A Response to the “Risk/Benefit” Analysis for Aerial Pesticide Release to Abate the Vectors of West Nile Virus

Jim Northup

    In their justification for the aerial distribution of pyrethrum and piperonyl butoxide the Sacramento-Yolo Mosquito Vector Control District has referred to publications by Peterson assessing the health risks (Peterson, Macedo and Davis 2005) and Kramer assessing the spray efficacy (Kramer 2005, California Department of Health Services). A careful review of this material gives rise to a number of questions and some clear refutations of the implicit assumptions in these analyses.

Unsupportable Assumptions:  There are a number of unsupportable assumptions in the health risk assessment. First, it postulates the perfect performance of the spray apparatus at distributing uniformly sized droplets with uniform concentrations of insecticidal materials. There is no consideration given to the possibility of failure in the agitator, nozzle, or computer controlled spectroscopic sensors. Any combination of failures in these components could lead to the distribution of highly concentrated pyrethrum or piperonyl butoxide. No assessment is made concerning the probabilities of such events in spray operations, though they surely do occur.

     However, even given the techno chauvinism of such an assumption about perfect performance of human artifacts, there is a great error in assuming the droplets will maintain uniformity in size, concentration, or spatial distribution once released. The certainty is that air movement created by the aircraft itself will disrupt the “perfect” distribution. Some droplets will collide and aggregate. There will be differential evaporation due to the differing size of the aggregating versus dispersing droplets, leading to differing concentrations of pyrethrum. The smaller concentrated droplets will have the opportunity to collide and aggregate, leading to larger, more concentrated droplets. In short, the material distributed from the airplane will vary greatly in dosage by the time it reaches the ground.

No Tests of Aerial Spray.  Ignoring these errant assumptions on the modeling of the SYMVCD spray protocol, the Peterson paper does not appropriately review the health risk of Sac-Yolo’s aerial application. Instead, it reviews the health effects of pyrethrum and malathion formulations being distributed by ULV (ultra low volume) ground spray applications and not aerial distribution. Even though SYMVCD and CDHS officials have declared to us that this Peterson study proves the safety of aerial spray over residential neighborhoods, the study doesn’t even test the question of aerial spray. Instead it makes a short caveat relative to materials and methods that modeling the problem dismissed the need for testing. It makes the same assumption about testing for sub-chronic exposures from persistent residues either for humans or the environment. This assumption has already been proved false by the data from creek bed surveys. While the Peterson study posits no environmental impact from persistent residues, the aerial releases over Sacramento are noted to have doubled the toxicities of the extant pyrethroid residues in the creek beds of Sacramento County (Weston, D.P. et al 2006) (“Spray adds creek effect,” Matt Weiser, Sacramento Bee, Saturday, July 29, 2006).

Inhalation Risks Ignored.  When accurate scientific assessments are performed about the human health risks from aerial ULV releases of pyrethrum, pyrethroids, and PBO over residential neighborhoods, the “modeling” mentioned will be shown to be equally invalid as the environmental assessment. Here the notion that the degradation of these larger molecules into their daughter species will have been completed before they can hit the ground does not dismiss the possibility of inhalation or contact exposures with toxins. Quite the contrary, we are advised by CDHS’s former chief for air quality risk assessment that this simply increases the duration of inhalation and contact exposures to these compounds and their byproducts. For example, many of the byproducts are more aggravating to pulmonary disease than the pyrethrum or PBO. It was noted to us by Rick Roush (Rick Roush, Vicki Kramer, Helen Lawler and I participated in a round table discussion of the WNV spray program, “The Inocculated Mind,” aired live on KDRT Aug. 12, 2006) that the release could be analogized to volatilizing alkanes into the air. We would agree and extend the analogy more accurately to be akin to volatilizing gasoline, mothballs, or creosote. Rather than minimizing exposures to negligible, it maximizes the time and areas of exposure. The aerial release really must be tested and not just dismissed by caveat.

    There are significant differences in the toxicities of oral, dermal, or inhalation exposures. Inhalation is the most critical route for exposure. Focused ground spray minimizes such exposure while aerial release maximizes the risk. Therefore studying ground spray cannot elucidate the risks of aerial spray and is irrelevant to the question of safety of aerial spray. Even though the Latin Square design to the Peterson assessment of inhalation exposure for adults and children is most elegant, there can be no extrapolation of the data to the question of an aerial release because the sensors weren’t changed at regular time intervals but instead only cumulative exposures were measured. This masks from analysis any view of the possibility of differential concentrations being released over different areas from the aircraft spray apparatus at different times during the spray event. This possibility of an errant release of large doses of the material over a small area actually represents the most serious possibility of acute exposure.

Various At-risk Groups Ignored.  Additionally the Peterson report gives no consideration of sub-acute exposure to the “at-risk” groups such as neonates (children less than six months old), the liver-impaired, or those with severe respiratory ailment. The reason that pyrethrum is only a low-level poison to humans is the ability of healthy, adult persons to detoxify the poison faster than it can be assimilated through the skin or gut. Neonates, who lack the full compliment of liver enzymes, however, are appreciably more at risk, as are those with liver disease. On top of that, piperonyl butoxide inhibits the liver’s production of detoxification enzymes, so the liver-impaired are doubly at risk from this formulation. Cumulative exposure to piperonyl butoxide will also enhance the toxicities of a number of other insecticides that are tolerated at low levels in our foods. Peterson doesn’t mention these synergistic phenomena.

    Another sub-acute effect of the pyrethrum-piperonyl butoxide formulation is the induction of asthma. One must note that this region is already experiencing an asthma epidemic due principally to air quality issues. Even though this at-risk group exceeds 100,000 in Sacramento County, there is no mention of asthma or the effects on those with chronic pulmonary disease in the Peterson study.

Deficient Review of Environmental Risk.  The review of environmental risk is equally deficient. At best this issue is simply dismissed by caveat with the assumption of negligible impact. In fact, formulations of pyrethrum and piperonyl butoxide sufficient to kill mosquitoes will have an impact on all arthropods in the same size class as mosquitoes. There are hundreds of non-target species that would be impacted by any spray protocol that succeeded at killing mosquitoes. This would include, for example, the larval stage of spiders that might otherwise prey upon mosquitoes.

     The Peterson report first claims to give consideration to persistence of residues in the environment by citing the EPA’s failure to consider the question, then goes on to dismiss it as negligible. This claim of no accumulation or persistence in the environment has already proved false in creek bed surveys in the spray area. Similarly, the lack of concern for persistent sub-acute exposure is also invalid.

Regional Differences Ignored.  The Peterson document is analogously invalid in the assessment of risks from the West Nile Virus “epidemic”. The risk of WNV exposure posited in the Peterson study was drawn from an outbreak in a humid, temperate forest environment. The dynamics and composition of the mosquito populations of a Northeast forest are nothing like the Sacramento Valley, to wit the symptomatic infection rate in that location was ten times greater than the one observed here in semi-arid California. No locality in California, spray or not, experienced the 100/100,000 “epidemic” of symptomatic cases referred to by this article, as expected. Sacramento ended up fourth highest with greater than 14/100,000. Adjacent Yolo and San Joaquin without the benefit of spray, ended up with less than 6/100,000. The stipulated risk in Peterson is clearly contradicted by the facts of Sac-Yolo.

No Demonstration of Spray Benefit.  None of these risk assessments means anything in the absence of any demonstrable benefit from the “cure”. The “proof” of efficacy rests in a table published by the Health Services Department (Kramer 2005, California Department of Health Services). On examining the Kramer table the only conclusion that can be drawn is that the author or editors failed to check the sums before publication. This table, which colorfully exhibits the enumeration of symptomatic WNV cases in two-week segments for four areas of Sacramento County as well as the rest of California, fails to correspond accurately between the listed totals and the sum of elements. While the total for Sacramento is listed as 154, the four regions listed for Sacramento County sum to only 110. The actual distribution of these missing 44 cases in the table may or may not change the conclusions drawn, but in their absence no valid conclusion whatsoever can be drawn.

Insufficient Measurement.  There are a number of underlying problems with the CDHS analysis of the efficacy of aerial insecticide applications over Sacramento County in 2005. The most important problems have to do with the insufficiency of the measurements.  Since as many as 97% of the cases of West Nile Virus infection in people are cryptic, showing either no disease or only mild and not noteworthy disease symptoms, relying upon their visits to a physician to detect the presence of transmission of the virus yields about a 3% effectiveness. And since the rates of infection in people are never exceedingly high there is little statistical significance to be had from this way of sampling. A proper scientific sampling would utilize serological assays from a random sampling of the population at regular time intervals.

Transmission Timeline Uncertainties.  Additionally, relying upon this type of sampling introduces the same vagaries into the precision of the timeline for transmission of the virus. Most of the 154 patients in Sacramento County with a stipulated onset date for WNV infection in 2005 didn't actually visit a physician for more than two weeks and some as long as two months after their symptoms had begun. The onset dates were then retrospectively constructed from the memory of the patients.

Inaccurate Transmission Assumptions.  The Health Services study also did not measure the location of any of these transmissions. Instead it assumed a uniform mapping of the probabilities of transmission through time and space, a status of epidemic transmission known as holoendemic, and treated the residential address as the location of transmission. If the probabilities of acquiring the infection were equal at all places at all times during the period reviewed in this table, this would be an acceptable assumption. Or if the disease were known to be transmitted principally inside the home this would be reasonable. But this assumption is not at all legitimate in the modeling of any mosquito-borne epidemic whether it is Yellow Fever, Malaria, or WNV when it is first introduced in a new environment. The actual mapping of transmission of newly introduced pathogens is characterized by a decidedly focalized distribution with certain epidemic foci accounting for the majority of transmission. The transmission does not achieve any degree of uniformity in the mapping until considerably after the pathogen has been introduced, if ever. With seasonally cyclic mosquito populations such as we have here it is not likely that an arbovirus like West Nile, for which humans are a dead-end host, could ever be construed as holoendemic. But rather than making assumptions about the question, there is excellent research data available from William Reisen that thoroughly refutes the assumption of holoendemicity for the transmission of West Nile disease to humans in California (Reisen, W.K. et al 2006). He maps six foci that account for 83% of transmission to people (Kramer, V. et al 2005) (Reisen, W.K. et al 2004). The distribution of transmission isn’t an academic or theoretical question here, and without including this erroneous assumption about the equivalence of residential address and the location of transmission the CDHS “study” can’t prove or correlate anything because it didn’t bother to assess the location of transmission for any of the cases.

Flawed Zone Analysis.  In addition to these difficulties of hypothesis formation and data gathering, the CDHS “study” has major flaws in its analysis. Most serious among these is the assignment of control and treatment zones. The CDHS left out of their study areas that were treated both north and south of the American River around the commercial center of Sacramento. It seems inexplicable not to evaluate the entirety of the treatment zones. Equally inexplicable is that the “control” areas had nearly twice the rate of infection in people prior to the spray date. This doesn't represent anything like a good control since it is entirely understandable that the areas with a higher rate of infection will take longer to recede than those with a lower rate. At any rate the evaluation excludes 44 of the cases with a known onset date of the disease. This also seems very strange. Why wasn't the data from the entirety of the zones that were sprayed included in the table? Why wasn't the distribution of all known cases accounted for in the table? Why were some areas outside of Sacramento County included in the zones called control?

Unrealistic Data.  Perhaps the most remarkable thing about this analysis is the appearance of such perfect control of the transmission by the treatment. These data look too good to be real! The issue here is that the data suggest that there was no migration of any infected mosquito or bird from the control zones to the adjacent treatment zones. In the case of the northern zone no migration occurred for more than two months. Neither did a single person residing in the treatment zones acquire an infection in the disease-laden control areas. Given the CDHS’s assumption of holoendemicity this seems most remarkable indeed.

Absurd Assumptions.  In examining corroboration for this table from the Health Services web page private citizens cannot discover where the errors occur, since the location of cases is listed by county only, not by residential address. However, even if the missing cases were discovered to be distributed only in “control” regions, the assumption that persons contracted their infections exclusively in their places of residence is patently absurd.  Most likely the majority of infections were contracted exterior to the domicile at some outdoor activity.  In absence of random serological assays, which would be essential in a proper scientific study, this report could have been made more credible with careful landscape epidemiology (detailed interviews with victims about their activities).  Without any attempt at such analysis and with the assumption that all transmission was in or near the home, the table moves from inaccurate to meaningless.

Spray-Independent Decline.  What can be seen in the Health Services “line list” (CDHS 2006) is that the peak expression of symptomatic cases for Sacramento County was 15 new cases on August 1. This number was declining to 8 new cases on August 6 and 7 and by August 8, the day the spray began, this number had declined to 4 new cases or a two-thirds decline in advance of the aerial spray program. Since the virus has an incubation of 3-14 days prior to expression of symptoms, the first possible reduction in symptomatic cases due to aerial spray couldn’t have occurred until Aug. 11. Apparently the Health Services believe that the threat to utilize aircraft frightened the mosquitoes into surrender prior to the spray! Viewing the data for Riverside County, where no such threat was made, should dispel this belief. It too had multiple cases in June had the peak expression of symptomatic cases on Aug 1 and had more than 60% of infections before Aug 10. The pattern of the epidemics through time in the two counties is very similar, although Riverside had no massive aerial spray campaign. Riverside had less than half the rate of infection as Sacramento. If one views the data gathered by Reisen it can be seen that the peak virus pool occurred in the middle of July for 2005 (Kramer, V. et al 2005). The decline from the peak was fairly constant throughout the rest of the mosquito season even though the mosquito population was increasing most of that time. The aerial spraying didn’t begin until after the first week of August, some three weeks after the virus pools had begun declining. Something else entirely accounts for the decline in transmission, since it is apparent that the spraying didn’t start until after the epizootic was on the way out.

Effective and Safe Measures Ignored.  Meanwhile, the SYMVCD is not utilizing the full compliment of safe and effective biological control agents. No mention is given to their efficacy as alternatives to this aerial spray protocol, so I will introduce these comparisons. The mosquito-parasitic nematode, Romanomermis culicivorax, has been shown to reduce mosquito populations by as much as 95% on inundative releases (Peterson, J.J., Chapman, H.C. and Willis, O.R. 1979), (Rojas, W.A. et al 1987). The mosquito-pathogenic fungus, Lagenidium giganteum, has exhibited similar efficacy and is even more specifically effective against the putative vectors of WNV (Lasko, J.F. and Washino, R.K. 1983). Both of these agents will establish ongoing populations and remain effective at controlling mosquito populations over a number of years. Both of these organisms are obligate parasites of larval Culicidae, mosquitoes, and infect nothing else. They present no risk to either human health or the environment.

    Unfortunately, neither of these agents can be cultured artificially in a way that retains its infectivity. For this reason they are not being produced commercially. However, SYMVCD once cultured its own Romanomermis in vivo, but there has been no mention of re-establishing such a culture program or developing one for Lagenidium.  With the expertise of this vector control district such programs should be easily devised. Such a mass culture facility could also produce Bacillus spaericus and Bacillus thuringiensis ser.H14, both of which have efficacy in reducing mosquito populations.

Incomprehensible Public Policy.  It seems almost ridiculous for the District to consider multiple millions of dollars in expenditures on insecticidal agents and not spend a dime producing these safe and effective biological agents. The risk-benefit comparisons should be obvious. On the one hand we have a material with dubious efficacy and a guaranteed universal exposure to an incompletely assessed risk; and on the other we have a proven safe and effective set of biological controls.

    When cost questions are introduced the comparison is even more exaggerated. The added labor required to culture the District’s own biological alternatives would end up both a benefit to community employment and substantially less costly for the District’s budget than procuring poisons and aircraft delivery systems. That the District is choosing ongoing outlays for 100% depreciable investment in distributing poison as opposed to investing in a permanent facility for production of a renewable resource is beyond any reasonable comprehension.

CDHS (2006) 2005 Human WNV Case Linelist (web publication at CDHS page 2005 Human WNV Case Linelist.pdf)

Kramer, V. (2005) Analysis of Sacramento County WNV Human Case Data with Onset of Illness During Time Periods Relative to the Aerial Spraying Conducted in 2005.  (web publication at CDHS page Analysis of Sacramento County WNV Human Case Data with Onset.pdf)

Kramer, V. et al (2005) 2005 West Nile Virus Activity in California. (web publication at CDHS page 2005WNVActivityinCA.pdf)

Lasko, J.F. and Washino, R.K. (1983) In Situ Studies on Seasonality and Recycling Pattern in California of Lagenidium giganteum Couch, an Aquatic Fungal Pathogen of Mosquitoes. Environmenal Entomology 12(3): 635-640

Peterson, J.J., Chapman, H.C., and Willis, O.R.  (1979) Release of Romanomermis culicivorax for the control of Anopheles albimanus in El Salvador I: Mass production of the nematode. Am. Journ. Trop. Med. Hyg. 27: 1265-1267

Peterson, J.J., Chapman, H.C., and Willis, O.R. (1979a) Release of Romanomermis culicivorax for the control of Anopheles albimanus in El Salvador II: Application of the nematode. Am. Journ. Trop. Med. Hyg. 27: 1268-1273

Peterson, R.K., Macedo, P.A. and Davis, R.S. (2005) A Human-Health Risk Assessment for West Nile Virus and Insecticides Associated with Mosquito Management. Environmental Health Perspectives  doi:10.1289/ehp.8667 (available at http://dx.doi.org/) Online 28 October 2005

Reisen, W.K. et al (2004) West Nile Virus in California. (web publication at CDC page http://www.cdc.gov/ncidod/EID/vol10no8/04-0077.htm)

Reisen, W.K. et al (2006) Role of Corvids in the epidemiology of West Nile virus. (web publication at CDHS page, rolecorvidsintheepidemiology.pdf)

Rojas, W.A. et al (1987) Reduction of malaria  prevalence after the introduction of Romanomermis culicivorax (Mermithidae:Nematoda) in larval Anopheles habitats in Colombia. Bull. WHO  65(3): 331-337
Weston, D.P. et al (2006) Aquatic Effects of Aerial Spraying for Mosquito Control over an Urban Area. Environ. Sci. Technol. 40:5817-5822