Seeds for one sample were grown in Poland, the remaining seeds used to press the oils were grown North America. Five Canadian-sourced samples were grown in provinces the manufacturer did not disclose. Of the remaining, one was grown in Alberta, three in Manitoba, and two were grown in Saskatchewan. Samples grown in the United States include three from seeds grown in Indiana, two from Kentucky, one from Montana, three from North Dakota, three from New York, and five from Virginia.
The researcher did not investigate cannabinoid variances based on growing locale since most suppliers reported only a single cultivar grown in a single location. Therefore, analysis by growing locale would not provide any new or significantly meaningful information given the available data.
Suppliers identified the cultivar for twenty-six of the samples, which included a mix of eleven varietals. The number of different cultivars ranged from one to four varietals per supplier. Of the four samples of an unknown varietal, one supplier did not respond to follow-up requests to identify the cultivar(s), and one sample came from a US manufacturer that sourced seeds from Poland without the varietal data being available.
Nineteen of the samples were classified as “conventional” by the supplier, and seven were classified as “organic.” The researcher did not attempt to verify compliance with any official organic certification. The conventional or organic status of four samples was not provided.

Analytical Procedures and Sample Analysis

The researcher contracted with a third-party ISO 17205 accredited laboratory that is also DEA, FDA, and USDA licensed to develop a validated GLP method to assay sixteen (16) cannabinoids in CPHSO with a lower LOQ of 50 ppb (0.05 ppm) using UHPLC-MS/MS. The FDA-CVM reviewed the method before the laboratory conducted the GLP-compliant analysis. All analytical measurements were conducted using a Shimadzu 8050 LC-MS/MS instrument and are reported in ppm (except density).
The development of a low-LOQ analytical method was required to quantify cannabinoids that naturally occur in minimal quantities that commercial laboratories could not assay. Also, there is no commonly-available method or approved AOAC or AOCS method for testing cannabinoids in CPHSO at very low LOQs, and the AOCS draft method did not provide sufficiently-low LOQ for the study requirements. Finally, this research is part of a series of safety and efficacy investigations relating to hemp in animal feeds. Existing research indicates some animals have a lower tolerance to cannabinoids than humans, with CBDA and CBD hepatotoxicity documented in multiple species, so these cannabinoids are of particular interest.
The third-party laboratory assayed the acceptable samples using the validated method and provided the results to this researcher. Each sample’s oil density and cannabinoid analytes were assayed multiple times per the GLP requirements. The mean values of density and each assayed cannabinoid content were reported in ppm and entered into a spreadsheet for further analysis. The analytical assay found no detectable levels of Δ8-THC, so this analyte was not included in the statistical analysis or reported herein. The analyte values (in ppm) were also summed for each sample.
Data were analyzed using Microsoft Office Professional® 2016 Excel®spreadsheets and IBM SPSS® version 26 statistical analysis software. The final documentation was constructed with Microsoft Office Professional® 2016 Word® and Adobe Acrobat Professional® 2017.

Data preparation before statistical analysis

For the phase 1 investigation, a value of 95% of the LOQ was substituted for cannabinoid analytes detected below the lower LOQ. This substitution value was chosen because the investigator determined that this value may represent the true data better than zero (0.0) or treatment as missing values for the analysis. In many cases, only a few values are substituted for any given analyte. Notable exceptions are CBN (20% <LOQ replaced) and THCVA (46.7% <LOQ replaced). The <LOQ substituted values are highlighted with light-gray in Table 7.
Descriptive statistics were computed for all values, including a 95% confidence interval and the inter-quartile lower and upper range limits. Table 8 shows the descriptive statistics for the trimmed data. Light-red backgrounds and red text indicate values that exceed the upper limit of the 95% confidence interval in Table 7.
The untrimmed data was copied to a second spreadsheet for phase 2, and all substituted <LOQ values and values higher than the upper limit of the initial 95% confidence interval were deleted. The descriptive statistics were recomputed for all values. Table 12 (a-b) shows the trimmed data with recomputed descriptive statistics.