Abstract
Background: Approximately 30% of all colorectal cancer (CRC) patients are diagnosed with stage II disease. Adjuvant therapy is not widely recommended. However, it is well established that a subgroup of patients with stage II are at high risk for recurrence within their lifetime and should be considered for adjuvant chemotherapy. The present work was designed to study the prognostic value of nuclear DNA content in stage II CRC of patients with long-term follow-up. Patients and Methods: Isolated nuclei from 50 μm-thick paraffin sections of tissue samples from 253 patients with stage II CRC, who had undergone bowel resection at Turku University Central Hospital were cytocentrifuged on slides, stained with Feulgen staining, and DNA was measured using a computer-assisted image analysis cytometry system. Different approaches were applied in analysis of DNA histograms. Results: DNA content did not show any relation with age (p<0.96), sex (p<0.35), tumor invasion (p<0.77), or grade (p<0.31). Aneuploid DNA content was significantly more frequent in the cancer of the left colon and rectum than the right colon (p=0.02). S-phase fraction analysis revealed that a higher proportion (62%) of the older patients (>65 years) had high proliferation rates than did the younger patients (p<0.05). Patients with narrow range histograms had a better disease-free survival (DFS) (narrow range: 70%, wide range: 60% at 10 years). Tumors with >9c nuclei were associated with significantly better DFS and disease-specific survival (DSS) as compared with the patients who did not have >9c nuclei in their tumor samples (p<0.003 and p<0.0001, respectively). Multivariate survival (Cox) model showed that only classification of the basic pattern of the histogram [odds ratio OR)=29.14; 95% confidence interval (CI) 2.350-361.57] (p=0.009) and recurrence (OR=165.35; 95% CI 48.42-564.7) (p=0.0001) proved to be independent predictors of clinical outcome. Conclusion: Our results seem to suggest it truly is possible, by using DNA cytometry, to find groups with different prognosis among stage II cases. Those with a high recurrence rate should be considered for adjuvant chemotherapy.
Colorectal cancer (CRC) is one of the most common malignant tumors worldwide (1, 2), with the disease incidence rising with advanced age (3, 4). The overall mortality from CRC is 60%, which represents the second leading cause of cancer death in Western societies. In Finland, the incidence of CRC is 25/100,000 and 20/100,000 among males and females, respectively. Annually, 1,150 new cases are detected among males and 1,200 among women, representing 9.2% and 10% of all cancer cases, respectively (Finnish Cancer Registry, 2006). Unfortunately, there has been no major improvement in patient survival despite the advances made in our understanding of the risk factors and pathogenesis as well as in development of new chemotherapy practices (5).
Approximately 30% of all CRC patients are diagnosed with stage II disease (node-negative patients) (6). The decision to use adjuvant therapy for patients after curative surgical resection for stage II CRC is often difficult (7) and its routine use is not recommended (8). There is hope, however, that these decisions could be made more rationally in the future, as soon as more solid data are available on disease predictors in these patients, e.g. molecular markers to disclose subgroups of patients suitable for adjuvant therapy. Because the 5-year survival in stage II patients is approximately 70% with surgery alone (9), adjuvant therapy is not widely recommended (10). However, it is well established that a subgroup of patients (20-25%) with stage II CRC are at high risk for recurrence within their lifetime (11) and should be considered for adjuvant chemotherapy (12).
Indeed, some recent studies have suggested a definite benefit from chemotherapy in this subgroup of node-negative patients with stage II disease (7). To date, most of the randomized trials have demonstrated a relative reduction in tumor recurrence but have not shown any significant impact on survival. It seems likely that this failure to demonstrate a survival benefit from adjuvant chemotherapy in stage II disease is due to the fact that these trials do not have enough statistical power due to small patient series. Nevertheless, the absolute survival advantage is only about 2% and clinicians need to weigh this against the costs and toxicities of the treatment when managing these patients (13). Hence it is mandatory to find new prognostic factors capable of identifying high-risk stage II patients for better targeting of treatment options (14).
As a part of our systematic search for prognostic factors in CRC (15), the present work was designed to study the prognostic value of nuclear DNA content in stage II CRC of patients with long-term follow-up. Particular emphasis was placed on the different methods of analysing the histograms provided by static DNA cytometry and their utility as potential prognostic tools.
Patients and Methods
Patients, treatment and follow-up. The material of the present study comprised tissue samples from 253 patients with stage II CRC who had undergone bowel resection in 1981-1997 at Turku University Central Hospital, collected from the archives of the Department of Pathology. Pertinent clinical and histopathological data of the patients were collected from the patients' case records and linked with the image cytometry data. The key clinicopathological data of the patients are summarized in Table I.
The patients were seen at 6-monthly intervals in the Clinic, followed up until death or when last seen alive (end of 2007). The mean follow-up time for the whole series was 94.7 months (SD=66.1, median=93.9, range 1-260 months). During the follow-up, 94/253 patients (38%) developed recurrence within the mean of 26.6 months (median=15.3 months). Disease-free survival (DFS) and disease-specific survival (DSS) were calculated as usual, i.e. i) time from diagnosis to the appearance of recurrent disease or to the date last seen disease-free, and ii) time from diagnosis to death (due to disease) or to the date last seen alive, respectively. In calculating DSS, patients who died of other or unknown causes were censored from the survival analysis.
Feulgen staining. Altogether, 253 formalin-fixed paraffin-embedded tumor samples were obtained for analysis. All analyses were conducted using the same tissue blocks from which the original histological diagnosis was made. Only biopsies with representative sample of cancer tissue were accepted. There were also benign stromal diploid cells in the samples but they were in a minority. Firstly, 5 μm-thick paraffin sections were cut and stained with hematoxylin and eosin (H&E) for confirmation of the histological diagnosis.
For the Feulgen staining, nuclei were isolated from 50 μm-thick paraffin sections according to the method described by Hedley et al. (16). The sections were placed in glass tubes and dewaxed in xylene, followed by rehydration by sequential immersion in 100%, 95%, 70% and 50% ethanol, washed in water, and treated with 0.5% pepsin (activity 2,500-3,200 units per mg protein adjusted to pH 1.5) in a 37°C water bath. The isolated nuclei were cytocentrifuged on glass slides and stained with Feulgen stain according to Gaub et al.'s method (17). The samples were washed in distilled water, followed by acid hydrolysis in 5 M hydrochloric acid at room temperature (20°C) for one hour. After washing in distilled water, samples were treated in darkness with Schiff's reagent (stain: pararosaniline) for 2 hours 45 minutes at room temperature (20°C), rinsed in distilled water, treated for 3×10 minutes in fresh aqueous sodium thiosulphate (180 ml distilled water, 10 ml 1 M HCl, 10 ml 10% Na2S2O5), and rinsed for 5 minutes. The Feulgen reaction labeled the DNA as magenta. The intensity of the stain is directly proportional to the amount of DNA present. After dehydration, the smears were treated with xylene, mounted and stored in light-tight boxes.
Image analysis cytometry. The intensity of Feulgen staining was measured using a computer-assisted image analysis cytometry system AHRENS ICM with a Nikon microscope (Eclipse E 400; Japan) (designed and produced by Olaf Ahrens; Meßtechnische Beratung, Bargteheide/Hamburg, Germany). The field of view from the camera (JAI DSP surveillance color CCD camera, CV-S 3200/3300) was stored in image memory with resolution of 736 by 560 pixels. The image was produced by a plan objective (Nikon; ×40) and the measurements were made from that image. Prior to each measurement session, the illumination of the microscope was adjusted according to the method of Köhler (18). Several histograms were produced twice and were found to be very similar.
Interpretation of the DNA histograms. Feulgen-stained samples were available from the primary tumors of all 253 patients. The whole slide was scanned at ×10 to find the most suitable area or areas for analysis, with atypical cells, when possible. The selected area or areas were scanned at ×40, and an average of 10-15 microscopic fields were selected, and at most 200 consecutive tumour nuclei with clear nuclear borders were outlined and measured. Overlapping or cut (from surfaces of the 50 μm section) nuclei were not measured. In addition, 30 small lymphocytes were measured as internal controls. The internal control peaks represent the location of the 2c diploid peak (c=haploid DNA content). The control peak was narrow with limited variation (coefficient of variation <5%).
In this study, we used different approaches to analyse the histograms obtained in DNA cytometry (Table II). The ABCDE system was a classification based on the appearance of the histogram and was used as previously described (19).
In range evaluation, the width of the histograms was originally divided into 12 gates, as listed in Table II. The first 6 were also called low-start gates (1.8c upwards) (=1st set), and the last 6 high-start gates (3.0c upwards) (=2nd set). In the original study, the absolute width of the histogram was measured and, based on the widths of the gates, the samples were divided into 2 categories (A and B) as follows: A (width 1.4 - 6.0 units, narrow range) consisted of gates 1.8c - 4.4c, 1.8c - 5.0c, 1.8c - 7.0c, 3.0c - 4.4c, 3.0c - 5.0c, 3.0c - 7.0c, 3.0c - 9.0c; and B (width 6.1->10.0 units, wide range) consisted of gates 1.8c - 9.0c, 1.8c - 10.0c, 1.8c - >10c, 3.0c - 10.0c, 3.0c - >10c. This original 2-tier grading system (19) was modified in the present study, by setting the cut-off for histogram width to 3.0. In addition, a 3-tier grading system was tested classifying the width of the histograms into narrow (1.4-3.0 units), intermediate (3.1-6.0 units) and broad (6.1->10.0 units) range categories, following the same principles. All the histograms could be fitted into these categories.
In the DNA cut-off approach, we calculated the fraction of the cells above five pre-defined cut-off points (5c, 6c, 7c, 8c, 9c), expressed as percentages, as shown in Table II. We tested the reproducibility of each approach of histogram evaluation on two separate occasions separated by 4 weeks, and obtained good intraobserver reproducibility (89% up to 100%).
Clinicopathological characteristics of the patients.
Finally, the S-phase fraction (SPF) as determined by the 10% cut-off for proliferating cells was used to distinguish two groups with different disease outcome.
Statistical analysis. Statistical analyses were performed using the SPSS® for Windows, version 16.0.1 (SPSS, Inc., Chicago, USA) and STATA/SE 10.2 (Stata Corp., Texas, USA) software packages. Frequency tables were analysed using the Chi-square test, with likelihood ratio (LR) or Fisher's exact test being used to assess the significance of the correlation between the categorical variables. Differences in the means of continuous variables were analysed using non-parametric tests (Mann-Whitney) or ANOVA (analysis of variance). Logistic regression models were used to analyse the power of different variables as predictors of the outcome variables, calculating crude odds ratios (ORs) and 95% confidence intervals (CI) in univariate analysis. Univariate survival (life-table) analysis for the outcome measure (DSS, DFS) was based on Kaplan-Meier method. In all tests, values of p<0.05 were regarded as statistically significant.
Results
Different evaluation approaches and clinicopathological features. By using the ABCDE approach, DNA content did not show any relation with age (p<0.96), sex (p<0.35), tumor invasion (p<0.77), or grade (p<0.31). When divided into two categories according to their localization (a) right colon vs. b) left colon and rectum, tumors with aneuploid DNA content were significantly more frequent in the left colon and rectum than in the right colon (p=0.02). By using cut-off points (>5c and >9c) to evaluate the tumors of the right colon and left colon, a higher proportion of tumors with more than 5c DNA content was located in the left colon and rectum than in the right colon (p<0.008), implicating more aneuploidy in tumors of the left side.
The different approaches used to analyse the DNA histograms.
Of the other approaches, SPF analysis revealed that more (62%) of the older patients (>65 years) had a high proliferation rate (more than 10% SPF) as compared with that (45%) among the younger patients (p<0.05).
Different evaluation approaches and survival. In the ABCDE method, we estimated the effect of 5 different categories of DNA histogram patterns on disease outcome. The DFS associated with the tetraploid (B) histograms and the aneuploid (C, D) histograms was very similar, with 10-year DFS of 60% and 65%, respectively. Tumors with diploid histograms were associated with worse prognosis, with a 10-year DFS of only 50% . In univariate (Kaplan-Meier) survival analysis, however, the differences in DFS were not statistically significant (p=0.320).
Disease-free survival related to presence or absence of >9c (presence: >1.0, absence: <1.0).
Disease-specific survival related to presence or absence of >9c (presence: >1.0, absence: <1.0).
Disease-free survival related to graded 9C cut-off below and above 1.0.
Disease-specific survival related to graded 9C cut-off below and above 1.0.
Survival after recurrence was dismal among patients with diploid (A) or aneuploid (C, D) histograms, only one of these patients being alive after 5 years. Interestingly, several (n=5) patients with tetraploid histogram survived longer than 5 years after detection of recurrence, and 3/33 (9.1%) of them were still alive after 10 years of disease recurrence. In Kaplan-Meier survival analysis, however, these differences were not significant (p=0.100).
DSS was significantly (p=0.024) related to histogram patterns in univariate survival analysis. DSS was markedly better for tetraploid and aneuploid cases, with about 55% survival at 10 years of follow-up, 52.3% and 47.5% of patients, respectively, being alive at the end of follow-up. On the other hand, the survival curve of patients with diploid histograms behaved very differently: at around 5 years, there was a dramatic drop in survival, resulting in 15% of 10-year DSS and 7.1% overall DSS.
Disease outcome related to histogram width was analysed using the 2-tier system with 3.0 cut-off. There was a difference in DFS between the cases with narrow range and wide range histograms, in that the patients with a narrow histogram had a better DFS (narrow range: 70%, wide range: 60% at 10 years), but the difference was not statistically significant (Figure 1). Similarly, no statistical difference (p=0.410) between the cases with narrow and wide range was observed in their survival after recurrence in Kaplan-Meier analysis (data not shown).
The difference in DSS between the patients with narrow and wide range was very similar to that of DFS, i.e. around 10% (Figure 2). Again, the patients with the narrow range had better survival, but the difference was not statistically significant (p=0.199). Of the patients with narrow range histograms, 60.9% were alive at 10-years as compared with 47.8% of those with wide range histograms (p=0.240).
The most dramatic effect on disease outcome was observed when the DNA cut-off approach (i.e. DNA-values above cut-off point clearly higher than 9c) was used. Somewhat unexpectedly, tumors with >9c nuclei were associated with significantly better DFS (Figure 3) and DSS (Figure 4) as compared with the patients who did not have >9c nuclei in their tumor samples. The difference is highly significant, p<0.003 and p<0.0001, respectively.
The S-phase fraction (SPF) as determined by the 10% cut-off for proliferating cells distinguished 2 groups with different disease outcome as well. The patients with SPF >10% tended to have shorter DFS (p=0.094), DSS (p=0.113) and survival after recurrence (p=0.868) in Kaplan-Meier analysis. This difference was significant (p=0.048) for 5-year DSS, less significant (p=0.104) for 10-year DSS and not significant for 5-year survival after recurrence (p=0.562).
Finally, a multivariate survival (Cox) model was used to test the independent prognostic predictors of DSS in this study, controlled by the possible confounding effects of age and sex. When the following factors were entered into the model: age, sex, lymph node involvement, localization, and recurrence (together with the cytometric variables), only the ABCDE approach (OR=29.14; 95% CI 2.350-361.57) (p=0.009) and recurrence (OR= 165.35; 95% CI 48.42-564.7) (p=0.0001) proved to be independent predictors of clinical outcome.
Discussion
The role of DNA content as a prognostic factor in CRC is highly controversial (20). While several studies have suggested that DNA ploidy is an independent prognostic factor (21-23) others have reported that DNA content is not associated with clinical outcome (24, 25). Some of these discrepant observations might be explained by differences in the technical aspects of recording the DNA contents or by differences of interpretation of the DNA histogram. This is also clearly shown by our previous study, where only some of the analytical approaches used to assess the DNA histograms resulted in prognostically valuable information (19).
The present series is unique in that the follow-up of the patients covers an unusually long period. Accordingly, we could calculate 5- and 10-year survival figures with good statistical power because enough cases with substantially longer follow-up were included in the cohort. This makes the series different from many of the published CRC studies, where only short survivals are reported. In addition, we used several different, in part complementary, approaches to analyse the image cytometric data.
In this study, we focused on stage II CRC, where molecular and other markers may help in identifying a subgroup of patients who might benefit from the use of adjuvant chemotherapy. The results clearly suggested that analysis of nuclear DNA content can provide such prognostic information.
In this cohort, none of the different approaches used to analyze the cytometric data disclosed any clinically useful correlations with the clinicopathological data, including age, sex, tumor invasion, grade and metastasis. This is contrary to our preliminary study of a small cohort which showed that the fractions of cells above each of the >5c cut-offs were always higher among the younger patients, and the difference was significant (p=0.021) (19). In this series, the only correlation with age was shown by the DNA histogram-based SPF analysis in which older (>65 years) patients had tumors with higher proliferation rates than the younger patients (p<0.05).
Interesting differences were found in these cytometric variables related to tumors at different localizations. It has been suggested that biological differences exist between tumors of the proximal and distal colon (26, 27). Indeed, mutations of the p53 gene were found to be more prevalent in tumors of the distal colon (28). Most studies found have reported that DNA diploid tumors are more frequent in the proximal colon than in tumors distal to the splenic flexure (29, 30). Furthermore, some of the allelic deletions and p53 gene mutations were more frequent in DNA aneuploid than in DNA diploid colorectal tumors (31, 32), whereas tumors with microsatellite instability were characterized by diploid nuclear DNA content (33, 34). The results of the present study provide further support to these observations, while disclosing that the tumors in the left colon and rectum had the highest frequency of cells with aneuploid DNA content (75%) as compared to the right colon, which showed only 25% of tumors with aneuploid DNA content.
In the present study, there were some observations that were somewhat unexpected in the light of previously published data (19). In our analysis, there is one feature which, in efficiency of survival prediction, successfully competes with the other recent powerful approaches such as cDNA profiling (35), or connective tissue fraction (36). In our database, the fraction of cases with nuclei of DNA >9c was highly significant and suggested that patients with nuclei with DNA >9c had a better prognosis than those without DNA >9c cells. This finding may suggest that patients with tumors which preserve polyploidic tendencies, with diploid or aneuploid baseline, may show better prognosis.
The DNA cytometric findings show that patients with diploid cancer cell DNA histogram had worse prognosis on average than patients with aneuploid cancer. It was rather problematic for us to explain this difference, but on the basis of known data on the development of CRC, we feel that the best explanation is hidden in the fact that CRC tumors with microsatellite instability are generally diploid. Because of the instable genome, these types of cancer may be more active in gaining new mutations, allowing faster progression than aneuploid tumors without microsatellite instability. The data on survival of microsatellite-instable tumors also suggests this (37, 38).
Also significant were aneuploid tumors at 15 years: some patients had surprisingly long survival. Patients with tumors with SPF >10% had a worse prognosis than those with tumors with SPF<10%, but the difference was not as clear as when groups with DNA >9c or DNA <9c were compared. It now seems that it is truly possible to find groups with different prognosis among stage II cases. Whether the patients in diploid/aneuploid groups differ or are the same in the DNA >9c group, high connective tissue fraction group, or the ominous cDNA expression pattern group (35) remains to be shown by further research.
One of the great problems in DNA cytometric data is related to the narrow and wide range histograms. Why are the narrow histograms associated with better prognosis? At the moment we cannot give a good explanation. Tumors with wide-range histograms might hide a heterogenic group of cancer cells as compared to tumors with narrow-range histograms which most likely are homogeneous.
One has to remember that it is potentially possible to combine the results of several tests. One can produce a DNA cytometric index by combining significant features found in DNA cytometry, or a combination index by combining results of connective tissue fraction studies, cDNA data and static DNA cytometry. Clearly, further studies on the mentioned issues are warranted.
Footnotes
- Received June 28, 2008.
- Revision received November 20, 2008.
- Accepted December 4, 2008.
- Copyright© 2009 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.