Death due to colorectal cancer (CRC) is the second most common cancer-related death in men and women in Europe, after lung cancer (Altobelli et al., 2014). Early detection is critical for survival; the 5-year survival rate of CRC patients diagnosed with localized disease is 90–95% following surgery, but only 5–10% if diagnosed with distant disease (Etzioni et al., 2003). Screening programs have been established to increase early detection (Altobelli et al., 2014). These typically involve fecal occult blood test or specific fecal immune test for hemoglobin (Hb) as initial step, followed by colonoscopy, sigmoidoscopy, or computed tomography colonography (Altobelli et al., 2014 ; Zavoral et al., 2014).
Fecal occult blood tests or fecal immune tests can detect heme or blood in stool. Blood loss from colorectal tumors may lead to iron deficiency and, in a later stage, iron deficiency anemia, depending on the size and location of the tumor (Li et al., 1999). The prevalence estimates of CRC in patients with iron deficiency anemia range between 5 and 10% in the UK; unexplained anemia is considered a warning signal warranting specialist referral to exclude gastrointestinal malignancies (Raje et al., 2007 ; Goddard et al., 2011 ; Pengelly et al., 2013). Important blood parameters in diagnosing anemia are Hb concentration, mean corpuscular volume (MCV), and serum ferritin (SF) (Moreno Chulilla et al., 2009). Iron deficiency likely precedes iron-deficiency anemia and its symptoms by several months. Low SF levels are a marker of iron deficiency and have been associated with CRC (Sawhney et al., 2007 ; Hamilton et al., 2008 ; Baicus et al., 2012). Most associations between laboratory parameters and CRC have been observed in secondary care and have been based on cross-sectional study designs. Blood parameters are, however, also routinely tested in primary care, and the availability of the results in large primary care databases provides an opportunity to investigate the timing of laboratory value changes and their association with CRC. A study using primary care data from the THIN database has already demonstrated that anemia markers recorded in primary care databases are associated with CRC (Hamilton et al., 2008).
The first objective of this study was to assess whether the known association between anemia and CRC can be replicated focusing on diagnostic markers of anemia, such as low Hb and SF, recorded in the Clinical Practice Research Datalink (CPRD) primary care database. Our second objective was to evaluate the timing of these associations. Finally, we assessed whether differences in the level of these markers predict survival in CRC patients.
We conducted a case–control study and a cohort analysis using data from the CPRD. The CPRD is a large UK-based database that encompasses some eight million people who are enrolled with selected general practitioners (GPs), as described in detail elsewhere (Jick, 1997 ; Lawson et al., 1998 ; Wood and Martinez, 2004). GPs have been trained to record medical information in a standard manner and to supply it anonymously (patients are identifiable only by means of a unique number, which is not available to researchers). The information recorded includes demographics such as sex and year of birth, medical diagnoses (based on ‘Read’ codes), drug prescriptions, and a broad range of routine laboratory parameters. The recorded information has been validated and proven to be of high quality (Jick et al., 1991 ; Herrett et al., 2010). The CPRD has already been used for several studies on CRC (Meier et al., 2002 ; van Staa et al., 2005 ; Yang et al., 2005 ; Srinivasan et al., 2007 ; Yang et al., 2008 ; Bodmer et al., 2012). This study has been approved by the Independent Scientific Advisory Committee (MHRA research, protocol number 14_153).
The study population for the case–control analysis comprised all patients in the CPRD aged 18–89 years with a Read-coded CRC diagnosis between 1 January 2008 and 31 December 2012. We excluded all patients with a history of any cancer except for nonmelanoma skin cancer before the index date. We matched one control patient without cancer to each case on age at the index date, sex, index date, GP, and years of history in the database before the index date.
The cohort analysis included CRC cases only, who survived for at least 1 day after the diagnosis, and was conducted to compare survival in CRC cases across the various levels of Hb, SF, and the other red blood cell (RBC) parameters. For sex-stratified and survival analyses we restricted the study population to patients aged 60 or older.
We conducted a case–control analysis using multivariate conditional logistic regression to assess the association between Hb, SF, MCV, and RBC count and risk for CRC, adjusted for BMI (<18.5, 18.5–24.9, 25.0–29.9, ≥30 kg/m2, unknown), smoking status (nonsmoker, current smoker, ex-smoker, unknown), and history of hypertension, diabetes, aspirin use or use of other NSAIDS, vitamin K antagonists, and platelet inhibitors. Medication use was grouped by number of prescriptions into never use, 1–5, 6–10, 11–30, 31–50, or more than 50 prescriptions. In women we additionally adjusted the analyses for hormone replacement therapy exposure. We converted all Hb measurements to the unit g/dl and classified them into the categories less than 9.0, 9.0–9.9, 10.0–10.9, 11.0–11.9, 12.0–12.9, 13.0–15.9, or of at least 16.0 g/dl. MCV values (in fl) were grouped into less than 80, 81–85, 86–95, 96–100, or more than 100. SF values (in ng/ml) were grouped into the following categories: less than 20, 21–40, 41–80, 81–100, 101–300, or more than 300. RBC counts (in 10×12/l) were grouped into the following categories: less than 3.5, 3.5–4.2, 4.3–4.9, 5.0–5.8, or at least 5.9. Patients with missing values were included in the analyses as separate indicator groups. We stratified the analyses by the time of the measurements before the index date (<180 days, >1 year, or >3 years), and we ran a sex-stratified sensitivity analysis in patients 60 years or older to exclude bias from anemia associated with menstruation in premenopausal women.
We used Cox-proportional hazard analyses to compare survival in CRC cases in the cohort study across the various levels of Hb, SF, and the other RBC parameters. We restricted the survival analyses to patients 60 years or older as the majority of cases were 60 years or older, and age is strongly associated with survival. We adjusted the survival analysis for age, sex, BMI, and smoking status. We performed two sensitivity analyses as a proxy for staging of CRC: one in patients who underwent surgery within 180 days after the diagnosis, and one in patients without surgery, assuming that the ones without surgery were more advanced.
We used the software program SAS, version 9.4 (SAS Institute Inc., Cary, North Carolina, USA) to conduct the statistical analyses.
We identified 9238 patients with colon cancer, of whom 4111 (44.5%) were female and 5127 (55.5%) male. The mean age at diagnosis was 69 (±12) years, with a majority of patients being 60 years or older at the index date. A history of iron-deficiency anemia at the index date was substantially more prevalent in cases (21.7%) than in controls (8.2%). The characteristics of the study population are displayed in detail in Table 1. Information on blood tests was available for 3–70% of patients, depending on the test parameter and the timing of the measurement.
The presence of low levels compared with normal levels of all assessed markers shortly (<180 days) before the CRC diagnosis was predictive of a CRC diagnosis. The odds ratio (OR) for Hb values less than 9 g/dl was 74.25 [95% confidence interval (CI): 34.69–158.91], the OR for MCV values up to 80 fl was 13.94 (95% CI: 10.31–18.85), the OR for SF values less than 20 ng/ml was 10.66 (95% CI: 6.88–16.51), and the OR for RBC counts less than 3.5×1012/l was 4.33 (95% CI: 3.15–5.96), as compared with normal levels of the respective parameters. Although the association was strongest for the most extreme levels, less extreme levels such as Hb levels (10.0–10.9 or 11.0–11.9 g/dl) were also significantly associated with a CRC diagnosis. When we evaluated measurements taken at least 1 year before the index date, we observed only 2–3-fold increased ORs for low levels of Hb, SF, and MCV, and no association for low RBC counts. The majority of CRC cases still had values in the normal range at this point in time. For measurements at 3 years or earlier before the index date we did not observe any substantial association between abnormal laboratory values and CRC risk. The results for the different levels and markers are displayed in detail in Table 2. The observed pattern for low levels was the same when we restricted the analysis to patients 60 years or older overall (data not shown) and in sex-stratified analyses (Table 2), but the strength of the association was slightly different for the different markers. A difference between men and women who were 60 years or older was observed for SF levels, 21–40 ng/ml measured at 3 years or earlier before the index; although women had an increased risk for a CRC diagnosis (OR: 2.49; 95% CI: 1.52–4.08, respectively), there was no statistically significant association in men (OR: 0.89; 95% CI: 0.52–1.50, respectively).
Low levels of Hb, SF, MCV, and low RBC counts were only weakly associated with survival in the cohort analysis; associations were stronger in patients not undergoing surgery after the diagnosis than in patients undergoing surgery. Low levels of SF were associated with a better survival in patients not undergoing surgery (SF<20 ng/ml, hazard ratio: 0.54; 95% CI: 0.30–0.99). The mean follow-up time after the diagnosis was 716 days. Table 3 provides detailed information on the association between Hb, SF, MCV levels, and RBC counts and survival.
Although there are numerous studies on disease epidemiology and on pharmacoepidemiology using data from the CPRD, only few studies are available in which laboratory data have been used; even less is known about the proportion of patients with available recordings and about the distribution of laboratory values. This is one reason why we used the CPRD to assess the previously known association between anemia and CRC (Twine and Morris, 1986 ; Raje et al., 2007 ; Hamilton et al., 2008 ; Pengelly et al., 2013). Our study was able to replicate this association – that is, low levels of Hb, SF, and MCV were associated with an increased risk for a CRC diagnosis. The recording of Hb and other laboratory measurements has increased since a study using data from the UK THIN database was conducted using data from the years 2000–2006. In the THIN study, 23% of the controls and 49% of cases had Hb measurements in the year before the index date, whereas MCV values were only available for 8% of the study population, and 11% of those with Hb values also had SF values (Hamilton et al., 2008). In our study, some 70% of cases and controls had a Hb measurement more than a year before the index date, and 60% of cases and 25% of controls had a Hb measurement less than 180 days before the index date.
Focusing on data from routine diagnostics in primary care also allowed us to assess the timing of anemia symptoms. We found that the association between low levels of Hb, MCV, and SF manifested only shortly before the CRC diagnosis. Hb levels measured at least 1 year before the CRC diagnosis were still normal for the majority of cases, with 88% of all cases with recorded measurements of Hb levels of at least 12.0 g/dl, and with ORs comparing low Hb levels with normal Hb ranging between 1.84 and 2.92. Thus, in some patients, but by far not in all, anemia markers pointed toward an increased CRC risk already more than a year before the diagnosis. It has to be emphasized that even for measurements taken less than 180 days before the diagnosis still more than half (55%) of the cases with recorded measurements had an Hb more than or equal to 12.0 g/dl. A similar pattern was observed for MCV and RBC values, in which the ORs for low MCV or RBC levels measured less than 180 days before the index date were also not as strongly elevated as those for Hb levels. Aside from timing, the magnitude of the association is likely dependent on tumor size (Li et al., 1999), site, and disease stage of CRC, factors we could not control in this analysis, as this information was not available for the majority of patients.
Although only a small proportion of patients had recorded SF measurements, we found a clear association between low levels of SF and CRC; the lower the SF level the stronger the association. Our study is thus consistent with previous observations between low SF levels and CRC (Sawhney et al., 2007 ; Hamilton et al., 2008 ; Baicus et al., 2012). Although only 18% of our cases had a recorded SF measurement less than 180 days before the diagnosis, this number comprised 1701 patients, which is a large sample size compared with many previously published studies of this association (Zavagno et al., 1987 ; Kato et al., 1999 ; Li et al., 1999 ; Sawhney et al., 2007 ; Hamilton et al., 2008 ; Baicus et al., 2012).
We were also interested in the prognostic value of anemia markers measured shortly before the diagnosis in relation to survival as tumor size had previously been associated with Hb, iron, and ferritin concentrations (Li et al., 1999). To account for disease severity, we performed two sensitivity analyses, one in patients who underwent surgery and one in patients with no recorded surgery within 180 days after the diagnosis; we made the assumption that patients with no surgery had more advanced cancer, or for another reason, had a lower life expectancy. Although there was a strong association between low levels of all four markers and the risk of having a diagnosis of CRC, we did not observe such strong associations with survival. Two previous studies comparing less than median with more than median ferritin levels in patients with colon and rectal cancer did not find any statistically significant association with survival (Giessen et al., 2014 ; Giessen-Jung et al., 2015). The role of ferritin and MCV might be more complex: increased levels of ferritin have also been associated with survival in relapsed or refractory metastatic CRC and increased MCV levels have been associated with a higher all-cause mortality in nonanemic patients (Lee et al., 2016 ; Yoon et al., 2016). It is also possible that the lack of association with survival in our study was due to short follow-up (the earliest cases occurred in 2008 in our study), which may have been insufficient to detect meaningful survival differences.
As previously described in the literature (Ballinger and Anggiansah, 2007 ; Charlton et al., 2012), older age was a strong predictor of developing CRC in our study; more than 80% of all diagnoses were made in people aged 60 years or older. We therefore matched our study population on age and restricted some analyses to patients over 60 years because we did not want to assess changes in blood values due to age. Although we were interested in predictive factors and not causal associations, we adjusted for some known risk factors because we were not interested in differences, which can be readily explained by known clinical factors, but were interested in additional information that may arise from blood test results. It is known that environmental or lifestyle factors – such as sedentary lifestyle, regular consumption of high quantities of alcohol, regular consumption of meat products, and a relative lack of consumption of fruit, vegetables, and fibers – are associated with an increased risk for CRC (Weitz et al., 2005). We were not able to adjust for these parameters in the analysis, as they are not recorded in the database. In addition, socioeconomic status and race/ethnicity may also be risk factors for CRC, two parameters which are not routinely recorded. However, we matched colon cancer patients to a comparison group of patients without colon cancer on the same general practice; thus, they were likely from the same neighborhood, which may somewhat control for potential confounding by socioeconomic status and we could control for diagnostic preferences of GPs. Although these limitations exist, it has to be noted that the study has important strengths, such as the large size and the CPRDs high quality for observational research, which has been demonstrated in numerous previous studies.
In summary, we explored the association between markers of anemia and the risk of developing a first-time diagnosis of CRC using the CPRD. The study provides evidence that the CPRD is well-suited to detect low levels of Hb, MCV, SF, and RBC in association with CRC, and that in most patients these changes only occur shortly before a diagnosis. These findings need to be interpreted in the context of the above-mentioned limitations of this observational study.
The authors thank P. Egger (Basel Pharmacoepidemiology Unit, Division of Clinical Pharmacy and Epidemiology, Department of Pharmaceutical Sciences, University Basel, Basel, Switzerland) for technical assistance.
Conflicts of interest
There are no conflicts of interest.
Altobelli E, Lattanzi A, Paduano R, Varassi G, di Orio F (2014). Colorectal cancer
prevention in Europe: Burden of disease and status of screening programs. Prev Med 62C:132–141.
Baicus C, Caraiola S, Rimbas M, Patrascu R, Baicus A, Gsspi (2012). Ferritin above 100 mcg/l could rule out colon cancer, but not gastric or rectal cancer in patients with involuntary weight loss. BMC Gastroenterol 12:86.
Ballinger AB, Anggiansah C (2007). Colorectal cancer
. BMJ 335:715–718.
Bodmer M, Becker C, Meier C, Jick SS, Meier CR (2012). Use of metformin is not associated with a decreased risk of colorectal cancer
: a case–control analysis. Cancer Epidemiol Biomarkers Prev 21:280–286.
Charlton R, Snowball J, Bloomfield K, de Vries C (2012). Colorectal cancer
incidence on the general practice research database. Pharmacoepidemiol Drug Saf.
Etzioni R, Urban N, Ramsey S, McIntosh M, Schwartz S, Reid B, et al (2003). The case for early detection. Nat Rev Cancer 3:243–252.
Giessen C, Nagel D, Glas M, Spelsberg F, Lau-Werner U, Modest DP, et al (2014). Evaluation of preoperative serum markers for individual patient prognosis in stage I-III rectal cancer. Tumour Biol 35:10237–10248.
Giessen-Jung C, Nagel D, Glas M, Spelsberg F, Lau-Werner U, Modest DP, et al (2015). Preoperative serum markers for individual patient prognosis in stage I–III colon cancer. Tumour Biol 36:7897–7906.
Goddard AF, James MW, McIntyre AS, Scott BB (2011). Guidelines for the management of iron deficiency
anaemia. Gut 60:1309–1316.
Hamilton W, Lancashire R, Sharp D, Peters TJ, Cheng KK, Marshall T (2008). The importance of anaemia in diagnosing colorectal cancer
: a case–control study using electronic primary care records. Br J Cancer 98:323–327.
Herrett E, Thomas SL, Schoonen WM, Smeeth L, Hall AJ (2010). Validation and validity of diagnoses in the General Practice Research Database: a systematic review. Br J Clin Pharmacol 69:4–14.
Jick H (1997). A database worth saving. Lancet 350:1045–1046.
Jick H, Jick SS, Derby LE (1991). Validation of information recorded on general practitioner based computerised data resource in the United Kingdom. BMJ 302:766–768.
Kato I, Dnistrian AM, Schwartz M, Toniolo P, Koenig K, Shore RE, et al (1999). Iron intake, body iron stores and colorectal cancer
risk in women: a nested case–control study. Int J Cancer 80:693–698.
Lawson DH, Sherman V, Hollowell J (1998). The General Practice Research Database. Scientific and Ethical Advisory Group. QJM 91:445–452.
Lee S, Song A, Eo W (2016). Serum ferritin
as a prognostic biomarker for survival in relapsed or refractory metastatic colorectal cancer
. J Cancer 7:957–964.
Li F, Kishida T, Kobayashi M (1999). Serum iron and ferritin levels in patients with colorectal cancer
in relation to the size, site, and disease stage of cancer. J Gastroenterol 34:195–199.
Meier CR, Schmitz S, Jick H (2002). Association between acetaminophen or nonsteroidal antiinflammatory drugs and risk of developing ovarian, breast, or colon cancer. Pharmacotherapy 22:303–309.
Moreno Chulilla JA, Romero Colas MS, Gutierrez Martin M (2009). Classification of anemia
for gastroenterologists. World J Gastroenterol 15:4627–4637.
Pengelly S, Fabricius M, McMenamin D, Wu E, Metzner M, Lewis SJ, et al (2013). Attendance at iron deficiency
anaemia clinic: audit of outcomes 5 years on. Colorectal Dis 15:423–427.
Raje D, Mukhtar H, Oshowo A, Ingham Clark C (2007). What proportion of patients referred to secondary care with iron deficiency anemia
have colon cancer? Dis Colon Rectum 50:1211–1214.
Sawhney MS, Lipato T, Nelson DB, Lederle FA, Rector TS, Bond JH (2007). Should patients with anemia
and low normal or normal serum ferritin
undergo colonoscopy? Am J Gastroenterol 102:82–88.
Srinivasan R, Yang YX, Rubin SC, Morgan MA, Lewis JD (2007). Risk of colorectal cancer
in women with a prior diagnosis of gynecologic malignancy. J Clin Gastroenterol 41:291–296.
Twine RW, Morris C (1986). Early detection of carcinomas of the right colon: role of hemoglobin
and mean corpuscular volume. J Natl Med Assoc 78:187–192.
Van Staa TP, Card T, Logan RF, Leufkens HG (2005). 5-Aminosalicylate use and colorectal cancer
risk in inflammatory bowel disease: a large epidemiological study. Gut 54:1573–1578.
Weitz J, Koch M, Debus J, Hohler T, Galle PR, Buchler MW (2005). Colorectal cancer
. Lancet 365:153–165.
Wood L, Martinez C (2004). The general practice research database: role in pharmacovigilance. Drug Saf 27:871–881.
Yang YX, Hennessy S, Lewis JD (2005). Type 2 diabetes mellitus and the risk of colorectal cancer
. Clin Gastroenterol Hepatol 3:587–594.
Yang YX, Hennessy S, Propert K, Hwang WT, Sarkar M, Lewis JD (2008). Chronic statin therapy and the risk of colorectal cancer
. Pharmacoepidemiol Drug Saf 17:869–876.
Yoon HJ, Kim K, Nam YS, Yun JM, Park M (2016). Mean corpuscular volume levels and all-cause and liver cancer mortality. Clin Chem Lab Med 54:1247–1257.
Zavagno G, Nitti D, Marchet A, Moschin N, Belluco C, de Marchi F, et al (1987). Significance of serum ferritin
in patients with gastric and colorectal cancer
. Eur J Cancer Clin Oncol 23:1077–1078.
Zavoral M, Suchanek S, Majek O, Fric P, Minarikova P, Minarik M, et al (2014). Colorectal cancer
screening: 20 years of development and recent progress. World J Gastroenterol 20:3825–3834.
Keywords:Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.
anemia; colorectal cancer; Clinical Practice Research Datalink; hemoglobin; iron deficiency; serum ferritin