Cross-sectional relationship between haemoglobin concentration and measures of physical and cognitive function in an older rural South African population

Background Age cohort differences in haemoglobin concentrations and associations with physical and cognitive performance among populations of lower income and middle-income countries have not previously been described. We examined the association between these factors among older men and women in rural South Africa. Methods We analysed cross-sectional data from a population-based study of rural South African men and women aged 40 and over (n=4499), with data drawn from questionnaire responses, a cognitive battery, objective physical function tests and blood tests. Anaemia was defined as a haemoglobin concentration <12 g/dL for women and <13 g/dL for men. We related haemoglobin concentrations to each of age, grip strength, walk speed and a latent cognitive function z-score for men and women separately. We used unadjusted correlations and linear models to adjust for comorbidities and inflammation. Results In total, 1042 (43.0%) women and 833 (40.1%) men were anaemic. Haemoglobin concentrations were inversely correlated with age for men but not for women; in adjusted analyses, haemoglobin was 0.3 g/dL lower per decade older for men (95% CI 0.2 to 0.4 g/dL). In adjusted analyses, haemoglobin concentration was independently associated with grip strength in women (B=0.391, 95% CI 0.177 to 0.605), but this did not reach significance in men (B=0.266, 95% CI −0.019 to 0.552); no associations were observed between haemoglobin levels and walk speed or cognitive score. Conclusions Anaemia was prevalent in this study population of middle-aged and older, rural South African adults, but in contrast to high-income countries, it was not associated with poor physical or cognitive function. Our findings need to be replicated in other populations.

Abbreviated Mental Test Score and the Mini-Mental Status Examination. Scores on the immediate and delayed recall items have previously been associated with risks of dementia and all-cause mortality in older adults, and the two numeracy items reflect an ability-based outcome of education (i.e. numeracy) that has direct relevance for daily functioning and well-being. The cognitive battery was translated and back-translated from English to the local Shangaan language and was pilot tested prior to the HAALSI study with a community-based advisory panel of 90 local adults aged 40 and over, none of whom were part of the later cohort enrolment.
A z-standardized, general latent cognitive factor was derived from the individual cognitive tests using confirmatory factor analysis, the methods and results of which are described in detail elsewhere [1]. The latent cognitive function z-score should be interpreted in the same way as any zscore (i.e. the mean is 0 and the standard deviation is 1), and conceptually it makes the assumption that scores on each individual cognitive test are reflective of a general underlying cognitive ability.
This method improves upon the use of a composite total score (i.e. summed cognitive battery score out of 26 points), as it reduces measurement error that may be present in any single cognitive item by only utilizing the covariation between items in generating the latent cognitive z-score, and it differentially weights the individual items according to their contribution to the latent cognitive zscore. For example, a composite total score would assign equal weight to knowledge of the current day (one point) with the ability to remember an additional word from the word recall list (one point), although these two cognitive tasks may vary in difficulty and may have differential relationships to functional outcomes and future health risk. The confirmatory factor analysis model was of good fit to the data, with a root mean square error of approximation of 0.032 (95% CI: 0.025, 0.039), comparative fit index of 0.998, and Tucker-Lewis index of 0.997 [1].

Covariates
Age was ascertained by self-report from participants, rounded to the nearest year. Sex was ascertained by self-report. Educational status was based on the highest year of school attended; none, years 1-7, denoting some primary-level education; years 8-11, denoting some secondary education, or years 12 and above, denoting completed secondary education or post-secondary education. Selfreported angina, chronic bronchitis and stroke were ascertained by a series of questions asking whether a physician had ever diagnosed these conditions or whether participants were on treatment for these conditions. In addition, the Rose angina questionnaire was administered to ascertain whether angina symptoms were present [2]. Diabetes mellitus was ascertained by asking if a participant was taking medication for the condition, if the participant had ever been told that they had diabetes, or by a random capillary glucose concentration >11.0mmol/L obtained as part of the HAALSI survey. Blood pressure was measured in the seated position; three readings were taken and the mean of the second and third reading was used as the measure of blood pressure. Hypertension was ascertained by asking if the participant was taking medication for hypertension at the time of interview, or by a systolic blood pressure >140mmHg, or diastolic blood pressure >90mmHg at the time of interview. C-reactive protein was measured from dried blood spots using a high-sensitivity sandwich ELISA method as described previously [3]. The lower limit of sensitivity for the assay was 0.05mg/L. HIV status was ascertained by analysis of dried blood spot samples using the Vironostika Uniform 11 assay (Biomerieux, France).

Supplementary Table 1. Mean haemoglobin levels according to the presence of individual comorbidities
Hb concentration (g/dL) (SD) -men Hb concentration (g/dL) (SD) -women