Is serum matrix metalloproteinase 9 a useful biomarker in detection of colorectal cancer? Considering pre-analytical interference that may influence diagnostic accuracy

Sir,

I read with great interest the article by Hurst et al (2007), which was recently published in the Br J Cancer. The authors described elevated serum concentration of matrix metalloproteinase 9 (MMP-9) in patients with colorectal neoplasia compared with the MMP-9 concentration in symptomatic patients without non-neoplastic conditions. On the basis of this result of increased serum MMP-9 concentration the authors developed a logistic regression model including age, sex, smoking history, abdominal pain, and weight loss as additional factors for the prediction of malignant colorectal disease. The authors concluded that increased serum MMP-9 concentrations could be useful to stratify patients into those with low- and high-risk of malignancy to spare patients unnecessary colonoscopy. Some of the authors of this article already elaborated detailed study protocols including sample size calculation, potential selection bias, confounders, methods of analyses study protocols to evaluate the potential of serum MMP-9 as a screening test for colorectal cancer (Ryan et al, 2006; Wilson et al, 2006). Apparently based on the pilot data described in the mentioned report of Hurst et al (2007) and the prevalence data of colon cancer, it was calculated that 23 100 people from 29 practices have to be recruited to identify the 700 participants needed for this study (Wilson et al, 2006).

However, I have the impression that the authors overlooked the fundamental problem of pre-analytical conditions for accurate MMP-9 measurements. They did not pay attention to the potential pre-analytical pitfall of blood sampling to measure true concentrations of circulating MMP-9 as serum instead of plasma was used for MMP-9 measurements and also recommended for this ambitious study. The misuse of serum as sample for determining circulating MMP-9 in peripheral blood has been frequently criticised both in analytical and clinical journals (Jung et al, 1998; Makowski and Ramsby, 2003; Mannello et al, 2003, 2007; Meisser et al, 2005; Souza-Tarla et al, 2005; Zucker and Cao, 2005). In addition, technical details of serum sampling and handling procedure, for example the time between phlebotomy and centrifugation of blood samples, should be described because they are critical determinants that could affect the concentrations of MMP-9 (Jung et al, 2005; Meisser et al, 2005; Mannello and Tonti, 2007).

To illustrate that problem comparative MMP-9 measurements were performed in serum and plasma samples that were obtained under different collection conditions. For that purpose, blood samples from 10 healthy adults were simultaneously collected in differently prepared plastic tubes (Monovette Systems, Sarstedt AG, Nümbrecht, Germany) and centrifuged at 1600 g at 4°C for 15 min within 30 min after venepuncture. Tubes either with kaolin-coated granulate as clot activator or without additive were used to prepare serum after enhanced coagulation (serum⁽⁺⁾) and native serum (serum⁽⁻⁾), respectively whereas tubes coated with lithium heparin, sodium citrate or dipotassium EDTA were used to collect plasma samples. The supernatants were carefully removed and stored at −80°C until analysis. MMP-9 was measured in duplicates with the Fluorokine MultiAnalyte Profiling assay (R&D Systems, Minneapolis, MN, USA) detecting pro-, mature, and TIMP complexed MMP-9 according to the note of the producer. The coefficient of variation calculated from the duplicate values was 9.7%. The striking effect of sample collection on the MMP-9 concentrations is presented as percentage data related to values obtained in serum⁽⁺⁾ collected with clot activator (Figure 1). Higher MMP-9 concentrations were observed in serum than in plasma. In addition, the kind of serum sampling procedure with or without clot activator was important. The concentrations found in serum⁽⁺⁾ samples collected with clot activator were about 3–4 times higher compared with those in serum⁽⁻⁾ samples collected without clot activator. However, also the MMP-9 concentrations in serum⁽⁻⁾ samples collected without clot activator corresponding to serum collected in plain tubes using the Vacutainer system by Hurst et al (2007) were 3–5 times higher than in heparin or citrate plasma. All these data show that serum MMP-9 concentrations do not correspond to the real circulating concentrations of MMP-9 in blood because the anticoagulants do not affect the recovery of MMP-9 in samples (Jung et al, 1998). Platelets and leukocytes abundantly contain MMP-9 among other proteases (Murphy et al, 1977; Cooper et al, 1985; Opdenakker et al, 2001; Sheu et al, 2004; Santos-Martinez et al, 2008). The increased MMP-9 concentration observed in serum may be due both to release of MMP-9 during coagulation from the blood cells and its secretion that is induced by the clot activator itself (Mannello and Tonti, 2007; Mannello et al, 2007). Thus, it can be assumed that this high unspecific background concentration of MMP-9 in serum probably impairs the potential diagnostic/prognostic performance of this parameter. Recent studies confirmed that plasma MMP-9 had better diagnostic accuracy than serum MMP-9 (Jung et al, 2001; Wu et al, 2007). Also the consistent use of serum collected under identical conditions was obviously not suitable to circumvent that effect.

Figure 1.

Relative concentrations of MMP-9 in serum and plasma samples. Percentage values (medians with interquartile ranges) of MMP-9 concentrations in serum and plasma samples simultaneously collected from 10 healthy adults were related to the values in serum⁽⁺⁾ samples (median: 536 μg l⁻¹; interquartile ranges: 384–692 μg l⁻¹) that were taken as 100%. Further details see text. Serum⁽⁺⁾ and serum⁽⁻⁾: serum prepared with and without clot activator (kaolin-coated granulate).

In conclusion, serum collected either with or without clot activator should not be considered as an appropriate sample for determining circulating MMP-9. Pre-analytical conditions of sample collection and handling have to be clarified before the diagnostic or prognostic performance of a corresponding marker should be explored in clinical trials (Mannello et al, 2005; Lomholt et al, 2007). It may be advisable to re-consider the sample collection procedure for the planned studies. Citrate plasma has been suggested to be the sample of choice for measuring circulating MMP-9 (Makowski and Ramsby, 2003; Mannello et al, 2003; Meisser et al, 2005; Gerlach et al, 2005).