Methods

Top Methods Results Discussion Author & Article Info References

Data Source

The National Center for Health Statistics at the Centers for Disease Control and Prevention (CDC) conducted NHANES III to assess the health and nutritional status of the civilian, noninstitutionalized population of the United States [15]. Details of the survey's design, sample, and operation have been published elsewhere [15]. A single phlebotomy was performed on the nationally representative sample examined in the survey at any time between 7 a.m. and 10 p.m. Persons who were scheduled for examination between 8 a.m. and noon were asked to fast for 12 hours, and persons who were scheduled for examination after noon were asked to fast for 6 hours.

We limited our study sample to the 15 839 men and nonpregnant women 20 years of age and older for whom data on serum transferrin saturation was available, regardless of time of blood collection or length of fast. Not all persons selected for inclusion in the NHANES III sample agreed to be interviewed or examined. Thus, our study sample comprises 68% of the nonpregnant adults who were originally selected for NHANES III, 84% of those who were interviewed, and 96% of those who had physical examinations. The NHANES III study was approved by the institutional review board of the National Center for Health Statistics. All respondents provided written informed consent [15].

Variables

The biochemical variables used in our study were serum iron level, total iron-binding capacity, serum ferritin level, and C-reactive protein level [16, 17]. Serum transferrin saturation was calculated by dividing serum iron level by total iron-binding capacity. Assays for serum iron level, total iron-binding capacity, and serum ferritin level were performed by the Division of Environmental Health Laboratory Sciences, National Center for Environmental Health, CDC (Atlanta, Georgia). Serum iron levels and total iron-binding capacity were measured colorimetrically by using an Alpkem RFA analyzer (Alpkem, Clackamas, Oregon); a 1% thiourea solution was added to complex Cu⁺⁺ to prevent copper interference [16]. Serum ferritin levels were measured by using the Quantimmume IRMA kit (BioRad Laboratories, Hercules, California).

Respondents were given the results of their serum ferritin tests, and they received a special alert if their serum ferritin levels exceeded 2000 µg/L. Because no widely accepted criteria for elevated serum transferrin saturation existed at the time of NHANES III, respondents were not given the results of tests for this variable. C-reactive protein levels, which reflect the presence or absence of inflammation, were measured at the Immunology Division of the University of Washington (Seattle, Washington) with latex-enhanced nephelometry.

The technician who performed each phlebotomy recorded the time of blood collection. Length of fast was the respondent-reported time since food or liquid, except water, had last been consumed. Ethnicity was also reported by the respondent, and ethnic categories were based on U.S. Bureau of the Census definitions. We limited comparisons of ethnic groups to non-Hispanic white persons (n = 6752), non-Hispanic black persons (n = 4227), and Mexican-Americans (n = 4233) because there were too few persons in other ethnic groups to allow us to report data separately for those groups.

Definition of Elevated Serum Transferrin Saturation and Elevated Serum Ferritin Level

Elevated serum transferrin saturation was defined as greater than 45%, greater than 50%, greater than 55%, greater than 60%, or greater than 62%; all of these cut-off values had previously been proposed or used in population-based studies of elevated serum transferrin saturation [1-3, 7-14]. We defined elevated serum ferritin levels as greater than 400 µg/L for men, greater than 200 µg/L for women 20 to 49 years of age, and greater than 300 µg/L for women 50 years of age and older [2, 18]. To limit the number of analyses when we compared prevalences of elevated serum transferrin saturation according to ethnicity or age, we used cut-off values of greater than 60% for men and greater than 55% for women rather than applying the full range of values. We chose these values because they had previously been used in a study of elevated serum transferrin saturation done by the CDC [19].

Data Analysis

We used sampling weights to calculate prevalence estimates; these weights account for oversampling and nonresponse to the household interview and physical examination but not for nonparticipation in phlebotomy or for missing data on serum transferrin saturation. The sampling weights were based on the March 1990 and March 1993 Current Population Survey, adjusted for undercounts, for the civilian noninstitutionalized U.S. population [20]. We performed all analyses with SUDAAN [21], a family of statistical procedures for analysis of data from complex sample surveys. To estimate the total number of persons in the United States with elevated serum transferrin saturation, we multiplied the prevalences found in our study with the appropriate census count for the noninstitutionalized U.S. civilian population 20 years of age and older at the midpoint of the survey. In our results, we note estimates that do not meet the criteria for statistical reliability on the basis of sample size (which ranged from 39 to 872 persons, depending on the estimate) and relative SE (SE/prevalence, <30%) [22]. These estimates are presented here because of the importance of hemochromatosis to the public health, even though NHANES III was not designed to reliably estimate the prevalences of rare conditions (conditions that occur in <10% of the population) [23].

In preliminary analyses, we found that time of blood collection, length of fast, and C-reactive protein level differed significantly by ethnicity and age. Diurnal variation, fasting, and inflammation are known to affect serum transferrin saturation independently of iron status [24-27]. Thus, we compared the prevalence of elevated serum transferrin saturation by ethnicity and age in both the total study sample and in a subsample of participants who had blood drawn before noon, fasted for at least 12 hours, and had a C-reactive protein level less than 1 mg/dL [16].

Results

Top Methods Results Discussion Author & Article Info References

The prevalence of elevated serum transferrin saturation ranged from approximately 1% to 6% in the total U.S. adult population (sexes combined), depending on the cut-off value used to define elevated serum transferrin saturation (Table 1). When the same value was used to define elevated serum transferrin saturation for men and women, the prevalence of elevated serum transferrin saturation was 1.3 to 1.8 times greater in men than in women.

The prevalence of elevated serum transferrin saturation was similar in non-Hispanic black men and non-Hispanic white men, but it was lower in non-Hispanic black women than in non-Hispanic white women (Figure 1). The prevalence of elevated serum transferrin saturation was not statistically significantly different in Mexican-Americans and non-Hispanic white persons for either sex. Results for non-Hispanic black women must be interpreted with caution because prevalence estimates for this group did not meet the criteria for statistical reliability. For both men and women, the prevalence of elevated serum transferrin saturation in adults younger than 50 years of age was as great as or greater than it was in the two older age groups (Figure 2). However, because estimates for several groups did not meet the criteria for statistical reliability, these findings must be interpreted with caution.

View larger version (21K): [in this window] [in a new window]   Figure 1. Prevalence of elevated serum transferrin saturation by ethnicity in National Health and Nutrition Examination Survey III (1988-1994). Elevated serum transferrin saturation is defined as greater than 60% for men and greater than 55% for women. White bars represent non-Hispanic white persons, black bars represent non-Hispanic black persons, and striped bars represent Mexican-Americans. Error bars represent 95% Cls. The asterisk indicates findings that may be statistically unreliable.

View larger version (23K): [in this window] [in a new window]   Figure 2. Prevalence of elevated serum transferrin saturation by age in National Health and Nutrition Examination Survey III (1988-1994). Elevated serum transferrin saturation is defined as greater than 60% for men and greater than 55% for women. White bars represent persons 20 to 49 years of age; black bars represent persons 50 to 79 years of age; and striped bars represent persons 80 years of age and older. Error bars represent 95% Cls. Asterisks indicate findings that may be statistically unreliable.

When we restricted our sample to persons who had phlebotomy in the morning, fasted for at least 12 hours, and had normal C-reactive protein levels, we found the same prevalence pattern by age (data not shown). The pattern of age-adjusted prevalence by ethnicity was also similar in the total and restricted samples, except that in the restricted sample, prevalence was slightly lower in non-Hispanic black men than in non-Hispanic white men (data not shown).

Among adults who had elevated serum transferrin saturation as defined by various cut-off values, approximately 11% to 22% had a concurrently elevated serum ferritin level (Table 2). Thus, the prevalence of concurrently elevated serum transferrin saturation and serum ferritin level in the total population is approximately 0.2% to 0.7% (data not shown). The greater the cut-off value for elevated serum transferrin saturation, the greater the prevalence of concurrently elevated serum ferritin levels.

Among persons with an elevated serum transferrin saturation, a higher proportion of women than of men also had an elevated serum ferritin level. This may be due in part to the lower cut-off values used to define an elevated serum ferritin level in women.

Discussion

Top Methods Results Discussion Author & Article Info References

Depending on the cut-off value used to define an elevated serum transferrin saturation in screening for hemochromatosis, the percentage of U.S. adults that would have a single elevated serum transferrin saturation and would therefore be targeted for further evaluation ranges from approximately 1% to 6%. When the same cut-off value is used, prevalences are higher in men than in women; however, some have suggested that a lower cut-off value be used to screen women [3]. The prevalence of initially elevated serum transferrin saturations seen in U.S. adults is generally similar to that seen in other population-based studies using similar cut-off values for an elevated serum transferrin saturation [7-12].

It is important to emphasize that our results indicate only the prevalence of initially elevated serum transferrin saturation that would be expected if a screening program for hemochromatosis were implemented. Blood was drawn only once from each respondent in the representative sample in NHANES III; thus, it was not possible to determine the prevalence of persistently elevated serum transferrin saturation in the population. Only some of those persons with an elevated serum transferrin saturation on one occasion would have an elevated value on a second occasion. The percentage of persons with an initial single elevated serum transferrin saturation who have persistently elevated serum transferrin saturation is not well defined, but estimates range from 35% to 50% [1, 6, 7, 9, 10, 13, 14]. Thus, although our results provide vital information needed to estimate the probable outcome of a screening program, they cannot alone justify implementation of such a program. As Cogswell and colleagues note [28], much remains to be learned about the natural history of hemochromatosis and iron overload. Furthermore, many other issues, ranging from ethical and legal considerations to laboratory standardization, must also be considered in the decision about whether to implement a screening program [28].

The lack of a second measure of serum transferrin saturation in NHANES III prevented us from precisely estimating the prevalence of persistently elevated serum transferrin saturation coupled with an elevated serum ferritin level; the College of American Pathologists screening guidelines [2] recommend liver biopsy for persons with both of these conditions. However, we found that 11% to 22% of those with an initially elevated serum transferrin saturation (0.2% to 0.7% of the total population) had a concurrently elevated serum ferritin level. This may approximate the percentage of persons who would be considered for liver biopsy according to this screening algorithm.

We found similar or only slightly different prevalences of initially elevated serum transferrin saturation in the various age and ethnic groups. The significance of this finding is uncertain because we do not know whether the proportion of those with an initially elevated serum transferrin saturation who ultimately receive a diagnosis of hemochromatosis differs by age or ethnicity. The prevalence of both classic, HLA-linked hemochromatosis and HFE mutations seems to be less common in nonwhite persons [2, 29, 30]. However, a primary iron overload syndrome that differs from classic hereditary hemochromatosis in clinical symptoms, iron deposition site in the liver, and linkage with the HLA gene has been described in African-Americans [31, 32]. Serum transferrin saturation in African-Americans who had received a diagnosis of iron overload was lower than that typically associated with hereditary hemochromatosis [31]; this suggests that a lower cut-off value for elevated serum transferrin saturation might be more appropriate for screening in black persons. More information (including an evaluation of serum transferrin saturation as a screening tool) is needed about hemochromatosis and primary iron overload syndromes in nonwhite persons.

Limitations of our study include the variability introduced by collection of blood at different times of day, variable fasting times, and possible confounding of serum transferrin saturation and serum ferritin levels by the presence of infection or other inflammatory conditions. Serum iron levels, and hence serum transferrin saturations, are subject to diurnal variation: They are highest in the morning, intermediate in the afternoon, and lowest in the evening [17, 24, 25]. Some authors have reported that patients with hemochromatosis have elevated serum transferrin saturations throughout the day [33], but diurnal variation may still affect the cost of a screening program if it increases the number of persons without hemochromatosis who are inappropriately identified for follow-up testing. Serum iron levels increase after consumption of iron [26], and fasting reduces within-person variability in serum iron levels [34]. Finally, confounding resulting from inflammation may have affected overall prevalence rates because serum transferrin saturation is depressed and serum ferritin levels are elevated in both acute and chronic inflammatory conditions [27, 35, 36]. However, we found that patterns of elevated serum transferrin saturation by age and ethnicity were similar in the complete study sample and in the subsample that was created to minimize confounding from fasting, diurnal variation, and inflammation. This suggests, but cannot firmly establish, that the effect of these potential biases was small. In addition, because a population-wide screening program would probably be conducted under circumstances similar to those used in NHANES III, the estimates that were not adjusted for these factors are probably the most relevant to anticipation of the outcome of a population screening program.

Another possible limitation of our study is non-response bias. We did not obtain usable data on serum transferrin saturation for approximately 32% of the originally selected study sample. Nonresponse bias in the NHANES III sample was reduced by an adjustment factor included in the calculation of the sample weights [37]. This adjustment does not address nonresponse among the 4% of examined adults for whom data on serum transferrin saturation were missing, but this small proportion is unlikely to have introduced serious bias. In addition, a previous nonresponse study found that adult respondents and nonrespondents to various parts of the survey did not seem to differ substantially in characteristics relevant to iron status [38]. Thus, our estimates are unlikely to be seriously biased by non-response. Of note, our estimates apply only to non-institutionalized adult populations because institutionalized persons were excluded from the NHANES III sample by design.

By using various cut-off values for an elevated serum transferrin saturation, we have estimated the percentage of adults that would be targeted for further testing for hemochromatosis if the first step of a population-based screening program for this disease were implemented in the United States. With the cut-off value for elevated serum transferrin saturation recommended by the College of American Pathologists (60%) [2], the prevalence of elevated serum transferrin saturation in our study multiplied by the most recent census data [21] shows that approximately 1.4 to 2.5 million adults would be targeted for further testing. With cut-off values of greater than 60% for men and greater than 55% for women, approximately 1.8 to 3.8 million adults would receive further testing. In either case, the numbers are not trivial. Our results should be useful to those considering implementation of a population-wide hemochromatosis screening program that uses serum transferrin saturation.