The assessment methodology for biological effects measurements is essentially the same as that for chemical concentrations in biota. However, modifications are required for some biological effects and these are described below.
Glutathionine transferase (GST), acetylcholine esterase activity (ACHE), aminolevulinic acid dehydratase (ALAD)
Low values of these variables indicate unhealthy organisms, so status is assessed using the lower one-sided 95% confidence limit on the fitted mean value in the most recent monitoring year. For example, if the lower confidence limit is above the Background Assessment Concentration (BAC), then the mean value in the most recent monitoring year is significantly above the BAC and levels are said to be ‘at background’.
Scope for growth (SFG)
The measurements are not log transformed because scope for growth can be negative and because the data are approximately normally distributed on the untransformed scale. Consequently, all models are of temporal changes on the original scale. Further, low scope for growth indicates unhealthy organisms, so status is assessed using the lower one-sided 95% confidence limit on the fitted mean scope for growth in the most recent monitoring year. As the data have not been transformed before modelling, the lower confidence limit is compared directly to the assessment criteria; i.e. there is no need for any back-transformation.
Neutral red retention time (NRR), lysosomal labilisation period (LP), stress on stress (SURVT)
The measurements are interval-censored time-to-event data. For example, NRR is the time until cell membranes, flooded with a neutral red dye, stop working and the dye leaches out of the cells. They are interval-censored because the cells are inspected at set time points (15, 30, 60, 90, 120, 150 and 180 minutes) with an NRR of 30 minutes indicating that the cells were intact at 30 minutes but not at 60 minutes. For NRR and LP, the measurement is the inspection time before the event. However, for SURVT, the measurement is the inspection time after the event; thus a SURVT measurement of 6 days indicates that the organism died at some point between 5 and 6 days.
The measurements are modelled using standard survival analysis techniques using the flexsurv package (Jackson, 2016) in the R statistical environment (R Core Team, 2020). Specifically, a gamma survival model is fitted with the form:
where E(time-to-event) is the expected time-to-event and f(year) depends on the number of years of data (see methods for chemical concentrations in biota). Fitted values are calculated on the time-to-event scale, with pointwise confidence intervals estimated by simulation.
Note that:
Christopher Jackson (2016). flexsurv: A Platform for Parametric Survival Modeling in R. Journal of Statistical Software, 70(8), 1-33. doi:10.18637/jss.v070.i08
R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/
Comet assay (%DNATAIL)
The measurements are proportions, obtained by measuring the intensity of light in the tails of comet cells relative to the intensity of light in both the heads and tails of the cells. Low values indicate healthy organisms. The data are modelled as a generalised linear mixed model with a beta distribution and a logistic link of the form:
where f(year) depends on the number of years of data (see methods for chemical concentrations in biota) and where the random term accounts for random between-year variation. The measurements are weighted by the number of cells in each sample.
Micronucleus assay (MNC)
The measurements are proportions and should be modelled assuming they have a (beta-)binomial distribution or, because the proportions are very low, as a poisson or negative-binomial distribution with an offset to account for the number of cells sampled. However, at present no time series has more than two years of data, so such methods have not yet been implemented. Instead, an ad-hoc assessment of status is made by calculating the mean measurement each year and comparing this value (1 year of data) or the maximum of the two values (2 years of data) to the assessment criteria. Low values indicate healthy organisms.