The intrinsic fluorescence spectrum of dilute gastric juice as a novel diagnostic tool for gastric cancer

Authors


Li Ya ZHOU, Department of Gastroenterology, Peking University Third Hospital, Beijing 100191, China. Email: liyazhou@medmail.com.cn

Abstract

OBJECTIVE:  To investigate the intrinsic fluorescence spectrum of gastric juice as a diagnostic method for gastric cancer.

METHODS:  We collected gastric juice by gastroscopy in 1870 patients from May 2001 to March 2006, of whom 202 were involved in a preliminary test, 162 in experimental optimization and 1506 in clinical verification. The best dilution and pH value were chosen in the experimental optimization phase. Clinical verification was based on optimized samples. Intrinsic fluorescence spectra were measured in all samples with a fluorescence spectrophotometer using an excitation wavelength of 288 nm.

RESULTS:  The first peak of fluorescence intensity (P1FI) of the intrinsic fluorescence spectrum was significantly higher in gastric juice from patients with gastric cancer than from those with benign lesions. There was no significant difference in the P1FI differences between patients with benign and malignant lesions with samples diluted by 20-fold to 80-fold and from pH 9 to pH 11. Clinical verification in 1506 patients showed that P1FI ≥ 76.5 was the optimal cut-off on the receiver operating characteristic curve for diagnosing gastric cancers: sensitivity was 83.2%, specificity 80.7% and accuracy 82.0%.

CONCLUSIONS:  P1FI of the intrinsic fluorescence at 288 nm is significantly higher in patients with gastric cancers than in individuals with benign lesions. As a clinical indicator of gastric cancer, its sensitivity, specificity and accuracy were high.

INTRODUCTION

Gastric cancer is one of the most common malignant tumors in humans,1 yet we still lack a simple and reliable laboratory diagnostic method for this disease. Clinical diagnosis currently relies on gastroscopy and histology or double contrast radiography of the upper gastrointestinal tract, both of which require large, specialized pieces of equipment with well-trained operators, making it impracticable for widespread, large-scale population screening.

Over the past 20 years several studies2–8 have demonstrated that endogenous (natural) fluorescence in body fluids can be used to detect and diagnose malignant tumors. Firsova et al.9 found that fluorescence emission spectra in serum from rats with malignant tumors was significantly different from that in normal rats. Grigorovich et al.10 found that the intrinsic serum fluorescence emission spectra of cancer patients (in response to monochromatic ultraviolet (UV) excitation at 313–317 nm) was significantly different from fluorescence spectra from the sera of healthy individuals. For diagnosing malignant tumors, the sensitivity and specificity of this method were greater than when carcinoembryonic antigen (CEA) was used as a biomarker.10 Lin et al.11 measured serum fluorescence spectra in 1191 patients with various gastric diseases, including 93 patients with gastric cancers, compared with 1098 individuals with gastric benign lesions, and showed that its sensitivity and specificity in diagnosing gastric cancer were 88.2% and 72.4%, respectively. Therefore, we hypothesize that changes of fluorescence spectra in components of gastric juice may directly reflect the development of gastric cancer. Accordingly, if the changes in the intrinsic fluorescence spectra of gastric juice can be accurately detected, it could be a valuable screening and diagnostic tool for gastric cancer. Our previous studies with small samples demonstrated that the intrinsic fluorescence spectra of gastric juice from cancer patients were significantly different from that of gastric juice in patients with benign lesions.12–16 In this study, we wanted to certify the utility of intrinsic fluorescent spectrum of gastric juice as a diagnostic method for detecting gastric cancer.

MATERIALS AND METHODS

The study protocol was approved by the Peking University Medical Ethics Committee (IRB00001052-10013) and all participants recruited gave their written formal consent.

Patients

This study was divided into three parts: (i) a preliminary study; (ii) experimental optimization; and (iii) large-scale clinical verification. From May 2001 to March 2006, patients underwent gastroscopy examinations for the clinical screening of gastric cancer in the Department of Gastroenterology of four hospitals in Beijing. The following were the criteria for inclusion: patients with a variety of benign diseases and malignant gastric tumors diagnosed by gastroscopic biopsy and histopathological examination. Patients with organic diseases in other organ systems diagnosed by clinical examinations and with metastatic carcinoma in the stomach were excluded.

Sample collection, preservation and preparation

After a 12-h fast the patients underwent routine gastroscopic examinations. Biopsies were done at the pyloric antrum and lesser curvature and abnormal lesions were sampled for histopathological examination. At the end of the gastroscopy procedure, 1–4 mL of gastric juice was collected and processed as follows.

Preliminary test

Within 4 h after sampling, the gastric juice samples were filtered using microcentrifugal filters with a pore size of 0.45 µm (Tianjin Chromatographic Science and Technology, Tianjin, China) and sedimented in an LG16-W centrifuge (Beijing Medical Centrifuge Manufacture, Beijing, China) at 5311×g for 5 min. The filtered juice samples were stored at −80°C. Within 3 days the samples were thawed, diluted tenfold with distilled water and tested.

Experimental optimization

The procedure for gastric juice sampling and centrifugation were as described in the preliminary test. To evaluate the effect of dilution on fluorescence spectra, the gastric juice samples from 54 patients with malignant lesions and 108 patients with benign lesions were diluted with a pH 7 phosphate buffer at 1:5, 1:10, 1:20, 1:40 and 1:80 ratios and then tested. To evaluate the effect of pH on the fluorescence spectrum, gastric juice samples from 45 patients with malignant lesions and 93 with benign lesions were diluted at 1:20 with ultrapure water and phosphate buffer at pH 3, 5, 7, 8, 9, 10 or 11, and then tested.

Clinical verification

The gastric juice samples from 132 patients with malignant tumors and 1374 patients with benign lesions were collected and centrifuged using the above procedures and then preserved either at −80°C or at −196°C (liquid nitrogen) within 4 h. Fluorescence tests were done within 1 month of collecting the gastric sample. Before the test, the samples were thawed to room temperature, diluted 62.5 times with pH 9 phosphate buffer and maintained at pH 9.

Measurement of intrinsic fluorescence spectra of diluted gastric juice

Because our preliminary study showed that the choice of fluorescence spectrophotometer had a negligible effect on test results (data not shown), both an RF-5000 fluorescence spectrophotometer (Shimadazu, Tokyo, Japan) and an Mp F24 fluorescence spectrophotometer (Hitachi, Tokyo, Japan) were used. Before the measurement was taken, 100 µL of gastric juice was diluted at different strengths. The excitation wavelength was set at 288 nm and the slit width at 3 nm. The emission spectrum was collected from 300 nm to 800 nm.

Statistical analysis

These variables were analyzed with SPSS 11.5 (SPSS Inc., Chicago, IL, USA) and the significance of the difference was determined by one-way analysis of variance (anova). All data were expressed as mean ± standard deviation (SD). CART 2.0 (Salford Systems, San Diego, CA, USA) was used to establish a model of the intrinsic fluorescence spectrum from the gastric juice for the diagnosis of gastric cancer. A receiver operating characteristic (ROC) curve was used to determine the optimal first peak of fluorescence intensity (P1FI), pH and the dilution of gastric juice for distinguishing the cancer from the non-cancer samples. P < 0.05 was considered statistically significant.

RESULTS

The characteristics of the study population are outlined in Table 1.

Table 1.  Patients' characteristics
Experimental groupPatients (n)GenderMean age (years, range)Benign lesions (n)Malignant lesions (n)
Male (n)Female (n)CSGCAGDU and GUATP1-2
  1. ATP1-2, mild and moderate dysplasia; CAG, chronic atrophic gastritis; CSG, chronic superficial gastritis; DU, duodenal ulcer; GU, gastric ulcer.

Systematic verification2029510749.8 (19–80)1034118535
Experimental optimization162907254.3 (15–81)10854
Clinical verification150688662047.4 (15–80)561372243198132
Total1870107179948.0 (15–81)772413261203221

Systematic verification

General characteristics of intrinsic fluorescence spectra of diluted gastric juice

There were three peaks in the fluorescence emission spectrum (300–800 nm): the first (P1F) (320–360 nm), the second (P2F) (576 nm), and the third (P3F) (670–690 nm). For P1F and P3F, the base was wide while the P2F base was high and sharp. The location and the number of fluorescent peaks of gastric juice from patients with gastric cancer and those with benign lesions were basically the same but the P1FI was stronger in patients with gastric cancer (Fig. 1). There was a significant difference in P1FI between malignant and benign cases (P < 0.05, anova), but no difference was found among the subgroups of the benign cases (P > 0.05, anova). The P2FI and the P3FI were not significantly different between patients with malignant tumours and those with benign lesions (P > 0.05, anova, Table 2).

Figure 1.

A typical intrinsic fluorescence spectrum of diluted gastric juice showing (a) benign gastric lesions; (b) gastric cancer. FI, fluorescence intensity; P1F, the first peak of fluorescence; P2F, the second peak of fluorescence; P3F, the third peak of fluorescence.

Table 2.  Comparison of the first (P1FI), the second (P2FI) and the third peak of fluorescence intensity (P3FI) in the intrinsic fluorescence spectrum of diluted gastric juice in patients with various gastric diseases (mean ± standard deviation)
 Patient (n)P1FIP2FIP3FI
  • *

    P < 0.05 compared with the other groups.

  • ATP1-2, mild to moderate dysplasia; CAG, chronic atrophic gastritis; CSG, chronic superficial gastritis; DU, duodenal ulcer; GU, gastric ulcer; SD, standard deviation.

CSG10359.86 ± 29.7416.85 ± 11.155.09 ± 3.03
CAG4161.31 ± 29.8123.71 ± 20.174.94 ± 2.64
DU and GU1853.77 ± 32.3513.21 ± 10.754.50 ± 2.41
ATP1-2583.18 ± 33.2121.24 ± 20.005.13 ± 2.38
Gastric cancer35172.15 ± 47.11*31.46 ± 22.397.38 ± 3.53

The best discriminant model of the intrinsic fluorescence spectrum established by CART V2.0 for the diagnosis of gastric cancer

A total of 202 samples were tested, including 35 gastric cancers and 167 benign diseases (including chronic superficial gastric, chronic atrophic gastritis, duodenal ulcer, gastric ulcer and mild and moderate dysplasia). P1FI ≥ 111.80 was used as the discriminant standard. Of these patients, 60 fulfilled this criterion (i.e., the diagnosis of gastric cancer), including 32 gastric cancer and 28 gastric benign lesions; while 142 patients did not fulfil, including 3 gastric cancers and 139 benign lesions.

The prior probability and a posteriori probability of various gastric diseases determined by this discriminant model

The prior probability determined by the CART model of intrinsic fluorescence spectra of diluted gastric juice included a sensitivity of 91.4%, a specificity of 83.2%, a false positive rate of 16.8%, a positive predictive value of 53.3%, a false negative rate of 8.6%, a negative predictive value of 97.9% and an accuracy of 84.7%. The posteriori probability of gastric cancer included a sensitivity of 85.7%, a specificity of 82.6%, a false positive rate of 17.3%, a positive predictive value of 50.9%, a false negative rate of 14.2%, a negative predictive value of 96.5% and an accuracy of 83.1%.

Experimental optimization

The effects of gastric juice dilution on the intrinsic fluorescence spectrum

Gastric juice dilution significantly affected the P1FI value. After the gastric juice was diluted with pH 7 phosphate buffers the P1FI value in patients with malignant and benign lesions increased and then decreased with increasing dilution, compared with the pre-diluted P1FI value. In patients with gastric cancers the increase of P1FI in diluted gastric juice was greater than that in those with gastric benign lesions.

The P1FI value of undiluted gastric juice was lower in patients with gastric cancers than in those with benign tumors. However, with a fivefold dilution the P1FI value was higher in patients with cancer than in those with benign lesions. The difference became further significant for 1:10 and 1:20 dilutions (Fig. 2a). The area under the ROC curves were the same (0.87) for 1:20, 1:40 and 1:80 dilutions (Fig. 2b).

Figure 2.

The effect of gastric juice dilution on the intrinsic fluorescence spectrum on (inline image) benign lesion (n = 108) and (inline image) malignant tumor (n = 54). (a) The first peak of fluorescence intensity (P1FI) with different dilutions; (b) the area under the receiver operating characteristic (ROC) curve with different dilutions.

The effect of the pH value on the intrinsic fluorescence spectrum of gastric juice

The effects of the pH value of the phosphate solution and the effect of ultrapure water on the P1FI value was negligible in patients with benign lesions, but significant in patients with malignant lesions. When gastric juice was diluted at 1:20, P1FI increased with the increasing pH. At pH 11, the difference (Fig. 3a) in P1FI between the benign and malignant cases and the area under the ROC curve (Fig. 3b) were the largest.

Figure 3.

The impact of the pH value of diluted gastric juice on the intrinsic fluorescence spectrum (a) The first peak of fluorescence intensity (P1FI) at different pH for (inline image) benign lesion (n = 93) and (inline image) malignant tumor (n = 45); (b) the area under the receiver operating characteristic (ROC) curve at different pH.

In summary, experimental optimization showed that using a 20-fold to 80-fold dilution and a pH of 9–11, the differences of intrinsic fluorescence in gastric juice from benign and malignant lesions were the greatest. Based on the ROC curve, P1FI ≥ 76.5 was determined to be the optimal cut-off point, that is, the diagnostic threshold for differentiating between benign and malignant lesions (Fig. 4).

Figure 4.

The receiver operating characteristic (ROC) curve for the diagnosis of gastric cancers by intrinsic fluorescence spectra of diluted gastric juice.

Clinical verification

Based on the above optimization results, we tested the intrinsic fluorescence spectrum in 1506 diluted gastric juice samples to determine if they came from patients with benign or malignant lesions. Our test variables were: phosphate buffer at pH 9, 62.5-fold dilution and a P1FI threshold of 76.5 for distinguishing benign from malignant lesions. The P1FI values for the different diseases are shown in Table 3.

Table 3.  The first peak of fluorescence intensity (P1FI) values of diluted gastric juice from patients with various gastric lesions
Pathological diagnosisPatients (n)P1FI minimumP1FI maximumP1FI mean ± SD
  1. ATP1-2, mild to moderate dysplasia; CAG, chronic atrophic gastritis; CSG, chronic superficial gastritis; DU, duodenal ulcer; GU, gastric ulcer; SD, standard deviation.

CSG561818355.22 ± 26.59
CAG3721015651.14 ± 24.77
GU and DU243719356.27 ± 29.06
ATP1-2198515358.31 ± 25.76
Malignant tumor1325260100.02 ± 45.45

The diagnosis based on pathological findings was used as the gold standard to determine the accuracy of the diagnosis using fluorescence spectra. When we set P1FI ≥ 76.5 as the optimal cut-off point on the ROC curve for diagnosing gastric cancer in clinical screening, the diagnostic sensitivity was 83.2%, the specificity 80.7%, the accuracy 82.0%, the false positive rate 19.3%, and the false negative rate 16.8%.

DISCUSSION

The results of this study revealed that P1FI in the intrinsic fluorescence emission spectrum from diluted gastric juice has a diagnostic value for differentiating benign from malignant lesions in the stomach. Using fluorescence spectroscopy to diagnose gastric cancer was supported by recent studies that showed differences in the endogenous fluorescent spectrum of tumor tissues. Several studies demonstrated that there are fluorescence spectrum differences between tumors and adjacent healthy tissues, between benign and malignant lesions and between cancer patients' serum and serum from healthy individuals, demonstrating its potential for early tumor diagnosis.10,11,17 In our approach, which used theoretical verification, experimental optimization and clinical verification, we validated the feasibility of using the fluorescence spectrum of gastric juice as a tool to diagnose gastric cancer.

Our key finding was that P1FI in the intrinsic fluorescence spectrum of diluted gastric juice was significantly different between patients with benign lesions and those with malignant tumors. The mechanism underlying this phenomenon is under investigation.18 A study using high-performance liquid chromatography showed that the concentration of a specific fluorescent material was significantly greater in the gastric juice from patients with gastric cancers than from those with other benign gastric lesions.14 Further, a mass spectrometry analysis showed that this material could be L-tryptophan or one of its analogues with a molecular weight of 216.18 It was previously demonstrated that the structure and metabolism of tryptophan were abnormal in cancer patients19,20 and this was a common phenomenon in tumors from different species and organs.21 Recently, some investigators have attempted to diagnose tumors using the fluorescence characteristics of L-tryptophan and its metabolites.22,23 The excitation and emission wavelengths of L-tryptophan (278/348 nm for free tryptophan, 278/322–345 nm for combined tryptophan) were similar to the wavelength of peak P1FI in our study. Therefore, we speculated that P1FI may be derived from L-tryptophan or its analogues. A definitive conclusion will require further studies.

After we determined the potential of using the intrinsic fluorescence spectrum of diluted gastric juice for diagnosing gastric cancer, we optimized the experimental procedures and established the best dilution and pH value of gastric juice so that a large-scale application of this method could be routinely used. In our study we collected gastric juice by gastroscopy, which is not an optimal procedure because it entails passing a rigid tube down the patients' esophagus and the anatomical peculiarities of the stomach make the procedure unsafe and unpleasant for the patient. The recent development of a sampling capsule technology has made gastric juice collection straightforward and safe. In this procedure, patients swallow the capsule-like juice collector24 that absorbs the gastric juice. Within minutes, the collector is withdrawn with a thread, which provides the gastric juice sample. This method is less painful, relatively inexpensive, less invasive and more acceptable to the patient, making it a far more acceptable procedure in routine use. The combination of diagnostic intrinsic fluorescence spectrum of diluted gastric juice with this sampling technique would allow the development of a safe and effective laboratory diagnostic tool that could be used for gastric cancer screening and early diagnosis.

In conclusion, compared with existing methods for diagnosing gastric cancers, the intrinsic fluorescence spectra of diluted gastric juice are significantly more sensitive, specific, reliable and accurate. This new method is a safe, simple and cost-effective diagnostic method and can be used widely, we believe, for gastric cancer screening. This method has a nearly 20% false positive rate and a 20% false negative rate, suggesting that it can be used only for initial cancer screening. In clinical practice, it should be combined with other diagnostic methods for the final diagnosis to be confirmed by a traditional pathological examination.

ACKNOWLEDGMENTS

The projects were supported by the National Natural Science Foundation of China (Grant No. 30371603 and 30973415). The authors are grateful to Professor Jia Rou PENG, Dr Hong Shan WEI, Yan Li ZHANG, Xin Lian JIN and Yue Ning ZHANG for their help with the study.

Ancillary