By Jim Rosenthal, Published in Spring Issue of Air Media
Does a Recent Analysis of BASE Data Really Show a Connection Between Synthetic Air Filters and (BRS) or Sick Building Syndrome?
A number of recent studies have shown that the products of ozone reactions can impact human health and perceived indoor air quality. An analysis was undertaken by researchers at the Lawrence Berkeley National Laboratory (LBNL) to determine if data compiled in the U.S. EPA Building Assessment Survey and Evaluation (BASE) study could be used to find out if there is an association between outdoor ozone concentrations and increased reporting of Building Related Symptoms (BRS) by occupants. BRS is also known as sick building syndrome.
The BASE study was conducted in 1994-1998 and included 100 commercial buildings. The BASE study compiled extensive information on environmental measures (including temperature, humidity, CO, CO2, sound, light, particles, VOCs, formaldehyde, biologicals), building characteristics (including use, occupancy, physical location, ventilation equipment, smoking policy, pest control, cleaning practices, etc.), HVAC characteristics (including filtration, air cleaning, humidification, maintenance, supply air flow, outdoor air flow, temperature, etc.) and occupant perceptions (work place characteristics, health and well-being, work place environmental conditions, job characteristics, etc.).
The LBNL researchers analyzed the BASE data to find associations between: 1. Outdoor ozone and building-related symptoms, 2. Outdoor ozone and indoor chemical reactions and 3. Outdoor ozone, ventilation system filter media and building related symptoms. BRS were defined as symptoms that occurred at least one day per week during the four weeks prior to answering the questionnaire and the symptoms improved when away from work. Indoor ozone measurements were not taken in the BASE study. Outdoor ozone levels for the LBNL analysis were obtained from the EPA who analyzed historical records of ambient air quality monitoring stations nearest the study buildings.
The results of these analyses were presented in two papers LBNL-62419: “Outdoor Ozone and Building Related Symptoms in the BASE Study” (by M. Apte, I. Buchanan and M. Mendell) and LBNL-62508: “Air Filter Materials, Outdoor Ozone and Building-Related Symptoms in the BASE Study” (by I. Buchanan, M. Mendell, A. Mirer and M. Apte). Both papers eventually were published in the Journal Indoor Air.
Of particular interest to us is the second analysis on outdoor ozone, air filter materials and BRS. This paper was also presented at the National Air Filtration Association (NAFA) Technical Conference in Memphis in 2008. In this analysis the team found a correlation between high outdoor ozone and the simultaneous use of “polyester-synthetic” filters with higher incidents of perceived problems of higher and lower respiratory symptoms, cough, eye irritation, fatigue and headache. The report goes on to state that “these findings suggest possible health consequences from chemical interactions between outdoor ozone and polyester-synthetic filters in buildings. Attributable risk proportion (ARP) estimates indicate that removing both risk factors might, given certain assumptions, reduce BRS by 26%-62%.”
In an article appearing in the January 28th Issue of Scientific American, the lead author, Michael Apte “speculated that symptoms are due to unstable ozone molecules chemically interacting with the wide range of materials found in an office building, beginning with the polyester filters.” Apte goes on to state: “Glass is a really inert material. On the other hand, polyester is a polymer and it’s got lots of bonds in there that are capable of being broken up by ozone.”
But, does the BASE data really suggest this conclusion? Could it be that the correlations between higher ozone levels combined with polyester-synthetic filters and BRS symptoms do not imply causality? Analysis of the assumptions made in the LBNL paper, further information about the air filters used during the time period of the BASE study (1994-1998), and other published peer-reviewed studies of the reactions between ozone and air filter materials suggest that the conclusions reached by the LBNL team are suspect.
A great deal of research has been done on the reactions of ozone with dirty air filters. The major effects include the creation of formaldehyde and other carbonyls, formic acid and ultrafine particles (Beko et al, 2006; Hyttinen et al., 2006). Hyttinen et al. (2003) reported increased intensity of odors downstream of filters that had been exposed to ozone. It was also found that air from dirty filters exposed to ozone degraded perceived air quality (Beko et al.;2006 and Clausen et. al.;2002), increased sick building syndrome symptoms (Clausen et al., 2002) and decreased performance (Wargocki et al., 2004). It has also been observed that when ozone is involved in creating reactions with filter materials and with the materials deposited on the surface of filters that ozone levels are decreased (Zhao et al., 2007). In effect, the ozone is consumed in the chemical reactions. There is also a direct correlation between the type and amount of soil on the surface of the filter with the reductions in ozone and the production of byproducts (Zhao et al., 2007). This body of research leads to the conclusion that in order to maintain good indoor air quality additional attention needs to be paid to changing dirty air filters more frequently during times of high outdoor ozone. On the other hand, the LBNL analysis suggests that it is not the soil on the filter that is causing the detrimental ozone reactions but rather the material used to make the filters.
To support this conclusion the LBNL analysts needed to show that the polyester-synthetic and fiberglass filters were of the same relative efficiency. Otherwise it would not be the filter type but rather the differences in the amount of dirt on the filter that could have caused the increase in BRS i.e., the higher the efficiency, the more dirt and thus the greater likelihood of ozone reactions.
The problem with using the BASE data for this analysis is that the details about the filters used in the buildings are not very complete and are somewhat dated. According to the EPA, filtration efficiencies for the BASE study were recorded using average dust spot efficiencies “determined from manufacturer’s specifications.” Values were reported as percentages “in accordance with ASHRAE 52-76 (sic).” In the LBNL study it states that the researchers “created a two-part variable for the overall filtration efficiency of all the filters in each study space. Filtration efficiency was measured on the “minimum efficiency reporting value” scale (MERV) as ‘less efficient filters’ with MERV<=7 vs. 'more efficient filters' with MERV>7.” ASHRAE 52.2 (which is the test used to determine MERV numbers) was not established until 1999 and the data was collected between 1994 and 1998. Some type of conversion from average dust spot efficiencies to MERV numbers had to have been done by the researchers to make the data fit this variable. It is not clear from the study what methodology was used for this conversion. There is not a reliable way of doing exact conversions since average dust spot efficiencies and MERV numbers are derived by two different test protocols. Therefore, conversions are estimates and should be expressed in ranges. For example in the NAFA Guide to Air Filtration, Fourth Edition 2007 pleated panel filters are shown to have an average dust spot efficiency of 20%-30% and a MERV of 6-8. It is very possible that a filter with a dust spot efficiency of 25% when tested according to the testing procedures in ASHRAE 52.1(which was the test standard in 1994) would have a MERV of either 6, 7 or 8 when tested accordance with ASHRAE 52.2. Consequently, the use of a two-part variable with just one MERV number seperating a “less efficient filter” from a “more efficient filter” is questionable and could lead to misclassification of the filters in the study.
It is difficult to understand the rationale for the construction of the two-part variable used in the LBNL analysis. Other than the fact that MERV 7 is the median of the MERV numbers used in the study, there is not a good reason to select this as the break point between a “less efficient” and a “more efficient” filter. Filters of a MERV 6 and above are almost always a denser media. Pleated filters of MERV 6, MERV 7 and MERV 8 would act in approximately the same way in terms of capturing dirt that could react with ozone. On the other hand, filters that are MERV 1-4 are so inefficient that they cannot be tested using the standard ASHRAE 52.2 protocol (to determine efficiency by particle size). ASHRAE test dust comprising linters, carbon black and Arizona Road Dust is used. This test for low efficiency filters produces an arrestance number – which is the percentage of dust collected by the filter by weight. In the filter industry these filters are known as “throw-aways” since they are inexpensive and relatively inefficient. A survey of filter industry executives and distributors with at least 15 years of experience estimated that at least 70% of these throw-away MERV 1-4 filters sold during 1994-1998 were fiberglass.(A microscopic view of the structure of a fiberglass filter is shown in the picture above. As an indication of the size of the holes in the filter the word “the”- in 12 point type – can be seen through the filter.) They capture far fewer particles that could react with ozone. In addition, they capture primarily larger particles that would have less surface area with which to react.
The survey participants also agreed that at least 80% of the pleated filters sold during this period were constructed using cotton-polyester media (which are classified in the LBNL analysis as polyester-synthetic filters). Less than 1% were fiberglass. High efficiency filters (with dust spot efficiencies between 60% and 95%) represented 23% of the filters in the BASE study. It was estimated that about 40% of those would be fiberglass. There were no HEPA filters found in the BASE study buildings. Therefore, it is likely that the largest number of fiberglass filters reported in the BASE study would be in the MERV 1-4 category.
Consequently, it is possible that the correlation that was found between filter media and BRS could be explained by differences in filter efficiencies rather than differences in filter materials. In fact, had the “filter efficiency variable” break point been set by the LBNL analysts at MERV 5 instead of at MERV 7, it is likely that the best correlation would have been with filter efficiency and not filter media.
Even with the MERV classification system used in the study, the LBNL researchers found that there WAS a correlation between the use of higher efficiency filters and higher levels of perceived BRS. This correlation was explained in the discussion of the study in the following way: “Contrary to expectations, occupants in the buildings with the more efficient air filters (MERV >7) had increased odds of some BRS (cough, eye, possibly skin and LR) relative to occupants in buildings with less efficient filtration (MERV <=7). More efficient filters remove from the indoor air a greater proportion of the smallest airborne particles, which are known to cause a variety of health problems from epidemiologic studies of ambient air pollutants. One possible explanation, which needs further investigation, is that higher efficiency filters provide greater surface area on which ozone initiated surface-reactions can occur, and thus increased release into the indoor environment of irritating compounds produced from such reactions." What is curious about this discussion is that the correlation between more dirt on the more efficient filters and lower perceived air quality is not "contrary to expectations." It is exactly what we would expect based on prior research. In addition, while the higher efficiency filters would remove more particles, this does not negate the fact that during periods of high outdoor ozone there could be a perception of lower indoor air quality caused by the byproducts of the ozone reactions with materials on the surface of the filters. The LBNL findings are based on the premise that there must be some type of reaction between the ozone and the synthetic materials used in the synthetic filters that contributes to the increases in BRS. The presumption is that these reactions do not take place with fiberglass filters. Other peer reviewed research does not support this premise. Zhao et al. in their study of ozone removal by HVAC filters found in their tests of clean filters that two of the synthetic filters did not remove any ozone, three from a different manufacturer removed an average of 8% of the ozone and three fiberglass filters removed 3% of the ozone. Contrast this with dirty commercial filters that removed anywhere from 40% to 100% of the ozone in the study. These findings support the conclusion that the vast majority of the reactions and the resulting byproducts come from dirty filters and the composition of the dirt itself rather than from the filter materials. The findings in the Zhao et al. study also show that there is a difference in the reactivity of new synthetic filter media to ozone. While one synthetic filter media had greater reactivity with ozone than the fiberglass filters, the second type of synthetic filter media had NO reaction with ozone. The conclusions in the LBNL analysis tend to paint all "polyester-synthetic" filters with the same broad brush. These findings show that this is not the case. In fact, at least one of the largest manufacturers of synthetic filter media today was not producing media in 1998. Filters from the media of this manufacturer could not have been in the BASE data used in the LBNL analysis. Similar tests for ozone reactions with new filters were reported in Hyttinen et al. (2006). In this study the researchers found that an unused polyester pre-filter did not remove ozone while an unused fiberglass filter removed 6% of the ozone and another type of unused fiberglass filter removed 19% of the ozone. This led the researchers to conclude that "these findings show that ozone removal properties of new filters vary by manufacturer and filter type." Beko et al. (2007) analyzed ozone reactions with new fiberglass media used to make a bag filter. They found that the new fiberglass filter media initially removed 45% of the ozone from the airstream; after one hour of operation it still removed 25% of the ozone and after allowing time in a neutral environment it regained its original ozone removal efficiency. The researchers explain this high ozone removal rate of a fiberglass filter by stating: "the ozone likely reacts with organic compounds remaining on the surface after the manufacturing process (e.g., tackifiers, binders, resins). Such compounds also appear to be present within the filter material, since the unused filter can regain its ability to remove ozone when left in a static environment." These findings do not support the proposition that fiberglass filters are "inert" and polyester-synthetic filters have "lots of bonds in there that are capable of being broken up by ozone." In fact, what we see is that test data from numerous researchers shows that the media in air filters vary considerably by manufacturer and that there is little reason to assume that all filters of a given type (ie. fiberglass or polyester-synthetic) would react in the same way. Indeed the research shows that fiberglass filters can produce more byproducts than synthetic filters. Because of deficiencies in the BASE data, the construction of the filter efficiency variable, other research into ozone reactions with dirt on filters and test results on ozone reactions with filter media, it is difficult to support the conclusions in the LBNL paper that synthetic filters are the possible culprit in increased building related symptoms. The correlation between synthetic-polyester filters and high outdoor ozone levels with high reported BRS could be coincidental and not causal. A different analysis design for the same BASE data could have produced different results and the conclusion that it is the reaction of ozone with dirt on filters that can lead to decreased indoor air quality and a rise in perceived building related symptoms. Therefore, the solution to the problem of increased BRS on high ozone days may not be a change in filter materials but rather a change in filter replacement cycles.
Beko, G., Halas, O.,Clausen, G. Weschler, C.J., 2006. Initial studies of oxidation processes on filter surfaces and their impact on perceived air quality. Indoor Air 16, 56-64.
Beko, G., Clausen, G., Weschler, C.J., 2007. Further studies of oxidation processes on filter surfaces: Evidence for oxidation products and the influence of time in service. Atmospheric Environment 41, 5202-5212.
Buchanan, I., Mendell, M., Mirer, A., Apte, M., 2008, Air filter materials, outdoor ozone and building-related symptoms in the BASE study, Indoor Air, 18, 2, 144-155
Clausen, G., Alm, O., Fanger, P.O., 2002, The impact of air pollution from used ventilation filters on human comfort and health. In: Proceedings of the Ninth Annual International Conference on Indoor Air Quality and Climate, Indoor Air 2002, Monterey, USA, vol. 1, pp. 338-343.
Hyttinen, M., Pasanen, P., Salo, J., Bjorkroth, M., Vartianen, M., Kalliokoski, P., 2003, Reactions of ozone on ventilation filters. Indoor and Built Environment 12, 151-158.
Hyttinen, M., Pasanen, P., Salo, J., Bjorkroth, M., Vartianen, M., Kalliokoski, P., 2006. Removal of ozone on clean, dusty and sooty supply air filters. Atmospheric Environment 40, 315-325.
United States Environmental Protection Agency website, www.epa.gov/iaq/base/summarized_data.html, Building Assessment Survey and Evaluation (BASE) Study – Summarize Data, Part 2. Test Space HVAC Characteristics
Wargocki, P., Wyon, D.P., Fanger, P.O., 2004. The performance and subjective responses of call-centre operators with new and used supply air filters at two outdoor air supply rates. Indoor Air 14, 7-16.
Zhao, P., Siegel, J.A., Corsi, R.L., 2007, Ozone removal by HVAC filters. Atmospheric Environment 41, 3151-3160.