It's the Dose, Not the Hypothesis: Reply to.

Dear Editor,

The basis for ecotoxicology is the relationship between dose and response. Hypothesis testing can be indicative of difference, but should never be relied on without considering whether there is a true dose-response relationship. Hypothesis testing cannot inform the researcher regarding causation or uncertainty associated with an observed response, particularly in the case of complex exposures.

We (Page ) evaluated experiments conducted by Heintz and Carls that assessed the toxicity of petroleum to developing pink salmon and herring embryos. The results of these experiments, based on hypothesis testing, were reported as evidence of toxicological effects from exposure to 1 μg/L or lower of dissolved total polycyclic aromatic hydrocarbons (TPAH), and as evidence that oil toxicity increased with weathering. Our evaluation was not, as Heintz incorrectly stated, focused on “supposed failure to adequately control exposure levels” in the experiments. Although we found a number of issues with the work as reported that led us to question the certitude of the conclusions reached, our main focus was on the lack of a dose-response relationship in those experiments rather than on the hypothesis testing relied on by Heintz and Carls . We found that these two articles and the articles that followed, which relied on these initial studies, did not provide technically defensible causal evidence for TPAH toxicity at 1 μg/L or lower (the toxicity benchmark was at least an order of magnitude higher), nor that oil toxicity increased with weathering.

Below we provide responses to all of the concerns raised by Heintz regarding our evaluations of their earlier work (Page ,b,c; Landrum , 2013; Neff ). We (Landrum ) previously congratulated Heintz, Carls, and co-workers “for providing more data and information for analysis than other papers that evaluated toxicological responses to time-varying concentrations of individual compounds or complex mixtures.” We again congratulate them for their pioneering work, but respectfully request that they fully consider the evidence we present, which is not “speculative misinterpretation.” Their interpretation of their findings, based on hypothesis testing, was not unreasonable as a preliminary analysis. However, with further consideration, particularly of the lack of dose-response relationships, and of the continuing failure to establish cause and effect, we find their interpretation is not technically defensible without corroborative testing. Such corroborative testing has not been reported (see below); in fact, Carls et al. in 1999 were unable to replicate the “very weathered oil” findings of Heintz as described by Page

Heintz state “others (Turcotte ) have performed the corroborating studies.” This is not the case. Turcotte showed that alkylated PAH (not high molecular weight polycyclic aromatic compounds [PAC]) are toxic at relatively high concentrations under constant exposure conditions—not the same exposure conditions as Heintz and Carls . Further, toxicity occurred at concentrations in the 40 to 100 μg/L range with the exception of retene, which was around 10 μg/L. TPAH toxicity at 1 μg/L was not demonstrated. Thus, Turcotte have not in fact “performed the corroborating studies.” Similarly, Barron focused on photo-induced toxicity for herring but not pink salmon, with semi-static exposure using daily renewal (i.e., they did not conduct similar testing to Carls ); photo-induced toxicity was not the focus of Heintz . Barron found that the “very weathered oil” exposures of Heintz did not fit their alkyl phenanthrene toxicity model, based on the high toxicity of retene, for pink salmon where the very weathered oil had lower toxic units than exposures not causing toxicity. Retene (a C4-phenanthrene: l-methyl-ispropyl-phenanthrene) is rare in crude oils, usually representing 2–20% of total C4-phenanthrenes in oil (Stout ), and has an atypical toxicity profile (Scott ). It is not representative of the alkyl-phenanthrenes in North Slope crude oil; thus, these papers were not corroborative but in fact support our concern regarding lack of a monotonic dose-response. Incardona et al. (2005) exposed fish embryos to saturated solutions of 4 non alkylated PAH, so didn't confirm that fish embryos are sensitive to 1 μg/L and lower concentrations of TPAH. They did, however, provide new evidence of the mechanisms of toxicity to fish embryos at high concentrations of different non-alkylated PAH. Kiparissis et al. (2005) showed that exposure depends on first order loss-rate kinetics but, again, this work did not corroborate Heintz or Carls .

Heintz accuse us of “selective reading of the literature” and specifically cite eight papers that they accuse us of not considering. In fact, we have been as broad in our reading of the literature as possible; we did consider all the articles we are accused of not considering, and more. None of these articles, or others we have recently reviewed and considered (Jung ), provide evidence that TPAH toxicity occurs at 1 μg/L or lower, nor that oil toxicity increases with weathering. Many of these papers rely on Heintz and Carls without questioning their conclusions (Carls and Meador 2010; Heintz ).

We have repeatedly and in different forums pointed out to Heintz and co-workers that they have not considered key evidence. We do not believe this evidence was purposefully withheld. However, we are concerned that it has been consistently dismissed as irrelevant to their hypothesis. For example, Heintz have not “addressed most of… [our] … criticisms … in a series of letters exchanged in Environmental Toxicology and Chemistry in 2012 (Heintz ).” Our criticisms (Page ,c) were dismissed, but the dismissal was not based on strong scientific proof. We hope that Heintz and co-workers are not relying on constant repetition to sustain their hypothesis, even though the evidence against it, based on their own data, is compelling. One cannot have exposures at lower concentrations that produce effects when exposures at higher concentrations do not; all data must be considered and explained with a consistent conceptual model.

Heintz provide a great deal of text criticizing us for raising the issue of possible confounding effects of oil microdroplets and bacterial/fungal growth. We do not in fact “feel entitled to presume … environmentally relevant information” from these possibilities. But we definitely feel entitled to raise the issue of confounding factors that could explain the curious, non-monotonic dose-response relationships reported by these workers. The data clearly demonstrate (Page ) that microbial degradation occurred and may have contributed to the observed toxic responses. Microdroplets also remain a possible confounding factor, as does binding to dissolved natural organic carbon. Redman showed that the less weathered and more weathered rock column effluents of Carls and the artificially weathered effluents of Heintz contained microdroplets and colloids enriched in higher molecular weight PAHs, the same PAHs to which Carls and Heintz attributed the high embryo-toxicity of weathered crude oil. We agree that there is no proof that these potential confounding effects were causative nor do we imply the extent of contribution of these factors to the response, but there are sufficient data to further question the hypothesis set forth by Heintz and Carls , independent of the major issue of no monotonic dose-response.

Heintz conclude their letter by stating that, “Were Page et al. willing to consider this literature, they would find the details that address their concerns.” We have considered all the literature they cite and more. The problem is that Heintz and co-workers are favoring a hypothesis that is not only inconsistent with their own data, but also with the key technical arguments we have put forward, which include: (1) the unexplained higher toxicity at lower PAH concentrations compared to higher PAH concentrations without toxicity, (2) the lack of a monotonic dose-response relationship, and (3) potential confounding factors.