Table 4 Surrogates

Approach (see Additional file 4 for example of this approach with real data):

1. Count high and low recoveries for each surrogate chemical across analytical batches.

 □ We typically apply an acceptable range of 50–150% recovery for most environmental samples, particularly when we are analyzing for new chemicals or combinations of chemicals for which methods are not well-established. For well-established methods, a more conservative range – 80-120% recovery – would be appropriate.

2. Identify any sample where all surrogate recoveries were low (e.g., < 50%). This suggests a potential problem with the extraction for that sample.

 □ Discuss with lab analyst. Consider dropping sample.

3. Visualize surrogate recoveries for QC samples (lab blanks, lab control or matrix spikes) across analytical batches. See Additional file 4: Figure S4 for an example.

 □ If these recoveries are out of range, this suggests a larger problem with the analytical method rather than with particular samples. Summarize information about the surrogate recoveries in the QC samples as well as lab control or matrix spike recoveries for the associated chemicals and discuss with lab analyst.

4. Visualize percent recoveries across all samples for each surrogate, by analytical batch. See Additional file 4: Figure S6 for an example.

 □ Note any trends (upward or downward) in the distribution of surrogate recoveries across batches. Such trends should be discussed with the laboratory analyst, even if all recoveries are in the 50–150% acceptable range (see Fig. 3 for an example).

 □ If the surrogate is a deuterated version of one of the target chemicals, it can be helpful to compare a plot of the surrogate recoveries by batch to the sample data for the corresponding un-deuterated target chemical by batch. We would be concerned – and would seek guidance from the lab analyst – if we saw a trend for the target chemical results that matched the trend in the surrogate recoveries.

 □ Note if many surrogate recoveries (e.g., more than half) are out of range in a particular lab batch. If yes, flag the results in that batch for the chemical(s) represented by that surrogate.

5. Visualize sample results flagged by surrogate recoveries. For each individual sample with an out-of-range recovery for a surrogate, flag the results for the chemical(s) associated with that surrogate. Plot all sample data with indicators (e.g., different colors) for whether the representative surrogate for each sample was out of range. See Additional file 4: Figure S7A-D for an example.

 □ Note whether samples with high surrogate recoveries consistently have the highest results for the associated chemical(s).

  ○ If yes, we would be concerned that samples with high recoveries are all overestimated. Discuss with lab analyst. Consider applying a surrogate correction factor to sample results (multiplying by 1/fraction recovery).

 □ Note whether samples with low surrogate recoveries were consistently non-detects or very low-level detects for the associated chemical(s).

  ○ If yes, we would be concerned that samples with low recoveries are all underestimated. Discuss with lab analyst. Note in publications that levels and detection frequencies for associated chemicals might be underestimated.

Reporting:

 □ In summary statistics, we note whether any maximum value is from a sample associated with a high (> 150%) surrogate recovery and note that in this case the maximum might be overestimated. Similarly, if 100% of samples are detects, we also flag the minimum value if it is from a sample associated with a low (< 50%) recovery and note that in this case the minimum might be underestimated.

 □ For any statistical analyses, if possible (i.e., if large enough dataset) we run sensitivity analyses:

  ○ Excluding samples with out-of-range surrogate recoveries.

  ○ Controlling for lab batch, if surrogates were problematic for a particular batch.