conditioning of water samples prior to filtration (Table 1). In addition, it is an attractive economic alternative to the popular 1 MDS electropositive filter, which has lost popularity over recent years due to its high cost.

In the early 1990s, glass wool became a popular tool for concentrating viruses from large volumes of water. The glass wool is packed into a column, and the column is then connected to the water source. Early studies reported poliovirus recoveries of 72 and 75% from drinking water and sea water, respectively (Vilagines et al. 1997). More recently, recoveries of 98% for poliovirus, 28% for adenovirus 41 and 30% for a norovirus GII strain have been reported from tap water by Lambertini et al. (2008). The main advantage of glass wool is that it is inexpensive, and because of its electropositive charge, its use requires no preconditioning of the water sample prior to filtration (Table 1). Because of its ease of use, glass wool is an attractive option for large-volume sampling and/or in-line continuous sampling. However, glass wool filters are currently packed by hand by the laboratory performing the testing rather than being commercially available in a column format, such as other filter types. This makes QA/QC difficult, as packing could vary from person to person, resulting in potentially large variations in filter performance.


Hollow-fibre ultrafiltration has gained popularity in the last decade as an attractive method for virus concentration from large volumes of water. The method of ultrafiltration relies on size exclusion and is conducted by passing the water sample through capillaries or hollow fibres or through flat sheets using tangential flow, all of which have nominal molecular weight cut-offs of 30–100 kDa. Because of the pore size, water and low molecular weight substances are allowed to pass through the fibres and into the filtrate, whereas larger substances, such as viruses and other micro-organisms, are trapped and retained in the retentate. Ultrafiltration can be dead end (single pass of the sample through the ultrafilter) or tangential flow (multiple circulations of the sample through the ultrafilter). Typically, the ultrafilter is then backwashed to remove any micro-organisms that are retained on the filter, and this backwash is combined with the retentate of the final sample for analysis.

In 100 l spiked groundwater samples, Olszewski et al. (2005) reported recoveries of 57–71% for bacteriophage T1, 70–74% for bacteriophage PP7 and 82–95% for poliovirus 2. In the same study, 100 l spiked surface water samples averaged recoveries of 70–123, 86–104 and 56–69% for bacteriophage T1, bacteriophage PP7 and poliovirus 1, respectively (Olszewski et al. 2005). Hill et al. (2009) reported recoveries of 58% for echovirus 1, 100% for bacteriophage MS2 and 77% for bacteriophage phi X174 from seeded tap water samples. Holowecky et al. (2009) performed a comparison of 5 different ultrafilter types in concentrating several micro-organisms, including the bacteriophages MS2 and phi X174, from spiked drinking water samples. For MS2, recoveries averaged 52–88% depending on the type of filter used and 55–95% for phi X174, depending on filter type. Overall, there was no statistical significance from one type of ultrafilter to the other (Holowecky et al. 2009). In another study, Gibson and Schwab (2011) demonstrated recoveries of 48 and 84% for bacteriophage MS2, 43 and 80% for bacteriophage PRD1 and 40 and 16% recoveries of poliovirus under high and low seed levels, respectively, in 100 l surface water samples. Murine norovirus recovery in 100 l surface water samples averaged 74% using low seed levels, and in 100 l drinking water samples, recoveries averaged 42% using high seed levels (Gibson and Schwab 2011). More recently, Rhodes et al. (2011) recovered 89 and 97% of poliovirus from 100 L of tap water under optimized filtration rate conditions, with spike titres of 103 and 10PFU 100 l−1, respectively. Recoveries of bacteriophage MS2 were up to 98% under low filtration rate conditions (Rhodes et al. 2011).

One major advantage to the use of hollow-fibre ultrafiltration is that the cost of the filters is much less than other sampling systems in the field, as hollow-fibre ultrafilters are made for use in dialysis treatment of medical patients (Rhodes et al. 2011). The major disadvantages to the use of ultrafilters are slow filtration rates, difficulty in field deployment and tendency to clog under conditions when particulates exceed 1·6 g l−1 concentration (Olszewski et al. 2005) (Table 1).

Meta-Analysis of Available Performance Data

Parameters for statistical analysis

These studies examined several different virus types and water types and diverse experimental conditions such as differently treated filters, an assortment of eluents, flow rates, volumes, concentration methods, detection assays, etc. Constrained by the number of studies that are available for any given combination of virus, water type and experimental condition, analysis was limited to poliovirus, norovirus and bacteriophage results from environmental (surface and sea water) and tap water samples, where electronegative versus electropositive filters (excluding ultrafiltration) are used. In particular, analysis by direct comparison of recoveries from different matrices using different filter types was important for understanding performance. However, the majority of the studies were concerned only with performance of a single filter type and do not afford direct comparison of tap water versus environmental samples, electropositive versus electronegative filters, nor sample volume. Thus, statistical analysis was limited to using only the mean recovery, rather than comparative differences, as effect size for a meta-analysis.

Few of the studies reported a standard deviation. Instead, ‘study quality’ is expressed in terms of sample size, and data are weighted by the number of samples evaluated within each study for the respective combination of virus, water type, filter type and volume. A random effects model is considered. Effects of different aspects of the experimental set-up among the various studies are deemed to represent random variability that can be expected to occur among different laboratories using their own unique set-ups.

Weighted analysis of variance was used to evaluate effects of filter type, water type, volume filtered and virus type on recovery efficiency as the effect size. Analysis was performed using R version 2·14 (R Foundation for Statistical Computing, Vienna, Austria). A critical (‘alpha’) level of 0·05 is used for evaluating differences among filter type (electronegative versus electropositive), matrix (environmental or tap water), sample volume (≥100, 10–99, or <10 l) or virus type (poliovirus, norovirus or bacteriophage).

Meta-analysis results and discussion

Water type, filter type and sample volume were found to have no significant effect on recovery (P > 0·05). Although the overall highest recovery was in distilled water, only three of the studies used distilled water as a test matrix (Haramoto et al. 2009; Bennett et al. 2010; Rhodes et al. 2011). The remaining water types tested by the studies included in this review had average recoveries that ranged from 38 to 49% (Fig. 2). This indicates water matrix effects may not be the most important factor in virus recovery from water samples.

Figure 2.

Mean recovery of viruses from water matrices. Error bars indicate% standard deviation. Laboratory water encompasses both Milli-Q and distilled water. Numbers on bars indicate the number of studies from this review that used the type of water matrix indicated.

Although recoveries from hollow-fibre ultrafiltration were higher than either electronegative or electropositive filters (Fig. 3), differences in recovery among filter types were not statistically significant (P > 0·05). There are fewer studies available for evaluating ultrafilter performance, while there is a larger number of studies available demonstrating electropositive and electronegative filter performance. Although ultrafiltration was used as early as 1977 for virus concentration (Fig. 1), only in the last 10 years have hollow-fibre ultrafilters been employed for sampling larger volumes of water for viruses (Rhodes et al. 2011). Compared with charged filters, there is still a large data gap in performance that needs to be filled, but if ultrafilters continue to perform well in recovering viruses, the added benefit of being a multi-pathogen concentrator (Hill et al. 2009; Holowecky et al. 2009; Gibson and Schwab 2011) makes them a very attractive choice for water sampling.

Figure 3.

Mean recoveries of viruses by filter type. Error bars indicate% standard deviation. Numbers on bars indicate the number of studies from this review that used the type of filter indicated.

Sample volume did not have a statistically significant effect on overall virus recovery (P > 0·05). For the studies included in this review, sample volumes of 100 l or greater averaged 45%; sample volumes of 10–99 l had the highest overall recovery at 64%, and sample volumes less than 10 l averaged 50% recovery (Fig. 4). In most of the studies included in this analysis, direct comparison of the effect of sample volume on recovery was not a consideration in the original study. This highlights a large data gap in understanding how sample volume affects recovery. Large-volume water samples are often required due to the fact that enteric viruses are typically at low levels in the environment (Gibson et al. 2012). Sample concentration methodologies can co-concentrate inhibitors that affect downstream analyses such as molecular methods and/or cell culture assays (Ijzerman et al. 1997; Gibson et al. 2012); the larger the sample volume, the potential for more inhibitors to be co-concentrated and therefore the more highly likely that recovery will be affected. While this meta-analysis indicates that sample volume may not be the most important factor in viral recovery, there is a strong need for data which includes the combined effects of different concentration methodologies (electropositive, electronegative or ultrafiltration), virus types and sample volumes on overall recovery.

Figure 4.

Mean recoveries of viruses by volume sampled. Error bars indicate% standard deviation. Numbers on bars indicate the number of studies from this review that used the volume indicated.

Regardless of filter type, water matrix or sample volume, recoveries were statistically significant between poliovirus and norovirus (P = 0·0063), with poliovirus averaging overall higher recoveries than norovirus among the studies examined (Figs 5 and 6). Adenovirus and astrovirus recoveries were also low in the studies included in this review, with adenovirus averaging 13% recovery between three studies (Lambertini et al. 2008; Gibbons et al. 2010; Ikner et al. 2011) and astrovirus averaging 19% recovery from a single study (Victoria et al. 2009). Due to a low number of replicates in these reported studies, adenovirus and astrovirus were excluded from analysis between virus types. However, the overall lower recovery from these virus types is important to note and could be due to inherent properties of the viruses themselves, such as isoelectric point (Michen and Graule 2010), structural differences and/or virion size (Kidd et al. 1993; Yeh et al. 1994). At this time, there are no studies in the literature describing the recovery of noroviruses using hollow-fibre ultrafiltration, with the exception of Gibson and Schwab's study using murine norovirus (Gibson and Schwab 2011). While the Gibson study showed recoveries of 74% from surface water, this was a single replicate sample, and drinking water samples averaged recoveries of 42% in triplicate experiments (Gibson and Schwab 2011), which are both higher than the overall average recovery for noroviruses in our review and analysis of performance data.

Figure 5.

Poliovirus recoveries from spiked water samples. Poliovirus recovery efficiencies from tap and laboratory water (white), and environmental (shaded) and sewage (black) samples are indicated by filter type as determined from papers used in this survey. Results using electronegative filters are indicated by open circles, those using electropositive filters by circles filled with crosshairs and those from ultrafiltration by diamonds. Laboratory water samples, distinguished from tap water by ‘L’, comprise both distilled and Milli-Q water. Environmental samples comprise surface water samples except where indicated by ‘gw’ for ground water or ‘sea’ for sea water. The area size of the symbol is proportional to the ‘quality’ of the respective study, taken as the number of samples used in the evaluation. Numbers correspond to references: 1-Bennett et al. (2010), 2-Cashdollar and Dahling (2006), 3-Dahling (2002), 4-Farrah et al. (1976), 5-Gerba et al. (1978), 6-Gibson and Schwab (2011), 7-Haramoto et al. (2009), 8-Ikner et al. (2011), 9-Karim et al. (2009), 10-Lambertini et al. (2008), 11-Olszewski et al. (2005), 12-Preston et al. (1988), 13-Rhodes et al. (2011), 14-Shields et al. (1985), 15-Sobsey and Glass (1980), 16-Victoria et al. (2009).

Figure 6.

Norovirus recoveries from spiked water samples. Norovirus recovery efficiencies from tap and laboratory water (white), and environmental (shaded) samples are indicated by filter type as determined from papers used in this survey. Results using electronegative filters are indicated by open circles, those using electropositive filters by circles filled with crosshairs and those from ultrafiltration by diamonds. Laboratory water samples, distinguished from tap water by ‘L’, comprise both distilled and Milli-Q water. Environmental samples comprise surface water samples except where indicated by ‘gw’ for ground water or ‘sea’ for sea water. The area size of the symbol is proportional to the ‘quality’ of the respective study, taken as the number of samples used in the evaluation. Numbers correspond to references: 1-Gibbons et al. (2010), 2-Gibson and Schwab (2011), 3-Haramoto et al. (2009), 4-Karim et al. (2009), 5-Lambertini et al. (2008), 6-Victoria et al. (2009).

Future Directions in Concentration of Water Samples for Virus Detection

Taken together, the data evaluated in this review suggest that virus target, rather than water matrix, filter type or sample volume, is more important in predicting the performance of a given method. While hollow-fibre ultrafiltration provided the highest average recoveries as compared to electropositive and electronegative filters (Fig. 3), more research is warranted, particularly in connection with norovirus, which had statistically lower recovery over all studies than poliovirus. Ultrafiltration might prove to result in higher recoveries of many virus types, as it is based on size exclusion rather than charge interaction. As isoelectric points are so variable not only between types of viruses, but even within different strains of the same viral type (Michen and Graule 2010), methods that utilize properties other than charge might prove to be more reliable in future virus concentration methodology.

Variations in recovery that can obscure the effects of primary concentration by charged filters or hollow-fibre ultrafiltration may be due to secondary concentration procedures and downstream detection assays. Secondary concentration procedures further concentrate a sample down to a more manageable volume for detection assays. There are various secondary concentration procedures that can be employed, ranging from organic flocculation (Sobsey and Glass 1980; Cashdollar and Dahling 2006) and celite (Dahling and Wright 1986; McMinn et al. 2012) to polyethylene glycol (Lambertini et al. 2008), and from various ultrafilters (Hill et al. 2009; Victoria et al. 2009; Ikner et al. 2011) to centrifugation protocols (Schultz et al. 2011). Such procedures should be validated separately from the primary concentration method so as to determine recovery variability at these steps. Additionally, viral recovery can be affected by downstream detection approaches. Detection assays such as RT-PCR, PCR, RT-qPCR, qPCR, cell culture, immunological-based assays, microscopy or direct sequencing reactions may be employed. However, the downstream application may play an important role in the decision as to which upstream filtration method should be used, as some types of filtration methods also concentrate inhibitors of molecular assays as well as compounds that can be toxic to cell culture assays. For this reason, the use of appropriate controls for the downstream assay being employed is essential for understanding and interpreting resulting data. Quality controls such as the use of internal controls (Parshionikar et al. 2004; Bosch et al. 2011; D'Agostino et al. 2011), matrix spikes (Fout et al. 2010) and/or inhibition assays (Fout et al. 2010; Gibson et al. 2012) can help to clarify any issues related to inhibition of molecular assays and/or problems associated with cell culture assays.

As existing and new technologies are further developed, it will be important to provide method comparisons, multi-laboratory validations and adequate performance and quality control measures to facilitate reliable interpretation of data. Of the numerous occurrence studies reported in the literature, many lack this type of validation. The environmental virology community should work towards a coordinated approach for analysing and reporting method performance, including the use of common quality control processes, so that resultant data are both reliable and meaningful for measuring public health risk from waterborne viruses. Results of the meta-analysis presented here emphasize the potential value of selecting a standard set of viral targets for effective assessment and comparison of existing and newly developed methods.


We would like to thank Nichole Brinkman, Dr. G. Shay Fout, Dr. Jay Garland, Brian R. McMinn and Dr. Eric R. Rhodes for critical review of this manuscript. A special thanks to Dr. Ann C. Grimm for critical review of this manuscript and numerous helpful suggestions throughout the writing process.

Conflict of Interests

The authors report no conflict of interest in this work.


The United States Environmental Protection Agency through its Office of Research and Development funded and managed the research described here. It has been subjected to the Agency's administrative review and approved for publication.