Patterns of Occult Bleeding in Asymptomatic Colorectal Cancer

DAVID A. AHLQUIST, MD, DOUGLAS B. MCGILL, MD, JON L. FLEMING, MD, SAMUEL SCHWARTZ, MD, H. SAMUEL WIEAND, PHD, JOSEPH RUBIN, MD, AND CHARLES G. MOERTEL, MD

The assumption that asymptomatic colorectal cancers bleed provides the rationale for widespread stool screening. The authors studied 12 patients with unoperated colorectal cancer but without colorectal symptoms and six healthy volunteers as laboratory controls. All stools were collected for 2 weeks and analyzed by the HemoQuant and Hemoccult tests. In controls, the mean HemoQuant value was 0.7 mg hemoglobin (Hb)/g stool (range, 0.1-1.8) and all stools were Hemoccult-negative. In cancer patients, the mean HemoQuant was 3.3 mg Hb/g (range, 0.3-13.2); stools were within the normal HemoQuant range (<2 mg Hb/g) in 38%and negative by Hemoccult in 70%. The mean cancer detection rate testing a single stool per patient was 57% for HemoQuant and 25%for Hemoccult (P < 0.001). The detection rate rose testing multiple stools and was maximal with five stools at 83%for HemoQuant compared to 31%for Hemoccult (P < 0.001). The authors conclude that fecal blood levels are commonly normal with asymptomatic colorectal cancer. Although higher with HemoQuant than Hemoccult, cancer detection rates by fecal blood testing appear to be lower than previously reported. Cancer 63:1826-1830, 1989.

screening by fecal o c d t blood testing is widely practiced in industrialized nations. This massive effort is rooted on the assumption that asymptomatic cancers bleed frequently and sufficiently enough to be detected by available fecal blood tests. In fact, little is known about the natural bleeding history of colorectal cancer. Reported fecal blood levels with colorectal neoplasia may not be relevant to the screening setting as they have been observed largely from groups of referred, symptomatic, or recently instrumented patients.'-' There have been no published studies on the longitudinal patterns of occult bleeding from truly asymptomatic cancer patients. Several obstacles complicate the meaningful study of such patients. Identification of subjects without selection bias is difficult in referral centers. The time between diagnosis and definitive treatment is typically very short. Results from stools collected during this interval may be uninterpretable due to artifact from recent biopsies or purgative procedures. Also, there are obvious logistic and
OLORECTAL CANCER From the Departments of Internal Medicine and Medical Oncology, Mayo Clinic, Rochester, Minnesota, and Minneapolis Medical Research Foundation, Minneapolis, Minnesota. Supported by the Mayo Medical Foundation. The authors thank Susan Laging, Mark Ellefson, Dr. Ralph Ellefson, Bernadine Stevens, Mark Larson, and Stephen Cha for invaluable technical assistance. Address for reprints: David A. Ahlquist, MD, Gastroenterology Division, Mayo Clinic, Rochester, MN 55905. Accepted for publication November 14, 1988.

ethical constraints to delaying potentially curative resection for the purpose of a nontherapeutic study. Finally, technical limitations of fecal blood tests have existed. Commonly used guaiac pad tests are qualitative, have variable sensitivity for blood,'-6*s-'0and are affected by many factors which either abolish"-'4 or t ~ i g g e r ' ~the -'~ nonspecific color reaction. Radiochromium-labeled erythrocytetechniques" are quantitative but cumbersome for observations over an extended time period and cannot themselves be used for screening. In the current study, we sought to define fecal occult blood levels from colorectal cancer patients who presented with evidence of metastatic disease but without colorectal symptoms. Stools were serially collected and quantitatively analyzed with the HemoQuant test (Mayo Medical Laboratories, Rochester, MN, and Smith-Nine BioScience Laboratories, Van Nuys, CA) which is not subject to those factors that interfere with guaiac reaction^.^^^^^^'^^^ Hemoccult I1 test (Smith-Nine Diagnostics, Sunnyvale, CA) results were compared with those of HemoQuant on all stools.
Methods

Subjects Colorectal cancer patients: Inclusion criteria are listed in Table 1. No patient experienced bleeding or other colorectal symptoms at the time of study and none had been identified or referred because of an abnormal fecal blood

1826

No. 9

COLORECTAL CANCER BLEEDING *

Ahlquist et al.

1827

test result. Eleven of the 12 patients had distant metastases and, in the absence of colorectal symptoms, no indication for resection of the primary lesion. The remaining patient had a localized sigmoid carcinoma but for personal reasons elected to delay resection for 2 months. Identification of subjects proved difficult as only 12 appropriate patients were accrued from over 1600 colorectal cancer patients seen at the Mayo Clinic (Rochester, MN) between November 1985 and October 1987. The overwhelmingly most common reasons for exclusion were the presence of colorectal symptoms and an inadequate interval for study between cancer diagnosis and resection. Median age was 58 years (range, 42-74) and there were eight men and four women. The major presenting findings were an abnormal chest radiograph or liver chemistries on routine medical examinations in eight patients, ascites in one, palpable hepatomegaly in one, a pathologic rib fracture in one, and mild microcytic anemia in one. Median blood hemoglobin was 11.4 g/dl (range, 10.1-1 5.3). All tumors were adenocarcinomas, Broders Grade 2 in ten patients and Grade 3 in two others. Cancers were located in the cecum in two patients, ascending colon in one, transverse colon in one, splenic flexure in three, and sigmoid in five. The median tumor size was 4 cm in maximum luminal dimension (range, 3-9) based on radiographic or endoscopic assessment. Six patients had received chemotherapy, a median of 5 weeks (range, 2-12) before study. None received cytotoxic drugs during the study. Controls:Because laboratory technicians were aware of the nature of this study, six healthy volunteers were included as controls. Controls were asymptomatic, had no history of gastrointestinal diseases, took no medications chronically, and reported a normal blood hemoglobin value within the past year. Their median age was 52 years (range, 43-62) and there were three men and three women.

TARI F 1.

Inclusion Criteria for Colorectal Cancer Patients

1. Unoperated primary adenocarcinoma of the colorectum 2. No colorectal symptoms 3. No other known hemorrhagic gastrointestinal disease 4. No biopsies or other gastrointestinal procedures 2 2 wk 5 . No chemotherapy 2 2 wk 6. No prior radiotherapy 7. No nonsteroidal antiinflammatory drugs, vitamin C, anticoagulants, or iron

Analytical Techniques
Mailed-in stool samples were received between 1 and 4 days from reported collection times. Samples were analyzed either on the day of receipt or after several days of storage at -20C; such storage does not affect re~u1ts.l~ HemoQuant and Hemoccult tests were performed by separate experienced technicians without knowledge of subject identity or alternate test results. The HemoQuant test was run on a single 16 mg aliquot from each sample, as previously described. Results are reported as mg of hemoglobin per gram of stool. HemoQuant levels 2 2 mg/g were considered ele~ated.~ The Hemoccult I1 test was performed as recommended in kit instructions. Two aliquots from each sample were smeared onto both windows of two separate Hemoccult I1 cards and tested immediately. A test for a given stool was recorded as positive if either window from either card developed a blue color within 30 seconds after addition of the peroxide catalyst. Hydration of the smeared windows before testing was not done, as we have found that such hydration on freshly smeared stools has little effect.I4

Statistical Considerations
Means and confidence intervals were obtained using the (natural) log-HemoQuant values, as these were more normally distributed than the original data. Because the sequential HemoQuant values obtained from the same patient were correlated, an overall weighted mean and corresponding variance were computed using a minimum variance unbiased linear estimate.24Tests for differences in results by subgroup, e.g., male versus female, were based on these weighted log-HemoQuant values. Estimates of the proportion of times that a HemoQuant or Hemoccult was positive were also based on weighted values that incorporated the correlations of serial HemoQuant values. When multiple stools were evaluated, a set was considered positive if at least one stool result within the set was positive. Only full sets were considered for analysis; for example, if multiples of three stools were tested in the first patient (Fig. l), that patient was considered to have two sets and the seventh stool was ignored. For direct comparison of HemoQuant to Hemoccult on each stool, a variable was defined as 1 if only

Stool Collection
Both controls and patients were instructed to omit red meat from their diets and to avoid nonsteroidal antiinflammatory drugs or other nonessential medications for the 5 days before and throughout a 2-week period of stool collection. To prevent artifact caused by sampling stools from toilet water, a disposable collection device that mounts to the toilet seat was supplied.I6 A marble-sized sample was obtained from the midportion of each stool using a small spatula affixed to a plastic tube cap (Sarstedt, Princeton, NJ). The sample was then screw-sealed into the plastic tube and mailed to our laboratory on the same day. Medians of 13 stools per patient (range, 7-20) and 13 stools per control (range, 1 1-20) were returned.

1828
1 4I
Hemoccult 0 Positive

CANCER May 1 1989

1

Vol. 63

Patients
FIG. 1 . Occult blood levels in stools collected for 2 weeks from 12 patients with asymptomatic primary colorectal adenocarcinoma as determined by HemoQuant (HQT) and Hemoccult.The shaded zone below a HemoQuant level of 2 mg Hb/g stool represents the conventional normal range.

HemoQuant was positive, - 1 if only Hemoccult was positive, and 0 otherwise. We again obtained correlations within patients and obtained a weighted mean value (and corresponding variance estimate) for this variable. This P value was based on a two-sided test for the hypothesis that the mean of this variable was 0.
Results

termined by either the HemoQuant or the Hemoccult test. The median HemoQuant level in controls was 0.7 mg hemoglobin/g stools (range, 0.1-1.8). Fecal blood levels from colorectal cancer patients varied considerably over a 2-week period (Fig. 1). HemoQuant levels were elevated in all fecal samples from four patients, fluctuated between normal and elevated levels in six others, and remained within the normal range in two patients. Fecal blood elevations were usually modest; from 155 stools collected by cancer patients, 59 (38%) contained <2 mg hemoglobin/g, 109 (70%) had <4 mg/g, and 144 (93%) had <8 mg/g. The median HemoQuant level for cancer patients was 3.3 mg hemoglobin/g stool (range, 0.3-1 3.2). The Hemoccult test was uniformly positive in two patients and variably positive in two others. Hemoccult remained negative in all stools from eight patients. A negative Hemoccult result was obtained in 108 (70%)of the 155 stools collected by cancer patients. Based on differences in weighted means, fecal blood levels were not influenced by age, gender, or prior chemotherapy in this group of cancer patients (P> 0.05).

HemoQuant Versus Hemoccult in Cancer Detection

The mean colorectal cancer detection rate for a single stool was 57% for HemoQuant and 25% for Hemoccult (P < 0.001). Multiple stool testing improved detection rates (Fig. 2). Based on analysis of consecutive sets of stools, the mean cancer detection rate rose to a maximum of 83% for HemoQuant with sets of five stools compared to a maximum rate of only 31% for Hemoccult (P < 0.00 1). Increasing set size to six or more stools did not further improve the detection rate for either test. The colorectal cancer detection rate observed for HemoQuant on a single stool per patient (57%)was nearly twice that for Hemoccult on six stools per patient (31%). Based on three stools per patient, as is commonly practiced in screening settings, the detection rate was 3 1% for Hemoccult compared to 76% for HemoQuant ( P < 0.001). Fecal blood levels tended to be higher with proximal than with distal colon cancers. The average mean HemoQuant level was 3.7 mg hemoglobin/g stool (range, 0.8-7.5) in the seven patients with cancers located proximal to the descending colon compared to an average mean level of 2.6 mg/g (range, 0.6-5.1) in those five patients with cancers in the sigmoid. The maximal detection rates were 86% for the more proximal cancers and 80% for the sigmoid cancers by the HemoQuant test. Maximal detection rates were 29% and 32%, respectively, by the Hemoccult test. Within either site, HemoQuant had a higher detection rate than did Hemoccult ( P < 0.02). Differences in detection rates were not significant ( P > 0.05)

Bleeding Patterns
None of the 86 stools from the small group of healthy controls contained an elevated level of hemoglobin as de-

Consecutive stools, no.

FIG.2. Colorectal cancer detection rates by HemoQuant and Hemoccult based on the number of stools tested. Sets of consecutive stools were considered for comparison. Tabulations representdata on 155 stools from 12 patients with asymptomatic colorectal cancer.

No. 9

COLORECTAL CANCER BLEEDING

Ahlquist et al.

1829

between sites for either the HemoQuant or Hemoccult test.

TABLE 2. Relationship Between HemoQuant* Levels and Hemoccult? Positivity in Stools From Patients With Colorectal Cancer
HemoQuant Hemoccult positive Range (me Hb/e stool) No. Percent No.

Chemical Comparison of Tests

Hemoccult failed to react with stools containing an elevated HemoQuant value over 50% of the time. Hemoccult was positive in 45 (47%) of the 96 stools with HemoQuant levels 2 2 mg hemoglobin/g. On the other hand, HemoQuant was elevated in nearly all stools that were Hemoccult positive, as HemoQuant was 2 2 mg/g in 45 (96%)of the 47 Hemoccult-positive stools. Hemoccult positivity was influenced by the amount of hemoglobin in stools (Table 2). However, there was no fecal hemoglobin concentration that separated Hemoccult positives from negatives. Most fecal blood levels fell within a minimally elevated range where Hemoccult proved to be insensitive. Hemoccult was positive in 35 (41%) of the 85 stools with HemoQuant levels between 2 and 7.9 mg of hemoglobin/g.

* Mayo Medical Laboratories (Rochester, MN) and Smith-Mine

BioScience Laboratories (Van Nuys, CA). t Smith-Kline Diagnostics (Sunnyvale, CA).

H b hemoglobin.

Discussion
As shown in this study, asymptomatic colorectal cancers are frequently associated with normal fecal blood levels. Fecal blood elevations were typically modest and usually beneath the sensitivity threshold of Hemoccult. Hemoccult remained negative in all stools during a 2-week collection period from fully two thirds of our patient group with colorectal cancer and its maximal detection rate was only 31%. The maximal cancer detection rate by the HemoQuant test was 83%. Asymptomatic colorectal cancers, the intended target of screening, probably bleed less than do symptomatic cancers. Cancer detection rates on a single stool per patient in our asymptomatic group were 57% by HemoQuant and 25% by Hemoccult compared to the appreciably higher detection rates of 97% and 67%, respectively, that we previously observed in a group of 30 symptomatic colorectal cancer patient^.^ Reported cancer detection rates by Hemoccult in groups of largely symptomatic patients have ranged from 40% to 9 1%,1-7,25,26 but all higher than Hemoccult detection rates in this study. Various models constructed to analyze the cost-effectiveness of Hemoccult ~ c r e e n i n g ,and ~~thereby ~~ help direct policy, have usually incorporated estimates of Hemoccult detection rates at about 70%. The disturbingly low Hemoccult sensitivity in the current study may be more representative of asymptomatic colorectal cancer and would have farreaching implications on such modeling. Biochemical limitations of guaiac tests like Hemoccult have become increasingly a ~ p a r e n t . ~ Hemoccult's chemical insensitivity for fecal blood is influenced by reducing substances,' ',I4 fecal storage,'3''4and degradation

of hemoglobin by colonic f l ~ r a . ~Consistent , ' ~ , ~ ~ with intracolonic hemoglobin degradation, we noted that Hemoccult detection rates were comparable for proximal and distal cancers despite higher fecal hemoglobin concentrations with proximal lesions, as has been previously ob~erved.'* Some ~ - ~ additional hemoglobin degradation may have occurred while stools were mailed to our laboratory, but such delays are a usual part of the screening process and were minimized in this study. The HemoQuant test is quantitative and represents a biochemical improvement over guaiac tests as a measure of occult fecal b l ~ o d . ~ * Importantly, '~,~' it measures both degraded and intact hemoglobin and, in contrast to guaiac tests, more accurately reflects the total amount of blood shed into the colonic lumen. It is sensitive enough to detect modest elevations in fecal blood and, thus, has proven to be a valuable tool for studying occult bleeding. Maximal cancer detection rates for both tests were achieved when sets of five stools per patient were evaluated. The increment from one to two stools per patient resulted in the greatest proportional increase in the HemoQuant detection rate going from 57% to 7070, respectively. The detection rate increased by 6% or less with each additional stool increment to a maximum at five stools. Others have documented improvement in detection rates with multiple stool The number of stools resulting in maximal detection may not be optimal for screening. The optimal number for screening depends not only on sensitivity, but on specificity,cost, and patient compliance considerations as well. Based on three stools per patient, as originally recommended by G r e e g ~and r~~ commonly practiced currently, cancer detection rates were 76% by HemoQuant and 3 1% by Hemoccult. The current study was not designed to assess test specificities. The small control group was included primarily to obviate any analytical biases, as assay technicians were aware of the nature of the study but blinded to subject sources of stools. Nevertheless, based on this small diet-

1830

CANCER May 1 1989

Vol. 63

prepared control group, specificity was 100% for both HemoQuant and Hemoccult tests. Reported specificity has been 97% for HemoQuant3 and 95% to 99% for Herno~cult~ based ' - ~ ~ on much larger control groups. Fecal blood is not a perfect marker for colorectal cancer. Cancers do not always bleed and most causes of occult bleeding are benign. Detection of fecal blood with insensitive guaiac tests further confounds the use of this marker in colorectal cancer screening. Based on observations of cancer patients without colorectal symptoms in this study, we conclude the following: (1) fecal blood levels commonly fall into the normal range and are usually beneath the detection threshold of Hemoccult; (2) HemoQuant is more sensitivethan Hemoccult; however, cancer detection rates with either test are lower than those previously reported; and (3) Multiple stool testing improves the detection rate.
REFERENCES
1. Macrae FA, St. John DJB. Relationship between patterns ofbleeding and Hemoccult sensitivity in patients with colorectal cancers or adenomas. Gastroenterology 1982; 82:891-898. 2. Dybdahl JH, Daae LNW, Larsen S et al. Occult fecal blood loss determined by a "Cr method and chemical tests in patients referred for colonoscopy. Scand J Gastroenterol 1984; 19:245-254. 3. Ahlquist DA, McGill DB, Schwartz S et al. Fecal blood levels in health and disease: A study using HemoQuant. N Engl J Med 1985; 3 12:1422-1428. 4. Herzog P, Holtermuller KH, Preiss J et al. Fecal blood loss in patients with colonic polyps: A comparison of measurements with 5'chromium-labeled erythrocytes and with the Haemoccult test. Gastroenterology 1982; 29:206 1-2067. 5. Doran J, Hardcastle JC. Bleeding patterns in colorectal cancer: The effect of aspirin and implications for fecal occult blood testing. Br JSurg 1982; 69:711-713. 6 . Farrands PA, Hardcastle JD. Accuracy of occult blood tests over a 6-day period. Clin Oncol 1983; 9:217-225. 7. Crowley ML, Freeman LD, Mottet MD et al. Sensitivity of guaiacimpregnated cards for the detection of colorectal neoplasia. J Clin Gastroenterol 1983; 5:127-130. 8. Ostrow JD, Mulvaney CA, Hansel1 JR et al. Sensitivity and reproducibility of chemical tests for fecal occult blood with an emphasis on false-positive reactions. Am J Dig Dis 1973; 18:930-940. 9. Stroehlein JR, Fairbanks VF, Go VL et al. Hemoccult detection of fecal occult blood quantitated by radioassay. Am J Dig Dis 1976; 2 1: 84 1-844. 10. Bassett ML, Goulston KJ. False positive and negative Hemoccult reactions on a normal diet and effect of diet restriction. Aust N Z J Med 1980; 10~1-4. I I . Jaffe RM, Kasten B, Young DS et al. False-negative stool occult blood tests caused by ingestion of ascorbic acid (vitamin C). Ann Intern Med 1975; 83:824-826. 12. Layne EA, Mellow MH, Lipman TO. Insensitivity ofguaiac slide tests for detection of blood in gastric guice. Ann Intern Med 1981; 94: 774-776. 13. Strohlein JR, Fairbanks VF, Go VL et al. Hemoccult stool tests:

False-negative results due to storage of specimens. Mayo Clinic Proc 1976; 51~548-552. 14. Ahlquist DA, McGill DB, Schwartz S et al. HemoQuant, a new quantitative assay for fecal hemoglobin: Comparison with hemoccult. Ann Intern Med 1984; 101:297-302. 15. Lifton LJ, Kreiser J. False-positive stool occult blood tests caused by iron preparations: A controlled study and review of literature. Gastroenterology 1982; 83960-863. 16. Ahlquist DA, Schwartz S , Isaacson J et al. A stool collection device: The first step in occult blood testing. Ann Intern Med 1988; 108:609612. 17. Illingworth DG. Influence ofdiet on occult blood tests. Gut 1965; 6:595-598. 18. Schentag JJ. False positive "Hemoccult" reaction with cimetidine. (Letter to the Editor) N Engl J Med 1980; 303:1 10. 19. Ahlquist DA, Schwartz S. Use of leuco-dyes in the quantitative colorimetric microdetermination of hemoglobin and other heme compounds. Clin Chem 1975; 21:362-369. 20. Owen CA, Bollman JF, Grindlay JH. Radiochromium-labeled erythrocytes for the detection of gastrointestinal hemorrhage. JLab Clin Med 1954; 44:238-245. 21. Schwartz S, Dahl J, Ellefson M et al. The HemoQuant test: A specific and quantitative assay of heme (hemoglobin) in feces and other materials. Clin Chem 1983; 29:206 1-2067. 22. Schwartz S, Ellefson M. Quantitative fecal recovery of ingested hemoglobin-heme in blood: Comparison by HemoQuant assay with ingested meat and fish. Gastroenterology 1985; 89:19-26. 23. Ellefson RD, Larson MS, Schwartz S, Ahlquist DA. The HemoQuant test remodeled for the busy clinical laboratory. Clin Chem 1986; 32:1142-1143. 24. Brownlee KA. Statistical Theory and Methodology in Science and Engineering. New York: John Wiley and Sons, Inc., 1967; 95-97. 25. Griffith CDM, Turner DJ, Saunders JH. False-negative results of Hemoccult test in colorectal cancer. Br Med J 1981; 283:472. 26. Songster CL, Barrows GH, Jarrett DD. Immunochemical detection of fecal occult blood: The fecal smear punch-disc tests. A new noninvasive screening test for colorectal cancer. Cancer 1980; 45:1099-1 102. 27. Eddy DM. Computer models and the evaluation of colon cancer screening programs. In: Colorectal Cancer: Prevention, Epidemiology, and Screening. Winawer SJ, Schottenfeld D, Sherlock P, eds. New York: Raven Press,-1980; 285-298. 28. Barn MJ. Mullev AG. Richter JM. Effect of work-up strategy on the cost-effectivenessof iecal occult blood screening for colo&tal cancer. Gastroenterology 1987; 93:301-3 10. 29. Kristein MM. The economics of screening for colorectal cancer. SOCSci Med 1980; 14:275-284. 30. Neuhauser D, Lewicki AM. What do we gain from the sixth stool guaiac? N Engl JMed 1975; 293:226-228. 3 1 . Simon JB. Occult blood screening for colorectal carcinoma: A critical review. Gastroenterology 1985; 88:820-837. 32. Frank JW. Occult blood screening for colorectal cancer: The benefits. Am J Prev Med 1985; 1:3-9. 33. Frank JW. Occult blood screening for colorectal cancer: The yield and costs. Am JPrevMed 1985; 1:18-24. 34. Frank JW. Occult blood screening for colorectal cancer: The risks. Am J Prev Med 1985; 1:25-32. 35. Ahlquist DA, Beart RS. Use of fecal occult tests in the detection of colorectal neoplasia. Probl Gen Surg 1985; 2:200-210. 36. Goldschmeidt M, Ahlquist DA, Wieand HS, McGill BG, Taylor WF, Ellefson RD. Measurement of degraded fecal hemoglobin-heme to estimate gastrointestinal site of occult bleeding: Appraisal of its clinical utility. Dig Dis Sci 1988; 33:5:605-608. 37. Greegor DH. Occult blood testing for detection of asymptomatic colon cancer. Cancer 1971; 28:131-134.