New Studies Closing Door for Roundup Plaintiffs?

25 min readApr 8, 2019

Widespread attention has focused on current Roundup trials involving claims the extensively used Monsanto herbicide causes Non-Hodgkin’s Lymphoma (NHL). The lawsuits are all filed in the U.S. where jurors decide the fate of these cases by evaluating studies from several types of scientific and medical disciplines alongside allegations of deceptive intent developed through interpretation of selected memos and reports of the 12 million Monsanto documents generated through the discovery process.

An understanding of the historical context is necessary for understanding why the Roundup litigation is occurring now despite more than 40 years of world-wide regulatory approvals. The primary factor was the disputed 2015 decision assigning a hazard classification of “probable carcinogen” for glyphosate by the International Agency for Research on Cancer (IARC), an autonomous agency of the World Health Organization. IARC’s institutional charge is the identification of cancer hazards for national regulators who in turn perform a human exposure assessment for determination of risks for human populations. Leaving aside the unanimous disagreement of all key world-wide regulators with IARC conclusions for animal and laboratory studies, the IARC also concluded there was limited evidence from the population or epidemiological studies of exploratory case control studies. Another consideration is the legal distinction between the “hazard” conclusion of the IARC and the human “risk” based conclusions of international regulatory agencies. In this respect, the federal judge in charge of the Roundup multi-district litigation indicated the IARC hazard conclusion, alone, cannot support a finding of general causation, much less a specific causation conclusion required for attributing a plaintiff’s NHL to glyphosate based formulation (GBF) exposures. Moreover, the recent epidemiological studies were not yet formally published in 2015 and under IARC’s rules could not be considered in forming their conclusion.

Product liability lawsuits involving health effects resulting from product usage nearly always place great weight upon epidemiology studies. During Roundup trials to-date, jurors evaluating the risk of NHL cancer from GBFs such as Roundup have been presented fragmented and contradictory expert testimony regarding the entire body of epidemiology studies, including early exploratory studies which have minimum or null scientific relevance. The presentation of this confusing epidemiological evidence coupled with jury instructions allowing jurors to conclude Roundup need only be a contributing factor for plaintiff’s cancer independent of medical, genetic or other significant factors provides a substantial impetus for Roundup culpability. Reliable epidemiological studies provide the best evidence because they measure the direct effect of GBF exposure and incidence of NHL cancers within human populations, unmistakably a better and more direct “proof” than animal and laboratory based studies. However, recent trials and associated public and social media reports have largely ignored the importance of the most recent epidemiological studies. Three recently published or soon to be published epidemiological studies encompassing four countries, hundreds of thousands of participants and millions of GBF days of exposure provide consolidated proof there is no causal association with NHL. The epidemiological history underlining the significance of three studies is summarized as follows:

Steadily rising NHL incidence rates of were observed from at least early 1970s to the1990s or later in the U.S. and other developed countries. More than half of NHL cases could not be explained nor attributed to GBFs because the Roundup introduction date and a NHL minimum latency period indicates NHL cases could not have appeared until the middle1980s. Exploratory case control studies were initiated in the late 1970s for the identification of unexplained risk factors. Initial case control studies typically included only small numbers of cases exposed to GBFs; subjects often tended to inaccurately recall pesticide usage for previous decades; a large percentage of interviews or questionnaires were conducted with next-of-kin or “proxies” because of subject unavailability; the method of selecting subjects may have biased results; and the number of subjects exposed to GBFs as well as levels of exposure were limited. Small numbers of exposed cases coupled with a large number of pesticide, medical history, lifestyle, heredity and occupational variables substantially impeded identification of confounding variables; for example, instances where more than one variable resides in subjects and causes the same health effect. As a result, exploratory case studies had only crude adjustments without pesticide adjustments or adjustments with results that were likely to be inaccurate. Aside from lacking internal reliability or validity, none of the exploratory case control studies presented statistically significant risk measures adjusted for other pesticides demonstrating a causal association of GBF exposures and NHL. Stated in an alternative form, there are no studies providing a reliable and scientifically valid measure for “ a more than likely” probability of GBFs causing NHL or its main sub-types.
A soon to be published study of the North American Pooled Project, but initially reported in 2014 pooled the datasets of two exploratory case control studies, a study of several Midwestern states (De Roos 2003) and a Canadian study (McDuffie) and the combined dataset analysis found no causal associations between glyphosate formulations and NHL or major sub-types. The pooled analysis also confirmed the invalidity of projecting GBF usage to NHL cancers of plaintiffs based on the De Roos and/or McDuffie studies as has occurred during Roundup trials. The Eriksson study of 2008 has also been used for projection of NHL to plaintiffs, but Eriksson also does not have a fully adjusted and statistically significant measure of causal association.
The Andreotti study published in 2018 of the Agricultural Health Study tracked, for decades, GBF exposures and overall NHL and sub-type cancer incidence of more than 50,000 pesticide applicators, farmers and commercial operators, many of whom had more than 100 days of accumulated exposure. The results found no evidence of association between GBFs and NHL or the main sub-types. The forward looking tracking of GBF exposures and cancer incidence along with sensitivity tests of internal consistency and data reliability marks the 2018 AHS study as the most comprehensive of the two internally validated and reliable studies performed to-date. The 2005 AHS study was the only other study with a sufficient level of internal validity to provide reliable risk estimates and it as well reported no associations between GBFs and NHL.
AGRICOH, a program of the IARC supports data pooling of agricultural cohort studies for improved analysis of disease-exposure associations. As part of this program, Leon et al, published a pooled cohort analysis on March 18, 2019. The cohort groups included those from AGRICAN, a program of the Mutualité Sociale Agricole, the French national health insurance system of agricultural workers; CNAP, an aggregate group of farm holders and families compiled by Statistics Norway; and the Agricultural Health Study (AHS), which included farmers and farm workers from Iowa and North Carolina. The pooled dataset included more than 300,000 subjects with potential exposure to 33 pesticides tracked for 3.5 million years of follow-up for an association with 2,400 NHL cases. The results indicated there was no association of GBFs with NHL, but there was an elevated association with one NHL sub-type largely attributable to the Norwegian cohort. An additional analysis was presented using a “minimally adjusted” number of variables and the risk rate for this sub-type approached a null association. AGRICOH did not include a level of exposure analysis as only the U. S. cohort had detailed exposure data, but the pooled cohort study confirmed the lack of causal association for ever use of GBFs and NHL as well as major sub-types.

Given the epidemiological history, the plaintiff’s practice of presenting the entire set of studies, including a majority of which lack internal validity and scientific significance for enabling a jury selection of a causal scientific conclusion of their own choosing is abusive of U.S. judicial systems as well as the science of epidemiology. The direct legal implication for Roundup cases of adding the recent studies to the available epidemiological evidence is the explicit denial for plaintiff’s experts of an epidemiological foundation for “ruling in” Roundup for “differential analysis” opinions stating GBF exposures lead to plaintiff’s NHL cancer. Similarly, the plaintiff tactics of presenting to jurors “trend lines” for a chronology of internally invalid studies with no causal associations or aggregating individual studies into a meta-analysis for studies with unequal quality or measurements also blatantly misrepresents the epidemiological evidence. For example, during recent trials, plaintiffs’ experts touted a recently published meta-analysis, but the underlying study methodology and conclusion was invalid and as a consequence the juries were presented an erroneous scientific conclusion for their consideration. The federal and state judicial systems have legal precedent based standards for qualifying expert witnesses, but there are no corresponding standards or instructions for facilitating jury understanding of the legal and scientific significance for presented evidence; specifically, the absence of a statistically significant causal conclusion precludes jury acceptance of an overall epidemiological conclusion favorable for the plaintiff.

A corollary issue is the necessity for the IARC to initiate steps leading to a prompt reevaluation of glyphosate and associated GBFs using a more appropriate and disciplined analytical methodology consistent with international guidelines. The reevaluation is obligated to consider the latest epidemiological studies; all of available animal study data (without excuses); and an appropriately weighted analysis of relevant and conclusive cellular mechanism studies. Experts in various scientific disciplines have detailed serious analytical flaws as well as omissions within the analysis of IARC’s epidemiology, animal and cellular mechanism studies. Identified issues, among others, include distorting of evidence relating to mouse studies by the exclusion of exculpatory studies; use of inappropriate statistical techniques; and the elevation of significance for studies in which the authors as well as regulatory agencies and other scientific organizations have stated there was low biological significance or inconclusive results.

The entire cancer prevention community would benefit from IARC’s implementation of measures to significantly increase transparency and communication such as those already employed by the Environmental Protection Agency, among others. Among other measures, reviews by outside expert panels and allowing public commentary regarding drafts reports prior to publishing a final monograph conclusion could substantially facilitate monograph acceptance among among the world-wide scientific, regulatory and cancer prevention communities.

Epidemiological History of Roundup and NHL

The following sections detail the Roundup epidemiological history, particularly the more recent studies.

Expert testimony for Roundup civil trials include three categories of general causation evidence regarding potential health effects of glyphosate and glyphosate based formulations (GBFs) such as Roundup; population, animal and cellular studies. Medical specialists subsequently use conclusive points derived from general causation evidence as a basis for attributing, or not, causation for one of the sub-types of Non-Hodgkin’s Lymphoma (NHL) or the overall NHL cancer group to plaintiff’s cancer. Animal and cellular studies because of distance from human biological systems require speculative translations for projecting “supposed” human health effects of glyphosate or GBFs. As a result, the highest significance, whether for civil trials or regulatory approvals, is generally placed upon reliable population or epidemiological studies because of their direct measurement of human health effects related to GBF exposures.

A steadily increasing NHL incidence rate trend was observed from the 1970s until at least the early 2000s or later in the U.S. and other developed countries. Blair et al 1995, estimated AIDS associated NHL accounted for 47% of the increased incidence leaving an estimated 53% to unexplained risk factors. The Roundup introduction in 1975 coupled with an estimated NHL latency period of 10 years also underscores the lack of GBF involvement with the sharply inclined trend ending in the early 1990s.

Roundup herbicide was introduced in 1975 and the first population or epidemiological studies reflecting potential health effects of (GBFs) began few years later as shown below thereby allowing, more or less, for a period of time or a latency period for disease development.

Initial studies, with the exception of those referencing the Agricultural Health Study (AHS) were exploratory case control studies and reflected a largely consistent set of attributes:

Selection of study subjects from confirmed Non-Hodgkin’s Lymphoma (NHL) patient cases and recruitment of control subjects without NHL.
The studies were “exploratory in nature”; i.e., subjects were interviewed regarding previous decades’ usage for a large number of pesticides and other chemicals exposures as well as medical history, family heredity issues, lifestyle factors and occupational histories.
Initial case studies typically had small numbers of cases exposed to GBFs, interviewed cases and controls were subject to inaccurate recall of pesticide usage, a large percentage of interviews were conducted with next-of-kin or “proxies” because of subject unavailability and the method of selecting case and control subjects may have introduced selection bias and distorted study results.
Small numbers of exposed cases coupled with a large number of pesticide, medical history, lifestyle, heredity and occupational variables significantly impeded identification of confounding variables; specifically when more than one variable resides in subjects and causes the same health effect. As a result, exploratory case studies had only crude adjustments without pesticide adjustments or statistical adjustments with misleading results.
The advantages of exploratory case control studies lie in the relative ease of organization and completion as well as the comparatively lower costs. In addition, case studies often provide information highlighting points of emphasis for subsequent studies.

Agricultural Health Study; De Roos 2005

The publishing of De Roos et al 2005 Cancer Incidence among Glyphosate-Exposed Pesticide Applicators in the Agricultural Health Study was the first use of a forward looking cohort study for examining the health effects of GBF exposures without the inherent exploratory case control difficulties. The following extracts summarize De Roos 2005:

The AHS is a prospective cohort study in Iowa and North Carolina, which includes 57,311 private and commercial applicators who were licensed to apply restricted-use pesticides at the time of enrollment. Recruitment of the applicators occurred between 1993 and 1997 (Alavanja et al. 1996). Cohort members were matched to cancer registry files in Iowa and North Carolina for case identification and to the state death registries and the National Death Index (National Center for Health Statistics 1999) to ascertain vital status. Incident cancers were identified for the time period from the date of enrollment until 31 December 2001 and were coded according to the International Classification of Diseases, 9th Revision (WHO 1977). If cohort members had moved from the state, they were censored in the year they left. The median time of follow-up was 6.7 years.

Using a self-administered enrollment questionnaire, we collected comprehensive-use data on 22 pesticides, ever/never use information for 28 additional pesticides, and general information on pesticide application methods, personal protective equipment, pesticide mixing, and equipment repair. Data were also collected on basic demographic and lifestyle factors. Applicators who completed this questionnaire were given a self-administered take-home questionnaire, which contained additional questions on occupational exposures and lifestyle factors.

Results:

There were rather few cases of NHL for inclusion in this analysis (n = 92); nevertheless, the available data provided evidence of no association between glyphosate exposure and NHL incidence. This conclusion was consistent across analyses using the different exposure metrics and in analyses using either never exposed or low exposed as the referent. Furthermore, there was no apparent effect of glyphosate exposure on the risk of NHL in analyses stratified by state of residence or in analyses of highly exposed groups comparing the highest with the lowest quintile of exposure.

Agricultural Health Study (AHS): Alavanja 2014

(Information from depositions, expert reports and testimony of federal Roundup multi-district litigation.)

Alavanja et al, drafted a report dated March 15, 2013 which presented a follow-up of the Agricultural Health Study cohort and De Roos 2005 study on the association between pesticide use, including GBFs, and risk of lymphoma. Findings for insecticides, fungicides, and fumigants and NHL were published in 2014, but herbicides, including GBFs were omitted for unknown reason(s). The Alavanja study of 2014 marked the first time availability of an epidemiological forward looking cohort study with significant long term subject tracking as the draft report includes an additional 7 years of follow-up for cancer incidence through December 2008. The Alavanja study used the same epidemiological and statistical methods as De Roos 2005. Results included 333 incident cases of NHL compared to 92 in De Roos 2005 as well as updated definitions for NHL cancers. Also notable was the high prevalence of co-exposure of pesticide use which was observed as 52% of the NHL cases were exposed to five or more herbicides and highlighted the need for fully adjusted multivariable models for treating potential confounding.

Cohort participants were asked to complete a follow-up questionnaire from 1998 to 2003 to update information on pesticide use since enrollment. This was completed by 63% of the original cohort. Dose exposure categories, including lifetime days of use, intensity-weighted days of use were also analyzed.

The relative risk estimates were age-adjusted and multivariable adjusted for potential confounders. Compared to never users, there was no evidence of a positive association for exposure to glyphosate and risk of NHL and no evidence of dose-response in the multivariable models:

The relative risk (95% CI) estimate for highest vs. lowest lifetime days exposed was 1.0 (95% CI 0.7–1.4).
The relative risk estimate for highest vs. lowest level of intensity weighted days of exposure was 0.97 (95% CI 0.7–1.4).
There was also no evidence of dose-response associations. In addition, sensitivity analyses indicated that results were similar when follow-up questionnaire results were excluded and the prior definition of NHL was used.

The head of the 2015 glyphosate panel of the International Agency for Research on Cancer (IARC) was fully aware of this study and the implications for IARC glyphosate conclusions, but apparently did not disclose to other panel members or to the IARC.

The North American Pooled Project (NAPP): Pahwa 2015

(Information from files of federal multi-district litigation.)

Pahwa et al 2015, of the North American Pooled Project (NAPP) is a pooled analysis of four case control studies, including the three studies in the pooled analysis by De Roos 2003 as well as the McDuffie 2001 Canadian study. An abstract was submitted in 2015 to the International Society for Environmental Epidemiology (ISEE) and was accepted as an oral presentation as well as two versions of power point presentations for the meeting.

The pooled analysis increased the overall size from De Roos 2003 to 1690 NHL cases, 533 of whom were proxy respondents and 5131 controls. Specific dose-response data were lacking from Cantor study data from Iowa/Minnesota and Hoar study data from Kansas and the analyses on frequency and lifetime days were restricted to the case-control data from Nebraska (Zahm) and Canada (McDuffie). GBF usage was a very low as only 113 of the 1690 total NHL cases had ever used GBFs and there are smaller numbers of cases in dose response analyses since not all of the studies had complete information collected.

The odds ratios in the abstract were only adjusted for age, sex, location, proxy respondent, family history and use of protective equipment, but not for other pesticides. In contrast, the power point presentation slides present multivariable models with these factors as well as adjustment for use of 2,4-D, dicamba, and malathion. Moreover, the slide deck compares the fully adjusted odds ratios separately for proxy and self-respondents to those of proxy alone.

More recently, Dr. Dennis Weisenburger, a co-author, testified during Hardeman v. Monsanto trial the Pahwa study is being finalized for publication with an online version possibly available before mid-year and a final publishing date prior to year end. Regardless, plaintiff and defense expert witnesses have testified regarding Pahwa 2015.

Pahwa 2015 provided the following results:

Multivariable analysis excluding proxy respondents, thus eliminating the recall bias associated with use of proxy respondents reflected an odds ratio for ever use of GBFs of 0.95 with a 95% confidence interval (CI) of 0.69–1.32 for the pooled analysis.
The number of years of use metric was available from all four NAPP datasets with highest number of years of exposure (>3.5 years vs. 0 years) provided for a non-adjusted (for other pesticides) odds ratio of 1.20 with 95% CI (0.82–1.75); an odds ratio adjusted for other pesticide use of 0.94 with 95% CI (0.62–1.42); and an odds ratio additionally limited to self-reported data of 0.78 with 95% CI (0.49–1.24).
The lifetime days of GBF exposure metric reported a non-adjusted odds ratio of 1.55 with 95% CI (0.99–2.44) which was reduced upon adjustment for other pesticide use to 1.08 with 95% CI (0.66–1.77) and subsequently when restricted to self-respondents resulted in odds ratio of 1.06 with 95% CI (0.62–1.81). The number of years of use, the most important dose response metric available for this study, did not show evidence of positive associations for NHL risk.
The dataset (largely derived from the McDuffie study) for the >0<2 and >2 days per year metric provided a non-adjusted odds ratio of 2.42 with 95% CI (1.48–3.96) for >2 days per year. Additionally adjusting for other pesticide use, the odds ratio was reduced to 1.73 95% CI (1.02–2.94). Limiting to self-respondents provided an odds ratio of 1.77 with 95% CI (0.99–3.17).
NHL sub-type odds ratios. The August 2015 slide deck reported 45 cases and a DLBCL non-adjusted odds ratio of 1.60 with 95% CI (1.12–2.29) for ever use and adjusting for use of the three pesticides reduced the ever use odds ratio for DLBCL to 1.23 with 95% CI (0.81–1.88). After adjusting for confounding by other pesticides, the associations between ever use of GBFs and risk of SLL (15 exposed cases) of 1.79 with 95% CI (0.87–3.69) and other sub-types (25 exposed cases) of 1.51 with 95% CI (0.87–2.60). The sub-type analysis did not present the effect for proxy respondents.

The NAPP pooled analysis demonstrated several points:

Controlling for three other pesticides (2,4-D, dicamba and malathion) shows no causal associations between GBFs and NHL.
North American exploratory case control studies using non meaningful measures of exposure such as days per year of use (largely from the Canadian McDuffie study along with a variety of other limitations and biases.) lead to contrived “dose response” comparisons. Alternatively, the pooled analysis demonstrated when more meaningful and validated measures are used, there is no association between years of use and GBFs or cumulative lifetime days of use of GBFs and NHL after adjustment for other chemicals.

American Health Study (AHS): Andeotti 2018

The Andreotti et al, Glyphosate Use and Cancer Incidence in the Agricultural Health Study was formally published in May of 2018, but online availability of November, 2017 allowed for use during Roundup pre-trial hearings and civil trials. Andreotti 2018 became the follow-on AHS study to De Roos 2005 instead of the Alavanja 2014 study. Again, extracts summarize the study:

The AHS is a prospective cohort of licensed pesticide applicators from North Carolina and Iowa. Here, we updated the previous evaluation of glyphosate with cancer incidence from registry linkages through 2012 (North Carolina)/2013 (Iowa). Lifetime days and intensity-weighted lifetime days of glyphosate use were based on self-reported information from enrollment (1993–1997) and follow-up questionnaires (1999–2005).

In this large, prospective cohort study, no association was apparent between glyphosate and any solid tumors or lymphoid malignancies overall, including NHL and its subtypes.

AHS Glyphosate Exposure Days Quartiles & NHL Incidence

There was also NHL sub-type analysis with no evidence of association or dose responsiveness for sub-types, including for Diffuse Large B-Cell Lymphoma:

Days Glyphosate* DLBCL Cases*Relative Risk*95% Confidence Interval

***NONE**********27***********1.0********* — ********** —

***1–13.74********26***********1.0************(.55–1.82)

13.75–38.74*******31***********1.24***********(.70–2.20)

38.75–108.4*******21***********0.94***********(.50–1.75)

**>108.4**********26***********0.97***********(.52–1.81)

AGRICOH POOLED COHORT ANALYSIS

A pooled cohort analysis, Pesticide use and risk of non-Hodgkin lymphoid malignancies in agricultural cohorts from France, Norway and the USA: a pooled analysis from the AGRICOH consortium, was published by Maria Leon et al during March, 2019. AGRICOH, a program of the International Agency for Research on Cancer (IARC), is “an international consortium of agricultural cohort studies formed in October of 2010 to encourage and support data pooling to study disease-exposure associations which individual cohorts do not have sufficient statistical power to study.

The three cohort groups included those from AGRICAN, a program of the Mutualité Sociale Agricole, the French national health insurance system of agricultural workers; CNAP, an aggregate group of farm holders and families compiled by Statistics Norway; and the Agricultural Health Study (AHS), which enrolled farmers and farm workers from Iowa and North Carolina. Overall, the study reflected more than 2,400 cases of NHL across more than 3.5 million person-years of follow-up for more than 300,000 subjects.

Pooled Cohort Risk Rates: NHL and Specified Sub-types

Glyphosate-NHL-CLL/SLL-DLBCL

Diffuse large B-cell lymphoma

During follow-up, 434 DLBCL cases were diagnosed and most analyses did not show any associations:

There was an elevated mHR of DLBCL with ever use of glyphosate (mHR = 1.36, 95% CI: 1.00–1.85; I2 = 0%).
The cohort-specific HR for ever use of glyphosate and DLBCL were HR = 1.06 (95% CI: 0.51–2.19) in AGRICAN. (28 cases and 36% farmers exposed) and
HR = 1.67 (95% CI: 1.05–2.65) in CNAP (100 cases and 38% farmers exposed) and
HR = 1.20 (95% CI: 0.72–1.98) in AHS. (93 cases and 83% farmers exposed).

Supplemental Table 3 shows ever use of 14 pesticides and 33 active ingredients and meta-analysis estimates, minimally adjusted, of NHL and sub-type DLBCL for glyphosate:

1131 NHL cases with Hazard Ratio random effects meta-analysis (minimally adjusted models) of .98 95%CI(.76–1.25) and P-value for heterogeneity of 80%.
221 DLBCL cases and mHR of 1.12 with 95%CI(.86–1.45) and p-value for heterogeneity of 21%.

Whereas the lack of association of ever/never use of glyphosate with NHL overall in our analysis is consistent with a recently published analysis from AHS reporting no association between lifetime days or intensity-weighted lifetime days of glyphosate use and NHL (440 exposed cases), our mHR observed for DLBCL with ever/never use of glyphosate for the three cohorts combined is higher than that for AHS alone. In addition to evaluating different exposure metrics, our analyses using the AHS data differed in several aspects from the recent AHS publication (2017). First, the AHS publication included 4619 commercial applicators (non-farmers, and thus excluded from our analysis), but excluded 1620 farmers with information on ever use of glyphosate but who did not report frequency of use (eligible for inclusion in our analysis). The follow-up time was longer in the AHS publication (up to 2012 and 2013), and thus more cases were included than in our analysis (130 vs 113 DLBCL cases). Finally, different variables were used to adjust the risk estimates (in our present analysis, we did not adjust for cigarette smoking, alcohol intake, or family history of any cancer but did adjust for animal production and for different pesticide active ingredients from those included in the AHS publication.

Differences among the cohorts may have affected risk estimates. Inclusion criteria for each cohort influenced the prevalence of exposure. AHS enrolled farmers at the time of obtaining or renewing licenses to apply restricted-use pesticides, whereas in AGRICAN and CNAP, identification of cohort members was not tied to potential use of pesticides. Indeed, 67% of AGRICAN and 45% of CNAP cohort members were classified as ever exposed to any of the evaluated pesticides, compared with 99% of AHS cohort members.

In addition, only (potential) use of pesticides on crops was considered in the two cohorts using CEMs to derive pesticide exposure. By design AHS, in addition to pesticides used on crops, also asked farmers about insecticides commonly used on animals. Consequently, the prevalence of a limited number of insecticides (for example, malathion) used in animal husbandry was higher in AHS than in the other two cohorts.

We did not observe an association between risk of NHL overall and ever use of glyphosate, a broad-spectrum herbicide used in agriculture and other settings. There was, however, evidence of heterogeneity of effects among the cohorts (I2 = 57%). Glyphosate was associated with an elevated mHR for DLBCL, and for DLBCL there was no evidence of heterogeneity of effects among the three cohorts. Cohort-specific associations had wide CIs, with only CNAP, which accounted for 45% of the exposed cases, excluding the null value. In CNAP, adjustment for ever use of other pesticides (linuron, aldicarb, mancozeb, DDT, lindane and deltamethrin) generated a fully adjusted HR for ever use of glyphosate of larger magnitude (1.67, 1.05–2.65) than the minimally adjusted estimate (1.26, 0.97–1.65), driven mainly by adjustment for animal production and ever use of DDT (data not shown). Stratification by DDT use, however, produced elevated DLBCL HRs for ever use of glyphosate in both never or ever use of DDT strata [HR = 1.54 (1.04–2.30), never DDT; HR = 1.95 (0.84–4.52), ever DDT], suggesting that the association with glyphosate was not due to concurrent exposure to DDT. Notably, in the meta-analysis, the mHRs for other subtypes of NHL in association with glyphosate were below 1, with CIs including the null value.

Differences in agricultural practices among the three countries might also have influenced exposure. These differences include agricultural practices in the production of any given crop, crops grown, a crop’s pest pressure and need for pest control, application method, use of protective clothing, worker’s age or a combination of these and other factors. For example in AGRICAN, 45% of farmers reported having cultivated vineyards, a production highly dependent on the use of fungicides, which represent about 80% of all pesticides used in vine production.48 Not surprisingly, >60% of farmers/farm workers in AGRICAN were classified as ever users of fungicides, compared with >30% in CNAP and 10% in AHS.

The most important strength of our study is the large number of exposed cases. Although the precision of many of the meta-risk estimates was still limited, it was nonetheless much better than for the individual cohorts and therefore provides additional insight into NHL and sub-type specific risks. An additional strength is that data were derived from prospective cohort studies, thus minimizing differential recall bias.

In our analysis, exposure misclassification is probably non-differential, because pesticide exposure was reported or assessed based on information available before the occurrence of the health outcome, introducing bias towards the null and possibly giving rise to false-negative results. Exposure misclassification may have also reduced the statistical power of our analysis. In addition, given some evidence for an increased NHL risk in farmers and some evidence that several pesticides may increase risk, our use of a reference group of never users of a certain pesticide, which included farmers exposed to many other pesticides, is another limitation. The carcinogenicity of several of the pesticides evaluated is unknown; therefore, if they were associated with NHL, we may have underestimated NHL risk in our study by including exposed farmers in the reference group.

Our exposure assessment approach in the two cohorts based on CEMs that relied on crop cultivation and pesticide registration and sales data will have resulted in a lower specificity than when pesticide use is self-reported at the active ingredient level. Registration of an active ingredient and recommendation for use on a given crop, or documentation that the pesticide was sold in the country, do not necessarily mean that it was used by an individual farmer, leading to overestimation of exposure prevalence. Probability of use of pesticides was not available in the CEMs. The magnitude of misclassification may be limited for commonly applied pesticides but could be substantial for rarely used pesticides. Our analysis demonstrated moderate to high correlations between ever use of active ingredients and between ever use of chemical groups in two cohorts (median correlations 0.63 and 0.64, respectively, in AGRICAN and 0.62 and 0.78, respectively, in CNAP), in contrast to much lower estimates of correlation in AHS (median correlations 0.06 and 0.13, respectively). For instance, between glyphosate and deltamethrin, we observed moderate correlation in AGRICAN (0.71) and CNAP (0.66) and very low correlation in AHS (0.01); between exposure to OC insecticides and phenoxy herbicides, we observed low (0.19), moderate (0.51) and high (0.83) correlations in AHS, AGRICAN and CNAP, respectively. The correlations between the active ingredients are a net result of farming practices and the tools we used to assess exposure. High correlation reduces our ability to distinguish independent effects of individual chemical groups or active ingredients.

An additional exposure assessment limitation involves using ever/never use of pesticide active ingredients as a metric of exposure to explore associations with cancer outcomes as a logical first step. This metric alone is not sufficient to characterize cancer risk from pesticide exposure.

Finally, false-positive associations among our findings may have occurred, given the large number of comparisons (14 chemical groups and 33 active ingredients with five cancer outcomes).

Discussion

The Pahwa 2015 study of NAPP included De Roos 2003 and McDuffie datasets and effectively superseded them from consideration because the pooled analysis clearly shows effects of proxy bias and more statistical power as well as effective multivariate modeling. Fully adjusted risk metrics of Pahwa are not statistically significant for NHL or sub-types even with the peculiar “days per year” metric derived largely from the McDuffie dataset.

Arguing Pahwa does not supersede De Roos 2003 and McDuffie, leaves only one statistically significant result, the De Roos result derived from logistic regression. However, in this instance the logistic regression technique was not used appropriately because of “data sparcity” issues. The logistic regression used 36 exposed cases of NHL and 47 pesticide variables when at least five or more exposed cases are needed for each of the 47 variables to obtain a valid result.

The only other study with claimed epidemiological significance is the Eriksson 2008 case control study. However, there are several issues related to the study which significantly diminish study validity and there were no statistically significant causal associations reported:

29 cases reported exposure to GBFs and using multivariable analyses which adjusted for other pesticides, the odds ratio of 1.51 with 95%CI (0.77–2.94) associated with NHL was not statistically significant.
The study defined subjects unexposed to GBFs to be unexposed to all pesticides, but conversely subjects exposed to GBFs, as a rule, were also exposed to other pesticides.
17 cases reported more than 10 days of exposure to GBFs and the non adjusted model reflected an odds ratio for those with >10 days of GBF exposure of OR 2.36 (95% CI 1.04–5.37), but fully adjusted (for other pesticides) risk estimates were not provided.
The study also presented associations between GBF use and risk of NHL sub-types, but the low exposures to GBFs severely limited the accuracy of sub-type analysis. The majority of the NHL cases were B-cell lymphomas (90%) and the non- adjusted models for ever use of GBFs was associated with an odds ratio of 1.87 (95% CI 0.998–3.51) of B-cell lymphoma. Fully adjusted models accounting for other pesticides were not presented. Notably, the odds ratios for several of the pesticides were also elevated for the B-cell lymphomas highlighting the potential for confounding in the GBF analyses. The other NHL sub-types show a number of positive associations, but there are less than five exposed cases in each group.
GBF exposure was rare with only 29 exposed cases out of a total of 910 and 18 exposed controls out of 1,108. A majority of the odds ratios for different pesticides and for different sub-types are elevated which is highly indicative of systematic bias driving the results. Similarly, as with all exploratory case control studies, recall bias is a concern.
Additional limitations are also imposed by the limited statistical power for subgroup, dose-response and multivariable analyses as evidenced by the wide confidence intervals.
There was also no accounting for multiple testing in their analysis despite the large numbers of comparisons.
There are also issues of residual confounding in this study. Results demonstrated the association between GBFs and NHL risk was not significant after adjustment for other pesticides. However, the study did not adjust in the dose response or sub-type analyses for other pesticides.