Introduction

With an estimated incidence of 1,023,000 new cases annually, colorectal cancer (CRC) represents the fourth most common malignant solid tumour worldwide [1]. About 20–25% of patients present metastatic disease at the time of initial diagnosis. In addition, up to 50% of patients with curatively resected CRC will develop metachronous metastatic disease. Palliative chemotherapy has shown to improve overall survival (OS) and quality of life. To date, several chemotherapy schemes including infusional and oral 5-fluorouracil (5-FU), irinotecan and oxaliplatin can be applied. Recently, targeted therapies with monoclonal antibodies against the vascular endothelial growth factor and the epidermal growth factor receptor have also become available [2, 3]. This increase of active drugs enables the choice between numerous therapeutic options. However, reliable prediction of tumour response, prognosis and toxicity for each therapy scheme for an individual patient is presently not yet available. Consequently, molecular markers which may predict the clinical course or toxicity of a certain patient may guide treatment choices.

One of these potential prognostic factors is high-level microsatellite instability (MSI-H). MSI-H is related to a deficient mismatch repair (MMR) system [4, 5]. Such a defect derives from a somatic hereditary mutation of a MMR gene or from hypermethylation of the MLH1 promotor. Germline mutations in MMR genes in the context of Lynch syndrome account for 2–5% of CRC [6]. An additional 12–15% of CRC display MLH1 promoter hypermethylation [7–11]. MSI-H CRC have characteristic clinical and histopathological features. In particular, MSI-H CRC has a favourable prognosis compared to non-MSI-H CRC [12]. However, the prognostic value of MSI-H has been determined mainly in surgical cases. The value of MSI-H in surgically resected colon cancers and adjuvant 5-FU-based chemotherapy remains controversial. While earlier studies suggested a benefit from adjuvant chemotherapy for MSI-H CRC [13, 14], more recent studies did not find a benefit from adjuvant chemotherapy for MSI-H CRC compared to surgery only [15–18]. The latter studies are in line with in vitro studies regarding chemoresistance to 5-FU for MSI-H CRC cell lines [19, 20]. In addition, Arnold et al. demonstrated an increased chemosensitivity after demethylation treatment in CRC cell lines with MLH1 promoter hypermethylation [21].

The data regarding the value of MSI-H in metastatic CRC undergoing 5-FU-based palliative chemotherapy are very limited. The results from four studies suggested mainly a benefit from 5-FU-based palliative chemotherapy [22–25].

Most recently, des Guetz et al. reported for the first time upon the prognostic value of MSI-H in patients treated with palliative first-line combination chemotherapy with 5-FU/oxaliplatin (FOLFOX) in 44 patients with metastatic CRC [26]. The authors did not find a significant difference between progression-free survival (PFS), OS and response rates (RR) between MSI-H and microsatellite stable (MSS) cases in this small series of patients not included in a clinical trial.

In vitro data suggest a chemoresistance of cancer cell lines with MMR deficiency upon treatment with cisplatin and carboplatin but not upon oxaliplatin [27]. These results were later confirmed by the same group using a xenograft model [28].

To further evaluate the prognostic and predictive value of MSI-H in metastatic CRC treated with 5-FU/oxaliplatin-based first-line chemotherapy, we investigated 108 tumour samples from patients included in a randomised prospective trial. We first aimed to investigate the incidence of MSI-H by performing microsatellite analysis and immunostaining of the MMR proteins MLH1, MSH2 and MSH6. Secondly, MSI-H was correlated to clinical data to evaluate its predictive and prognostic value for tumour response and PFS and OS.

Materials and methods

Study design

Tissue samples were derived from patients with metastatic CRC participating in a prospective randomised phase III first-line chemotherapy trial of the AIO Group (Arbeitsgemeinschaft Internistische Onkologie of the German Cancer Society) [29]. A total of 474 patients were randomised to be treated with either 5-FU/folinic acid (FA) and oxaliplatin (FUFOX: oxaliplatin, 50 mg/m2; FA, 500 mg/m2; continuous 5-FU, 2,000 mg/m2/22 h; on day 1, 8, 15, 22; q day 36; n = 51) or capecitabine and oxaliplatin (CAPOX: oxaliplatin, 70 mg/m2 on day 1 and 8; capecitabine, 2 × 1,000 mg/m2/day consecutively for 2 weeks, q day 22; n = 52). Eligibility and exclusion criteria are published elsewhere [29]. The study was approved by the ethical committee of the Central Hospital Bremen and of the Medical Faculty of the Ruhr-University Bochum.

Clinically, the combination of capecitabine and oxaliplatin showed no significant difference in terms of RR, PFS (as the primary endpoint) and OS when compared to 5-FU/FA/oxaliplatin. Porschen et al. found an overall RR of 48% in the CAPOX arm and 54% in the FUFOX arm (p = 0.7). The median PFS was 7.1 months (CAPOX) and 8.0 months (FUFOX, p = 0.117) and the median OS was 16.8 months (CAPOX) versus 18.8 months (FUFOX, p = 0.26). These data are consistent with the findings of other trials [30, 31]. Moreover, no clinical factor was found to be predictive for definition of a subgroup of patients benefiting more or less from each fluoropyrimidine backbone [29].

Patient characteristics

This analysis includes a subgroup of 108 patients (23% of all patients) from the clinical trial, recruited from July 2002 to July 2004, of whom tumour tissue were available. The median follow-up time was 16.5 months (range, 1–37 months). Patients’ characteristics were: 62 male, 42 female, with a median age of 64.7 (range 38–81) years. Fifty-seven patients had synchronous metastases, and 43 had metachronous metastases. In four patients, the date of appearance of metastasis was not recorded.

As expected in metastatic CRC (mCRC) cohorts, one third of patients had primarily rectal cancer whereas the primary tumour of the majority was in the colon (Table 1). The investigated tissue samples included primary tumours (n = 84) as well as metastases (n = 8). In 12 patients, the origin of the tissue was not recorded.

Table 1 Clinicopathological patient characteristics
Full size table

Immunohistochemistry

Immunostaining and interpretation were performed as described previously [32]. As primary antibodies, mouse monoclonal antibodies for MLH1 (BD Biosciences, 1:20), MSH2 (BD Biosciences, 1:20) and MSH6 (BD Biosciences, 1:50) were applied. Staining was considered only informative when there was normal nuclear staining in adjacent non-neoplastic cells, which served as positive internal control. The part of positive stained nuclei was determined semi-quantitatively by light microscope and classified as follows: <1% represented a lack of expression of the MMR proteins; 1–10% showed a reduction of expression; >10% indicated a regular expression. Immunohistochemistry of MLH1 and MSH2 has a sensitivity of 92% to 95% and specificity of 99% to 100% to detect MSI-H [33–35].

Microsatellite analysis

Tumour and normal tissue were microdissected by a skilled pathologist. DNA was isolated with the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). Microsatellite–polymerase chain reaction (PCR) reactions were performed as described previously [32]. To screen for microsatellite instability (MSI), all specimens were investigated with the microsatellite marker Bat-26 first. Cases with instability of Bat-26 were tested with the complete National Institute of Health reference marker panel (Bat-25, D5S346, D17S250, D2S123) [36]. Tumour tissue without normal tissue available (n = 3) was exclusively investigated with Bat-26 and Bat-25. The exclusive application of Bat-26 has shown a sensitivity for detection of MSI-H of about 95% [37]. Tumours were classified as MSI-H if at least two of five markers showed instability. MSI-L was defined if instability was evident in only one marker. Tumours were classified as MSS in cases where no marker was instable.

Statistical analyses

Fisher’s exact test was used to evaluate the associations between MSI and other dichotomous variables. Student’s T-test was performed to calculate the relationship between continuous variables and dichotomous variables. Significant differences between a continuous variable and different groups were identified by the Kruskal–Wallis test. The Kaplan–Meier method was used to determine the OS and the PFS [38]. The OS and the PFS of the subgroups (MSI-H positive vs. MSI-H negative) were compared by the log-rank test. All tests were two sided. For all tests, p values <0.05 were considered as significant.

Results

Incidence of microsatellite instability

Of the 108 patients included in this study, 89 (83%) tumours were analysable by PCR-based MSI testing. Additional 15 cases were evaluable by immunohistochemistry of MLH1, MSH2 and MSH6. Tissue samples from four (4%) patients were not analysable due to low DNA quality and staining artefacts (no staining of normal tissue). Among 104 analysable tumours, we identified four (4%) MSI-H tumours and one (1%) MSI-L tumour. One tumour exhibited a reduction of MSH6 while displaying MSS. The latter two cases were analysed within the MSS group.

Correlation between MSI and clinicopathological variables

Four (8%) out of 51 patients treated with FUFOX had a MSI-H tumour, while no patient out of 53 treated with CAPOX had a MSI-H tumour (p = 0.054) MSI-H was not correlated to any other clinicopathological variables recorded (age, sex, localisation, syn- vs. metachronous metastatic disease) (data not shown).

Correlation between MSI and remission status

Tumour response evaluation yielded that one of four MSI-H-positive patients had a WHO-classified complete remission and one more patient had a stable disease whereas two (50%) patients showed progressive disease. Among 100 non-MSI-H tumours, three (3%) patients had a complete remission and 52 patients (52%) had a partial remission. Thirty-one (31%) patients exhibited a stable disease and four (4%) had disease progression. No significant relation could be found with regards MSI-H status when comparing patients with objective response (PR + CR) to patients with stable disease or progressive disease (p = 0.30).

In contrast, MSI-H-positive patients compared with non-MSI-H patients actually showed a significant disadvantage considering the combined outcome marker with disease control (defined as PR, CR or SD) versus progressive disease (PD). Disease control was observed for 50% of MSI-H-positive and 95.6% of non-MSI-H patients (p = 0.019). PD appeared in 50% of patients with MSI-H-positive and 4.4% of patients with non-MSI-H tumours (Table 2).

Table 2 Treatment response in correlation to MSI-H status
Full size table

Correlation between MSI and progression-free survival and overall survival

This subcohort of patients, representing approximately 25% of all patients included in the clinical trial, had a median PFS of 7.8 months (95% CI, 5.8–9.8 months) and a median OS of 18.9 months (95% CI, 15.8–22.0 months) which were not significantly different from the data from the whole clinical trial [29]. PFS and OS were significantly better in patients with metachronous metastases as compared to patients with synchronous metastases (Table 3). The other patient characteristics did not demonstrate any relation to PFS or OS, respectively.

Table 3 Overall survival (OS) and progression-free survival (PFS) in correlation to MSI status and clinical variables
Full size table

During the follow-up time, 88 patients showed disease progression of which three patients had an MSI-H and one patient an MSI-L tumour. Sixteen patients, including one patient with an MSI-H tumour, had no disease progression at last follow-up. The median PFS was 7.9 months for non-MSI-H patients (95% CI, 6.1–9.74 months) and 2.5 months for patients with MSI-H tumours (95% CI, 0.9–4.1 months) (p = 0.59) (Table 3; Fig. 1).

Fig. 1
figure 1

Progression-free survival of MSI-H-positive patients compared to non-MSI-H patients (MSS/MSI-L)

Full size image

At the time of last follow-up, 69 patients died, including one patient with MSI-H tumour, one patient with MSI-L tumour and 67 with non-MSI-H tumours. Thirty-five patients were alive, including three patients with MSI-H tumours (75%) and 32 patients with non-MSI-H tumours. The median OS for patients with non-MSI-H tumours was 18.9 months (95% CI, 15.7–22.2 months). The MSI-H-positive patients did not reach the median during the follow-up. However, the difference was not statistically significant for OS (p = 0.13) (Table 3; Fig. 2).

Fig. 2
figure 2

Overall survival of MSI-H-positive patients compared to non-MSI-H patients (MSS/MSI-L)

Full size image

Discussion

So far, only a few studies investigated the incidence of MSI-H in metastatic CRC and its correlation to prognosis and treatment response. We therefore analysed data from a cohort of 25% patients included into a prospective randomised trial treated with oxaliplatin in combination with intravenous versus oral 5-FU, with respect to the incidence and prognostic and predictive value of MSI-H.

Incidence of microsatellite instability

The incidence of MSI-H in unselected CRC is reported to range from 12–18% [36]. The incidence of MSI-H in our study was 4%. Our data are consistent with most previously published data regarding metastatic CRC [39, 40]. However, some studies reported a markedly higher incidence of MSI in metastatic CRC [24, 41, 42]. The discrepancy of some studies with a higher incidence of MSI to our data might be a result of a variable definition of MSI-L/MSI-H, use of different markers or a consequence of selection of patients. The inclusion criteria of patients for the prospective clinical trial may have caused some selection bias, since only patients with ECOG 0–1 performance status were eligible for the trial. In addition, we preferentially analysed tumour tissue from the primary colorectal cancer (n = 84) rather than the metastases. However, this fact was unlikely to cause a bias since MSI-H is an early event in colorectal carcinogenesis.

Correlation of MSI-H and prognostic variables

Only a few studies have addressed the prognostic value of MSI-H in metastatic CRC treated with 5-FU-based chemotherapy. Liang et al. detected a significant favourable outcome of MSI-H patients treated with 5-FU-based chemotherapy (n = 169), regarding response rate (65.7% vs. 35.1%; p = 0.001) and median OS (24 month vs. 13 month; p = 0.0001), when compared to non-MSI-H patients [23]. For patients without chemotherapy (n = 75), there was no correlation between MSI status and survival. The authors concluded that the predictive value might be explained by a higher chemosensitivity of MSI-H-positive tumours but not by a lower aggressiveness of tumour growth. In addition, Brueckl et al. observed a significant better median OS for patients treated with a 5-FU-based chemotherapy (33 months vs. 19 months, p = 0.021) and a higher remission rate (72% vs. 41%; p = 0.072) for MSI-H-positive patients (n = 7) compared to non-MSI-H patients (n = 36) [22].

The first study which reported the correlation of MSI-H and the outcome of combination chemotherapy with FOLFOX 4 or FOLFOX6 first-line palliative chemotherapy in 44 patients metastatic CRC was published while we were conducting our analyses [26]. The authors did not find a significant difference between PFS, OS and RR between MSI-H (n = 9) and MSS (n = 31) cases. However, the limited sample size (n = 44) has to be considered. In addition, it has to be noted that patients were not treated within a controlled trial with defined inclusion and exclusion criteria which may have introduced unintended bias.

In our study, we did not observe a favourable or unfavourable prognosis of MSI-H-positive patients compared to non-MSI-H patients with respect to PFS and OS. However, our analysis must be interpreted bearing in mind the very low number of MSI-H cases.

Considering disease control (defined as CR, PR and SD) versus PD, we found a significant disadvantage of MSI-H-positive patients compared to non-MSI-H patients. If this finding could be confirmed in a larger set of patients, these findings differ from previous studies analysing 5-FU-based protocols and may reflect a possible unfavourable outcome for oxaliplatin administration for this subgroup.

However, if the low number of MSI-H-positive patients is consistent in other prospective mCRC trials, the use of this marker is of limited value. Furthermore, other potential prognostic markers such as p53, SMAD4, thymidylate synthetase and K-ras mutations should be included in multivariate analyses.