Proliferation indices
Immunohistochemical methods for the assessment of cell proliferation in small biopsies suffer from a number of potential disadvantages. Firstly, it can be difficult to obtain high yields of assessable crypts in histological sections. Secondly, there is concern over the impact of the so called "denominator effect" on the LI. Counting mitotic figures in whole-mount preparations avoids these issues, but does suffer the disadvantage that a second biopsy must be obtained if conventional histolological examination (or additional immunohistochemical staining, e.g. for apoptosis) is required. We compared the statistical reliability of both approaches using the Cronbach's alpha statistic -a measure of between and within-case variability, ranging from 0 to 1 with increasing reliability. The scores were 0.9476 and 0.9164 for Ki-67 and whole mount mitotic counts, respectively, indicating that robust measurements can be obtained without exhaustive sampling of crypts. Moreover, scores from both methods correlated well (Pearson correlation coefficient 0.655, p = 0.008). There is also concern that individual antibody targets may not display expression that is truly restricted to proliferating cells, and therefore give potentially spurious results. It is reassuring that the immunohistochemical markers used in this study both showed similar results.
Two immunohistochemical markers of cell proliferation, Ki-67 and pHH3, were measured. Colonocyte Ki-67 LI decreased after preparation with Klean-Prep, when compared with unprepared bowel from the same patients (Figure 1A). Despite the consistent change, significance was not achieved (p = 0.117, paired t-test). A decrease was also seen in the number of whole-crypt mitoses with Klean-Prep (Figure 1B). Again, the trend was consistent but significance was not achieved (p = 0.118). Klean-Prep had no apparent effect on pHH3 diffuse or granular staining, when compared to that seen in the same patients without preparation (p = 0.519 and p = 0.562, respectively, data not shown).
Picolax had an opposing effect on colonocyte proliferation. Ki-67 LI increased after preparation with Picolax, although this effect was not significant (p = 0.2778, Figure 2). Similarly, increased diffuse and granular pHH3 staining was observed with Picolax and reached significance in the granular fraction (p = 0.049, p = 0.269 for positive staining, data not shown). Paired samples from only one patient were available for whole crypt staining; the mean number of mitoses per crypt for this patient were 11.4 and 5.0 for Picolax and unprepared biopsies, respectively.
While obvious and consistent trends are seen in the above analysis, treating the samples as paired for statistical purposes is undermined by the small numbers (n = 4 for Klean-Prep, n = 3 for Picolax) and the unavailability of paired biopsies for all markers tested. All data, paired or otherwise, was therefore redistributed into treatment groups and subjected to analysis of variance (ANOVA) testing. One-way ANOVA revealed significant differences between groups (Ki-67: p = 0.035, mitoses: p = 0.045), demonstrating, as for the paired analysis, a decrease in proliferation with Klean-Prep, and an increase with Picolax, when compared to unprepared bowel (Figure 3). Scoring differentially for granular and positive pHH3 staining showed a similar pattern of response to bowel preparation (data not shown), but statistical significance was reached when the data sets were combined to give a total LI (p = 0.038, Figure 4). Post-hoc analysis revealed a significant difference between Klean-Prep and Picolax on Ki-67 and total pHH3 LI (p = 0.019 and 0.017, respectively).
This observation may be relevant to the interpretation of proliferation data in several recently published studies. In one study demonstrating no significant difference in Ki-67 or whole crypt mitoses between patients with Hereditary non-polyposis colorectal cancer and normal subjects, the authors concluded that crypt cell proliferation is not a suitable discriminative marker for this disease[15]. All patients had been administered Picolax and, as in our study, had Ki-67 LIs in the range 0.3 to 0.4[16]. A study on the effect of pre-biotic carbohydrates similarly showed no difference in Ki-67 LI between subjects and controls, reporting indices between 0.3 and 0.4. Again, all subjects received Picolax. Our evidence suggests that any marginal change in LI in response to disease or intervention in these studies may have been masked by the proliferative effects of Picolax. Equally, data from studies based on subjects prepared with PEG should be interpreted cautiously. No association was found between the PCNA proliferative index (PI) and the likelihood of developing adenoma in one prospective study; the majority of these patients were prepared with PEG[17]. It is possible that PEG suppressed colonocyte proliferation in a group of patients that might be expected to have a higher than normal PI.
Crypt length and cellularity
No significant differences were seen by t-test analysis between pairs for either Klean-Prep or Picolax (p = 0.766 and 0.183, respectively), or between groups by one-way ANOVA (p = 0.437). Similarly, there was no significant difference in crypt length between pairs (p = 0.209 for Klean-prep, p = 0.827 for Picolax), or between groups by one-way ANOVA (p = 0.0786) (data not shown). The cellular homeostasis observed despite increased proliferation could be attributed to two factors. Firstly, the interval between bowel preparation and biopsy (typically 12 hours) may have been insufficient for any cells entering cell division to complete the cycle. Secondly, the observed homeostasis may be attributable to increased apoptosis. Our assessment of apoptotic indices recorded very low rates of background apoptosis in all samples, and showed no alteration in response to bowel preparation (data not shown).
Limitations of study
A randomisation bias on clinical grounds cannot be ruled out in the allocation of patients to Klean-Prep or Picolax preparation; this, along with the small sample size, may in part account for the lack of statistical significance seen between prepared patients and unprepared controls. Similarly, whilst patients prepared with Klean-Prep or Picolax fasted prior to colonoscopy, control patients did not, and potential confounding influences of luminal nutrition and luminal workload on proliferation indices should be noted. However, our analyses, and the central finding of this study, show an opposing direction of change between Klean-Prep and Picolax that is consistent for all proliferation measures, and cannot be a feature of starvation. Although the small number of cases and the single sampling site inevitably means that the results should be interpreted with caution, this pilot study has informed our choice of procedural homogeneity for a separate, larger study requiring repeat endoscopies.