Faecal haemoglobin concentrations vary with sex and age, but data are not transferable across geography for colorectal cancer screening

Introduction

There are a number of options available for screening asymptomatic populations for colorectal cancer (CRC). It is considered that the best non-invasive investigation is currently provided by faecal immunochemical tests for haemoglobin (FIT) and these are now widely recommended in guidelines [1, 2]. FIT are available in two formats. Dichotomous qualitative FIT are generally used in point-of-care settings, but have a number of disadvantages including that the cut-off faecal haemoglobin concentrations (f-Hb) for positivity vary among FIT leading to very different clinical outcomes being achieved with different FIT [3], the cut-off f-Hb are set by the manufacturers and the visual interpretation required for reading the results is far from easy [4]. In consequence, automated quantitative FIT analytical systems based on immunoturbidimetry, which can measure the f-Hb, are much favoured [5].

Many have suggested that one of the major advantages of quantitative FIT is that the cut-off f-Hb used in any CRC screening programme can be set by the organisers so as to give the performance characteristics required, such as positivity rate appropriate to the available colonoscopy resource, sensitivity/specificity ratio, or positive predictive value. However, almost ubiquitously, a single cut-off f-Hb is applied to divide the screened population into two groups, those who require further investigation usually by colonoscopy and those who do not but are re-invited for repeat FIT screening annually or biennially. This single cut-off f-Hb used is often that recommended by the manufacturer. Thus, quantitative FIT which have excellent analytical and clinical performance characteristics are generally used in practice as simple qualitative investigations.

Although there are many publications on the clinical outcomes of CRC screening pilots and programmes using quantitative FIT with a single cut-off concentration, important data exist on the dependence of clinical outcomes on different cut-off f-Hb concentrations and the number of faecal samples analysed: positivity rate and sensitivity increasing as the f-Hb cut-off is decreased, at the expense of specificity and positive predictive value [6]. Moreover, Ciatto et al. [7] have reported that f-Hb is significantly related to the presence of the very lesions, cancer and advanced adenoma, that screening is aimed at detecting. There is also considerable evidence that f-Hb increases as disease becomes more serious, from the normal to low-risk adenomatous polyps, then to higher-risk polyps and then to cancer, although there is much overlap between these groups [8, 9]. Further, there is growing realisation that f-Hb alone, or in combination with other important factors [10, 11], can be used to predict future risk: Chen et al. investigated f-Hb as a predictor of incident colorectal neoplasia [12] and found that, in participants with f-Hb below the usual f-Hb cut-off used for referral for colonoscopy, the subsequent risk of incident colorectal neoplasia was predicted. Recently, this work has been extended and it has been shown that f-Hb is related to CRC mortality and even to all-cause death [13].

It is known that, for many aspect of CRC screening, including incidence and uptake, positivity rates and interval cancer rates in guaiac faecal occult blood test (gFOBT) and FIT-based programmes, men are different from women and these outcomes also change with increasing age. We have previously investigated a large cohort of apparently asymptomatic individuals and shown that f-Hb are higher in men than women and increase with age [14]. At any particular f-Hb cut-off, more men were declared positive than women and more older people were declared positive than younger individuals. Future risk of neoplasia was higher in men than in women and in older people. We concluded that these data should be of assistance in screening programme design. However, a germane question that arises is whether these published data on f-Hb are transferrable across geography to different countries. In consequence, we investigated the distributions of f-Hb in three countries that used identical FIT methodology from a single manufacturer.

Materials and methods

Results

The populations in Scotland, Taiwan and Italy studied comprised 38,680 (46.7% men), 113,366 (41.6% men) and 9167 (46.4% men), respectively: For men and women overall and for all four age groups studied, none of the distributions of f-Hb were Gaussian (D’Agostino-Pearson test, p<0.0001) and the coefficients of skewness and kurtosis were significantly >1 (p<0.0001). Since approximately 50% of the populations had undetectable f-Hb and the 75.0th f-Hb percentiles were below the limits of quantitation quoted for the analytical system (10 μg Hb/g faeces), these data were examined but are not tabulated here: the 90.0th, 95.0th and 97.5th percentiles (the last being potential upper reference limits) of f-Hb for men and women are shown in Table 1 for all data available for those aged 50–74 years.

The 90.0th, 95.0th and 97.5th percentiles of f-Hb for men and women for all data available for four 5-year age groups, namely 50–54, 55–59, 60–64 and 65–69 years, are shown in Tables 2 and 3. The 95.0th percentile data for the four 5-year age groups are shown in Figure 1A–C for Scotland, Taiwan and Italy.

Figure 1

Faecal haemoglobin (μg Hb/g faeces) for men and women in four age groups in Scotland, Taiwan and Italy.

(A) Faecal haemoglobin 95.0th percentile by sex and age group in Scotland; (B) Faecal haemoglobin 95.0th percentile by sex and age group in Taiwan; (C) Faecal haemoglobin 95.0th percentile by sex and age group in Italy.

Discussion

We have examined f-Hb in large groups of ostensibly asymptomatic people aged between 50 and 74 years in three countries. Table 1 shows that the quantity of data available on f-Hb is lower in men than in women in all three countries. Overall, men have higher f-Hb than women in all three countries. Many differences between men and women are already well documented with respect to CRC [20], but the reason for this sex difference in f-Hb is not known exactly, although there are some possibilities [14]. Men have higher blood haemoglobin concentrations than women, but the population of women here is likely to be mainly post-menopausal when these sex differences are not apparent: in addition, use of aspirin and non-steroidal anti-inflammatory drugs might be more common in men than women. We agree that a very plausible explanation is that colonic transit time is faster in men than in women: in consequence, since faecal haemoglobin is very unstable, more degradation of any haemoglobin released into the gut before the excretion of faeces could occur in women. In addition, Tables 2 and 3 show that for both men and women f-Hb rises with age in all three countries, although in Italy, the f-Hb decreases in men aged 65–69 years compared to the other three age groups. Perhaps this is because men in this age group have participated in a number of screening rounds and the prevalence of CRC in this group has become lower through successful screening with FIT. This decrease in f-Hb does not occur with women in Italy, possibly because, as documented here, women have lower f-Hb and, with the use of one only non-partitioned cut-off f-Hb, women are disadvantaged since fewer will be referred for colonoscopy.

The data in Table 1 also show that the distributions of f-Hb in the three countries are not the same and that the 95.0th and 97.5th percentiles of f-Hb are higher in Scotland than in Taiwan, and Italy, for both men and women. Moreover, the data in Tables 2 and 3 show that this holds for the four 5-year age groups studied. This is also shown in Figure 1 which demonstrates the differences between men and women in the four age groups and in the three countries. There are a number of plausible reasons for the differences between countries. The three cohorts studied are of different sizes. Known lifestyle risk factors, including diets high in red and processed meats, lack of physical activity, diets low in fibre, being overweight or obese, possibly drinking alcohol and smoking cigarettes, differ in the three countries. In addition, the number of individuals included with family history of bowel cancer or personal history of polyps or inflammatory bowel disease might be different. Further, the CRC screening programmes in Scotland, Taiwan and Italy have been in existence for different time spans and use different approaches: in consequence, the populations examined here might differ with regard to the number of prevalence and incidence screens and the previous number of screening test exposures. The differences might be due to variations in sample handling, transport and storage, temperature (ambient and laboratory) or subtle changes in the analytical methods which occurred during the time spans of the data accumulations in the three countries, such as small improvements made by the manufacturer. The differences are unlikely to be due to CRC incidence since the World Age-Standardised Ratios for incidence of CRC [21] are 30.3 per 100,000 for the UK and 33.9 for Italy (unfortunately not documented for Taiwan).

The question again arises as to whether this large database should be used to create sex and age partitioned reference values for each country for f-Hb: the 97.5th percentile in the tables represent the upper reference limits derived exactly according to the Clinical and Laboratory Standards Institute Approved Guideline C28–A3c [22]. This concept has been discussed and dismissed previously [14], fixed decision limits being more useful for classification of participants in screening programmes. However, the important deduction is that a single f-Hb cut-off for use in CRC screening programmes is inadequate for use in both sexes and all ages and the f-Hb cut-off used in one programme cannot be transferred to another with the expectation that the outcomes, including positivity rate, positive predictive value, sensitivity, specificity and other crucial clinical outcomes, will be the same.

Moreover, the data from Italy [17, 18] allowed some initial comparison of distributions by sex and age across three different analytical systems; the data (not documented here and possibly to be published elsewhere) show that the distributions differ between analytical systems and this holds for the four age groups studied. This is another important finding since it means that data on f-Hb, irrespective of whether partitioned by sex and/or age, are not transferable between FIT analytical systems. Progress in standardisation has been made over recent years, possibly stimulated by the work of an Expert Working Group on FIT for Screening set up by the Colorectal Cancer Screening Committee of the World Endoscopy Organization, which has published a series of informative discussion documents and papers [23]. Despite this, an important ancillary conclusion is that use of a single f-Hb cut-off for use in CRC screening with one analytical system cannot be transferred to another system even if μg Hb/g faeces units are used.

In conclusion, f-Hb do vary with sex and age, being higher in men and the more aged, but the degree of variation is inconsistent across countries. This means that data on f-Hb are not transferrable across geography and any single f-Hb cut-off will give different outcomes in different countries. Further, the three populations examined here are inhabited predominantly, or almost exclusively, by a single ethnic group: two or more populations in a single country with different ethnic backgrounds and lifestyles might have different distributions of f-Hb. Our analysis supports the view that setting of a single f-Hb cut-off in any CRC screening programmes is far from ideal. We suggest that individualisation of CRC screening is the optimum approach with f-Hb in an individual, alone or with other important factors such as sex and age, be used to determine important personal issues such as the need for colonoscopy, the screening interval between tests and the risk of future CRC. It is usual to monitor physiological variables, such as blood pressure and body mass index, and other quantities frequently examined in laboratory medicine, such as serum cholesterol and plasma glucose. Whether there is merit in monitoring f-Hb similarly in individuals over time and the optimal strategy to undertake this remain interesting research questions for the future.

Conflict of interest statement

Authors’ conflict of interest disclosure: The authors stated that there are no conflicts of interest regarding the publication of this article.

Research funding: This study had no specific research funding and funding played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.

Employment or leadership: CGF held consultancy contracts with Immunostics Inc., Ocean, NJ, USA, and Mode Diagnostics, Glasgow, Scotland, during part of this work.

Honorarium: None declared.

Ethical approval and informed consent: Neither ethical approval nor individual informed consent were required for this study.