Immunohistochemistry for CD31 was performed on 78 tumour samples (EPA n=37, placebo n=41) in 5 batches. Examples of the automated CD31-stained vessel detection are shown in Figure 5.19 and Figure 5.20.
Three sections, with a range of staining intensity, were chosen from the first batch and sequential sections from these tumours used as internal controls for each subsequent batch. Subjectively there was good consistency of staining across the five batches (Figure 5.21) although there was wide variation in both tumour vessel density and percentage vascular area across the five batches for each of the internal controls (Table 5.10).
Internal control
ID
Vessel density (vessels per million square microns)
Mean SEM Range Coefficient of
variance
11-0587 157.46 18.21 105.49 - 198.45 25.9
11-0261B 116.33 27.37 33.62 - 204.69 52.6
11-0263B 26.52 4.89 18.40 - 45.61 41.1
Internal control
ID
% vascular area
Mean SEM Range Coefficient of
variance
11-0587 0.29 0.04 0.17 - 0.37 29.8
11-0261B 0.38 0.10 0.09 - 0.66 55.9
11-0263B 0.04 0.02 0.03 - 0.07 43.2
Table 5.9. Measures of tumour vascularity for the internal control sections stained for CD31 in each batch.
Figure 5.19. Representative sections of immunohistochemistry for CD31 tumour vascularity. CD31-stained endothelial vessels are visible in brown, and those vessels detected by the automated image analysis algorithm are highlighted in green on the corresponding markup image (examples of corresponding vessels marked with solid black arrows). Note that the algorithm does not highlight areas of non-specific DAB staining as endothelial cells (examples marked with open arrows).
Figure 5.20. Further representative sections of immunohistochemistry for CD31 tumour vascularity. CD31-stained endothelial vessels are visible in brown, and those vessels detected by the automated image analysis algorithm are highlighted in green on the corresponding markup image (examples of corresponding vessels marked with solid black arrows). Note that the algorithm does not highlight areas of non-specific DAB staining as endothelial cells (examples marked with open arrows), which are more extensive in this section than in Figure 5.18.
Figure 5.21. Internal control sections in each batch of CD31 staining. Three tumours with a range of CD31 staining were selected for use as internal controls. Adjacent sections of these tumours were used as internal controls in the five batches of CD31 staining. A distinctive ROI was identified in each internal control, photographed, and presented here as an example of the consistency of staining between batches. Note that the measures of tumour vascularity were scored for the whole of each slide, not just the representative ROIs in this figure.
100µm 100µm 100µm
100µm 100µm 100µm
100µm 100µm 100µm
100µm 100µm 100µm
100µm 100µm 100µm
11-0587 11-0261B 11-0263B
Batch 1
Batch 2
Batch 3
Batch 4
Batch 5
In the EPA group, one section was scanned out of focus and was unable to be scored, and in the placebo group two tumour sections were erroneously scanned in duplicate in place of two other tumour sections. MVD was therefore scored for 36 tumours in the EPA group and 39 tumours in the placebo group. The measures of tumour vascularity for each group were log transformed to normalise the distribution, then summary statistics were converted back to the original scale for presentation in numerical and graphical form (Figure 5.22). The geometric mean vessel density in the EPA group was 18.54 vessels per million square microns compared to 19.50 vessels per million square microns for the placebo group (p=0.87), and the geometric mean % vascular area was 0.030% in the EPA group compared to 0.034% in the placebo group (p=0.74). There was no correlation between indices of tumour vascularity and either duration of treatment or tumoural PUFA content. Subanalysis performed after exclusion of patients who were taking concurrent aspirin also showed no difference in vessel density or % vascular area between the EPA and placebo groups (p=0.66 and p=0.76 respectively).
Figure 5.22. Tumour vessel density and % vascular area. Error bars represent geometric mean and 95% confidence interval.
Tumour % vascular area
Subanalysis was performed excluding those patients who were either pre-existing FO users or who were in the highest oily fish consumption category at baseline.
These patients are highlighted in red in Figure 5.23A. When these patients were excluded (Figure 5.23B) and the data re-analysed there was a trend to reduced tumour vascularity in the EPA group compared to placebo. Geometric mean tumour vascular area was 48% lower in the EPA group lower compared to placebo (0.027% vs. 0.052%, p=0.07), and geometric mean tumour vessel density was 43%
lower in the EPA group compared to placebo (16.87 vs. 29.58, p=0.09).
Tumour % vascular area
Figure 5.23. Subanalysis of tumour vascularity. A) Patients who were already taking ω-3 PUFA supplements prior to inclusion in The EMT Trial, or who reported the highest category of oily fish consumption at baseline, are highlighted in red. B) Sub-analysis comparing tumour vascularity in the EPA and placebo groups was then performed after excluding these patients. Error bars represent geometric mean and 95% confidence interval.
5.12 Urinary PGE-M
Urine samples were not analysed from the seven patients withdrawn from the study before undergoing surgery and the one patient who had a histological diagnosis of HCC. Baseline urine samples were therefore analysed for all 80 patients with CRCLM who underwent surgery (EPA n=38, placebo n=42). Pre-operative urine samples were not obtainable from the two patients in the placebo group whose surgery was brought forward without my knowledge. Pre-operatively urine samples were therefore analysed for 78 patients (EPA n=38, placebo n=40). One patient in the EPA group was unable to provide a final urine sample at the final study visit, and final urine samples were not obtainable from the seven patients in the placebo group and two patients in the EPA group who were discharged from the Hepatobiliary Unit's care after recovering from their surgery and did not attend the final study visit. Follow-up urine samples were therefore analysed for 68 patients (EPA n=35, placebo n=33).
Urinary creatinine levels were below the limit of detection (<1.5mmol/l) in four samples in the placebo group and a technical problem with the MS/MS apparatus prevented detection of PGE-M in one sample in the EPA group. PGE-M standardised to urinary creatinine could therefore not be calculated for these samples. Ten samples in the EPA group and 7 in the placebo group had PGE-M levels below the limit of detection (LOD, <10ng/g), and were therefore recorded as this LOD value for the purpose of analysis. Data was log transformed to normalise the distribution for analysis, then summary statistics were converted back to the original scale for presentation in tabular and graphical form.
EPA Baseline
Table 5.10. Urinary PGE-M. Geometric mean and 95% confidence interval (CI) for PGE-M in ng/mg creatinine at baseline, post-treatment (i.e. completion of study medication), and at the six weeks follow-up appointment (off study medication). 1 Paired t-test comparing paired patient data at baseline and post-treatment. 2 Paired t-test comparing paired patient data at post-treatment and follow-up. 3 Unpaired t-test comparing EPA vs. placebo groups at each time point.
There was no significant difference in mean PGE-M between the two groups at baseline (Table 5.11 and Figure 5.24). Post-treatment PGE-M was 51.8% lower in the EPA group compared to placebo (p<0.01). This represented a 27.6% reduction in urinary PGE-M in the EPA group between baseline and post-treatment, and a 27.9% increase in urinary PGE-M in the placebo group between baseline and post-treatment. Urinary PGE-M levels rose again by 52.2% at follow-up in the EPA group (p=0.02), and fell by 18.2% in the placebo group (p=0.14). There was no significant difference in PGE-M between the two groups at the six weeks post-operative follow-up appointment (p=0.89), and no significant change in urinary PGE-M between baseline and follow-up in either group (EPA p=0.61, placebo p=0.60).
There was no correlation between the duration of treatment in the EPA group and the change in urinary PGE-M between baseline and post-treatment (Pearson correlation coefficient 0.059, p=0.73). Sub-analysis by gender revealed no significant differences between the two groups or between male and females (Figure 5.25). Sub-analysis excluding those patients who were already taking ω-3 PUFA supplements prior to enrolment in the Trial or who were high consumers of oily fish did not reveal any significant differences between the two groups.
Subanalysis performed after exclusion of patients who were taking concurrent aspirin also showed no difference between the two groups.
u r i n a r y P G E - M m a l e
PGE-M (ng/mg creat)
Placebo baseline
EPA baseline
Placebo post-treatm ent
EPA post-treatm ent
Placebo follow-up
EPA follow-up 1
1 0 1 0 0 1 0 0 0
u r in a r y P G E - M f e m a l e
PGE-M (ng/mg creat)
Placebo baseline
EPA baseline
Placebo post-treatm ent
EPA post-treatm ent
Placebo follow-up
EPA follow-up 1
1 0 1 0 0
Figure 5.25. Scatterplots of urinary PGE-M stratified by sex. Error bars represent geometric mean and 95% confidence interval.