Associations Between Functional Biological Age and Cognition Among Older Adults in Rural Bangladesh: Comparisons With Chronological Age

Abstract

Objectives: We constructed a functional biological age (fBioAge) indicator by using four functional variables: grip strength, forced expiratory lung volume, visual acuity, and hearing. Our aim was to compare how chronological age (ChronAge) and fBioAge are related to cognitive abilities in older adults. Method: We used data from the Poverty and Health in Aging project, Bangladesh. Participants (N = 400) were 60+ years of age and diagnosed as nondemented. Examined cognitive abilities were four episodic memory measures (including recall and recognition), two verbal fluency indicators, two semantic knowledge, and two processing speed tasks. Results: fBioAge accounted for cognitive variance beyond that explained by ChronAge also after controlling for medical diagnoses and blood markers. Discussion: Compared with ChronAge, fBioAge was a stronger predictor of cognition during a broad part of the old adult span. fBioAge seems, in that respect, to have the potential to become a useful age indicator in future aging studies.

Keywords

age indicator cross-sectional cognitive abilities

Introduction

In many aging studies to date, chronological age (ChronAge) has been used as a proxy of the aging processes. However, ChronAge is something defined and measures time, not aging (Spiro, 2007). To be 1 year older does not necessarily mean that a person has aged the equivalent of 1 year. This is because the aging process is, to a high degree, individual. For example, a person of 85 years may in several respects be more resilient than a person who is 75 years of age. In seeking alternatives to ChronAge, some argue that other referral points should be used, such as time to death or time from onset of disease (Spiro, 2007), whereas others suggest using markers of biological aging (Anstey & Smith, 1999). The interest in biological markers started already in the middle of the past century (e.g., Birren, 1959). However, no single marker of biological aging seems yet to have enough predictive power to be considered as the complete biological age predictor, even if the epigenetic clock shows promising results (e.g., Jylhävä, Pedersen, & Hägg, 2017). Instead, composites of selected biomarkers seem to have more predictive value in predicting mortality, and would probably be more reliable indicators of biological age across the life span (Jylhävä et al., 2017). There is, however, little consensus in choosing the biomarkers to be included in a composite of biological age (Levine, 2013). The rate of aging is, for example, not the same for all parts of the body (e.g., Raz et al., 2005). One recent promising attempt to create a composite was made by Levine (2013), and the composite of 10 biological markers was a more reliable predictor of mortality than ChronAge in a sample of 9,389 participants aged 30 to 73 years. This composite was also later validated by Belsky et al. (2015) in a younger sample, where the 954 participants were assessed at 38 years of age. However, the selection of biological markers in the Levine (2013) study was partially due to availability in the Third National Health and Nutrition Examination Survey (NHANES III) study, and therefore, the author suggested to study other combinations of biological markers to better capture the complexity of biological aging. One type is functional biomarkers (e.g., grip strength and lung capacity), and some studies have combined them into a measure called functional biological age (fBioAge; for example, Anstey & Smith, 1999; Wahlin, MacDonald, deFrias, Nilsson, & Dixon, 2006). We wanted to use an age indicator relevant for the study of cognitive aging. Several cognitive abilities seem to deteriorate in general with increasing ChronAge (Kramer, Bherer, Colcombe, Dong, & Greenough, 2004), but there are large individual variations. Differences exist between cross-sectional and longitudinal results of age effects in cognitive abilities (Rönnlund, Nyberg, Bäckman, & Nilsson, 2005), where longitudinal results are generally much more positive and, for example, onset of decline is much later. In psychology, central aspects are behavior and perception (Anstey, 2008), and therefore, we used an age indicator that captures aging in those aspects. In the present study, the fBioAge indicator has included four measures: grip strength, lung function, vision, and hearing. Sensory and cognitive abilities seem to decline in a rather similar fashion in old age (Baltes & Lindenberger, 1997; Lindenberger & Baltes, 1994), and motor functions, such as grip strength and lung capacity, also show relationships with cognitive decline (e.g., Emery, Finkel, & Pedersen, 2012; Sternäng, Reynolds, et al., 2016). The choice of biomarkers was based on a review of functional age studies by Anstey, Lord, and Smith (1996), and on availability in the present study, and this fBioAge indicator does not need, for example, blood sampling or the use of complicated measures. Recent longitudinal findings from our own group indicate that fBioAge does capture individual differences in aging, both between and within individuals (Finkel, Sternäng, & Wahlin, 2017). A well-functioning cognitive performance is vital for any individual. Little, however, is known about the interrelationships between ChronAge, fBioAge, and cognition. The few studies that exist indicate that fBioAge may be a better predictor of cognitive performance than ChronAge (e.g., MacDonald, Dixon, Cohen, & Hazlitt, 2004; Wahlin et al., 2006). In a longitudinal study, MacDonald et al. (2004) used vision, hearing, lung function, and grip strength to create a composite of biological age. This composite showed association with cognition beyond that from ChronAge. Cross-sectional studies by Anstey and Smith (1999), using a composite of vision, hearing, lung function, grip strength, and vibration sense, and Wahlin et al. (2006), using vision, hearing, lung function, and grip strength in their composite, showed that fBioAge explained much of the ChronAge-related variance in cognitive abilities, such as working memory, reasoning, and episodic memory. To validate fBioAge as an aging indicator for the field of cognitive aging, an important part is how well fBioAge predicts cognitive performance in old age compared with ChronAge.

We used data from the population-based, cross-sectional Poverty and Health in Aging (PHA) project in Bangladesh (Kabir et al., 2006). Bangladesh is a low-income country, and the elderly population have to deal with problems such as poverty, malnutrition, and limited access to health care (Ferdous, 2009). Life expectancy at birth was 69.9 years for men and 72.3 years for women in 2015 (United Nations Department of Economic and Social Affairs: Population Division, 2015). In 2015, 7% or 11.2 million people in Bangladesh were aged 60 years or above, which is expected to increase to 11.5% or 21.5 million in 2030 (United Nations Department of Economic and Social Affairs: Population Division, 2015).

The ChronAge/fBioAge distinction is of relevance to study in low-income countries, where determining the participant’s age is not always straightforward. It is also of relevance for future comparisons within low-income countries or with data from high-income countries. To the best of our knowledge, no study of BioAge and cognition has yet been carried out in low-income countries.

The aim of the present study was, therefore, to compare how ChronAge and fBioAge relate to cognitive abilities (in this study: episodic recall and recognition, verbal fluency, semantic knowledge, and processing speed) among older adults (60+). Furthermore, we wanted to examine whether such associations differed between persons who have aged faster or slower than their ChronAge, and whether these relationships were at least partly accounted for by medical diagnoses and blood markers. Fast and slow aging has, to the best of our knowledge, never been examined and contrasted before. Our hypotheses were (a) fBioAge is a stronger predictor than ChronAge of cognitive performance in the older ages (60+). Longitudinal findings (Finkel et al., 2017) demonstrate that trajectories of fBioAge and ChronAge in old adulthood differ considerably, which probably influences their associations with cognition. (b) Most of the associations between age measures and cognition will stay significant even after controlling for medical diagnoses and blood variables. Related research has shown that relationships between age and cognition are only partly explained by morbidities and health status (Anstey & Christensen, 2000; Bäckman, Small, Wahlin, & Larsson, 2000; Salthouse, Kausler, & Saults, 1990; Wahlin, 2004). (c) The relationships between the age measures (i.e., fBioAge and ChronAge) and cognition do not differ across the fast and slow aging groups. Previous research has not contrasted these two groups, and we have no a priori reason to believe initially that these relationships are different just because a person belongs to a fast or a slow aging group.

Method

Participants

The participants were recruited from the PHA project in Bangladesh (Kabir et al., 2006). The main aim of the cross-sectional PHA project is to study how social, cognitive, and medical/biological factors are related to each other in a sample of older participants (60+) from Matlab, a rural region approximately 60 km southeast of the capital Dhaka. In all, 471 participants underwent interviews and cognitive tests, and after exclusion of participants with dementia and questionable dementia (Palmer et al., 2014), a final sample of 400 participants (214 women and 186 men) remained. Dementia is probably the most severe disease for memory abilities (Nilsson, Nyberg, & Bäckman, 2002). The prevalence of dementia increases rather quickly with ChronAge above 65 years of age. In an European population, the prevalence increased from 0.8% in the age group 65 to 69 years to 28.5% in the age group above 90 years of age (Lobo et al., 2000). There is evidence that the preclinical phase of dementia can affect different cognitive abilities some years before the actual diagnosis (for a meta-analysis, see Bäckman, Jones, Berger, Laukka, & Small, 2005). It was important to screen for preclinical dementia, especially since episodic memory (one of the dependent variables in this study) has proved to be the most affected cognitive ability in preclinical Alzheimer’s disease (Bäckman, Small, & Fratiglioni, 2004). Dementia screening in PHA was made by geriatricians based on criteria according to Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; American Psychiatric Association, 1994). Approximately 60% of the participants were nonliterate.

Cognitive Abilities

We used four types of memory measures (episodic recall, recognition, verbal fluency, and semantic knowledge) and processing speed. All the four memory measures can be considered as different aspects of long-term declarative memory, and processing speed as a measure of cognitive processing efficiency. Verbal fluency may also measure aspects of executive functioning (Shao, Janse, Visser, & Meyer, 2014). The structure of between-person differences in these five cognitive functions has been examined and fits data from this population in rural Bangladesh well (Sternäng, Lövdén, Kabir, Hamadani, & Wahlin, 2016). To be able to assess both nonliterates and literates, the tests materials comprised of, where appropriate, pictures and objects instead of words and letters. Cognitive factor scores from the study by Sternäng, Lövdén, et al. (2016), based on two tests in each cognitive ability, were used. Both in the study by Sternäng, Lövdén, et al. (2016) and in this study, the five cognitive variables were considered as separate, but interrelated abilities of cognition.

Episodic immediate recall

The first test was a recall test of random nouns where the investigator read aloud the 12 words (5 s per word). Number of correctly recalled words constituted the outcome. In the second test, recall of 12 pictures of objects was tested, where the investigator read and showed the pictures simultaneously (5 s per object). The score was the number of correctly recalled objects. A total of 2 min were allotted for each test.

Recognition

The first recognition test was a free-choice (yes/no) recognition of 12 random words (the same words as in recall Test 1) together with 12 distracters. The second test was a test of recognition of 12 pictures of objects (the same pictures of objects as in recall Test 2) together with 12 distracters. In both tests, the score was the number of hits minus false alarms. Each recognition test followed immediately after its respective recall version.

Verbal fluency

In the first verbal fluency test, the participant was instructed to say as many different animals as possible in 1 min. In the second test, the instruction was instead to say as many different exemplars of food as possible in 1 min. In both tests, the score was the number of correct responses.

Semantic knowledge

Semantic knowledge was tested first with a test of the meaning (synonyms) of 20 words. Three alternatives for each target word were given. The outcome was the number of correctly identified words. The second knowledge test included 20 knowledge questions. The score of this test was the number of correct answers.

Processing speed

The first processing speed task was to complete as many boxes with a missing side as possible during 1 min. The second task was to cross as many balls among other distracter symbols as possible during 30 s. The scores in the tasks were the number of boxes completed and the number of balls crossed, respectively.

Independent Variables

Age measures

ChronAge

ChronAge was indexed as time since birth. In many cases in this region, birth certificates are lacking. Therefore, as a well-established method in low-income countries, a participant’s age was determined with questions related to historical and life events (see Allain et al., 1997). In this sample, no one had birth certificates. In a few cases (number not known), age was self-reported by the participants who said they knew their age. For others, an “events (biological & historical events) calendar” was used to assess age. For example, do you remember the law introduced against child marriage? (this was a period when many small children, especially girls were married off to bypass the marriage act introduced in 1929). Do you remember the partition of India? Do you remember the great famine of 1943? In case the person was born before such an event, example follow-up questions would be as follows: Whether the person had started school by then or whether she got married then, whether she had her first menstruation when she was married (estimating age 12 for first period), or whether she or he had lost her first tooth. So the assessment was based on individual specific combination of historical, biological, and family-related events.

fBioAge

A composite of fBioAge was constructed from the following four variables: grip strength, forced expiratory volume (FEV), vision, and hearing. Grip strength was measured 3 times for each hand with a dynamometer (in kilogram). The mean of the highest grip strength score for each hand was used in the present study. Lung function was measured with FEV, and the best score of three attempts was recorded. This measure was corrected for body volume through dividing it by the individual’s squared height (in square meter). Visual acuity was measured for both eyes with an analog Snellen chart (McGraw, Winn, & Whitaker, 1995) during the clinical exam. To enable inclusion of nonliterate participants, symbols of different sizes were used on the chart instead of letters. The instrument was scaled from 1 to 13, where 1 was the best visual acuity and 13 was the worst (no vision). The score was the mean of both eyes. Auditory acuity was tested with a tuning fork during the clinical exam. Participants were examined for each ear, with respect to whether they could hear a tuning fork. The auditory acuity variable was trichotomous: 1 = the participant was able to hear with both ears, 2 = the participant could only hear in one ear, and 3 = the participant was not able to hear the tuning fork in either ear.

To create a fBioAge composite, all four variables above were z transformed. To control for trivial sex differences in average muscle strength, the z transformation of grip strength was done separately for women and men. The variables grip strength and FEV were also multiplied by minus 1 so that a higher score corresponded to a higher age. Finally, all four z transformed variables were summed into a total fBioAge composite. After having calculated the fBioAge composite, 10% of the sample was removed, that is, the 5% oldest and the 5% youngest participants according to their ChronAge. This allowed for the possibility of the remaining participants at both ChronAge ends to be biologically younger or older than their ChronAge (cf. Krøll & Saxtrup, 2000). This resulted in a total final sample of 360 participants (for a description of the sample, see Table 1). The correlation between fBioAge and ChronAge was .274 (p < .0005) in the final sample.

Table 1.

Description of the Final Sample.

Variable	Total sample	Fast aging group	Slow aging group	Mann–Whitney U test (z values)
n	360	201	159
Age	68.6 (5.4)	66.3 (4.2)	71.6 (5.5)	−8.92***
Women	52%	65%	36%	−5.52***
Years of schooling	2.10 (3.11)	1.39 (2.33)	3.00 (3.70)	−4.46***
Nonliterates	59%	68%	47%	−3.92***
Married	61%	58%	65%	−1.48
Cognitive variables
Verbal fluency	10.98 (2.49)	10.47 (2.34)	11.61 (2.53)	−4.37***
Recall	3.80 (0.67)	3.73 (0.67)	3.90 (0.67)	−2.48*
Recognition	5.81 (1.42)	5.69 (1.49)	5.97 (1.31)	−1.85
Processing speed	26.76 (8.95)	24.44 (7.70)	29.69 (9.58)	−5.20***
Semantic knowledge	12.70 (2.32)	12.15 (2.10)	13.40 (2.39)	−4.84***
Diagnoses
Depression	164	101	63	−2.01*
Back pain	282	169	113	−2.97**
Arthralgia	203	118	85	−1.00
Arthritis	209	121	88	−0.93
Elevated blood glucose	5	4	1	−1.09
Stroke	2	1	1	−0.17
Obesity	11	8	3	−1.14
Signs of hypofunction in the thyroid gland	36	21	15	−0.32
Headache	188	113	75	−1.71
Symptoms of helminthiasis	253	144	109	−0.64
Leucorea	41	34	7	−2.13*
Jaundice	7	5	2	−0.84
Gastro intestinal or upper alimentary tract disorder	267	154	113	−1.19
Lower alimentary tract disorder	46	20	26	−1.80
Uterine prolapse	31	21	10	−0.21
Obstructive pulmonary symptoms	25	20	5	−2.52*
Symptoms of heart failure	17	12	5	−1.25
Respiratory tract infection	29	22	7	−2.26*
Impaired vision	136	94	42	−3.95***
Skin disease	182	102	80	−0.08
Hearing impairment	7	7	0	−2.37*
Oral conditions	92	63	29	−2.83**
Elevated blood pressure	139	72	67	−1.22
Low blood pressure		151	121	−0.21
Severe headache	130	92	38	−4.28***
Cataract	225	112	113	−2.98**
Blood markers
Hemoglobin (g%)	11.51 (1.31)	11.31 (1.18)	11.76 (1.43)	−3.19**
Albumin (g/L)	36.54 (2.94)	36.31 (2.95)	36.84 (2.91)	−1.64
Calcium (mmol/L)	2.40 (0.10)	2.39 (0.10)	2.41 (0.11)	−1.16
TSH (µIU/mL)	2.47 (5.97)	2.78 (7.66)	2.07 (2.54)	−0.89
Cholesterol (mmol/L)	4.60 (0.93)	4.64 (0.96)	4.55 (0.89)	−0.73
HDL (mmol/L)	0.81 (0.17)	0.81 (0.17)	0.81 (0.18)	−0.02
LDL (mmol/L)	3.03 (0.81)	3.06 (0.84)	3.00 (0.78)	−0.56
HDL/LDL ratio	0.29 (0.14)	0.29 (0.15)	0.29 (0.13)	−0.50
Triglycerides (mmol/L)	1.67 (0.92)	1.69 (1.00)	1.64 (0.81)	−0.08
Homocysteine (µmol/L)	16.28 (5.69)	15.37 (5.20)	17.43 (6.08)	−3.87***
Vitamin B12 (pg/mL)	625.9 (707.8)	646.7 (596.9)	599.5 (828.3)	−1.52
Folate (nmol/L)	15.33 (8.89)	15.98 (10.10)	14.50 (7.02)	−0.57
Blood glucose (mmol/L)	5.30 (1.96)	5.34 (2.08)	5.25 (1.81)	−0.21

Note. Significance testing of group differences was done using the Mann–Whitney U test (approximated to z values). The same results (i.e., which differences were significant), not shown, were obtained by chi-square tests for frequency data (e.g., diagnoses, sex, and literacy) and by independent-samples t tests for continuous data (e.g., blood markers, cognition, ChronAge, and years of education). TSH = thyroid stimulating hormone; HDL = high-density lipoprotein; LDL = low-density lipoprotein.

p < .05. **p < .01. ***p < .001.

Medical diagnoses

Differences in age-related cognitive performance may be partially explained by morbidities (Anstey & Christensen, 2000; Wahlin, 2004). In the present study, 26 medical diagnoses were used: depression, back pain, arthralgia, arthritis, elevated blood glucose, stroke, obesity, signs of hypofunction in the thyroid gland, headache, symptoms of helminthiasis, leucorea, jaundice, gastro intestinal or upper alimentary tract disorder, lower alimentary tract disorder, uterine prolapse, obstructive pulmonary symptoms, symptoms of heart failure, respiratory tract infection, impaired vision, skin disease, hearing impairment, oral conditions, elevated blood pressure, low blood pressure, severe headache, and cataract. All diagnoses were made by a physician based on the clinical examination. All these variables were dichotomous (1 if diagnosed with the disorder and 0 otherwise).

Blood markers

Thirteen different blood markers were included: hemoglobin (g%), albumin (g/L), calcium (mmol/L), thyroid stimulating hormone (TSH; µIU/mL), cholesterol (mmol/L), high-density lipoprotein (HDL; mmol/L), low-density lipoprotein (LDL; mmol/L), HDL/LDL ratio, triglycerides (mmol/L), homocysteine (µmol/L), vitamin B12 (pg/mL), folate (nmol/L), and blood glucose (mmol/L). Blood sampling was done in the morning of the same day as the medical and cognitive examination. The participants were not fasting. The blood markers in the present study were used as continuous variables.

Two groups with different rates of aging (fast and slow)

To increase our understanding of the difference between ChronAge and fBioAge, we constructed two groups, where the participants in the first group had a fBioAge lower than their ChronAge (the slow aging group) and the participants in the second group had a fBioAge higher than their ChronAge (the fast aging group). Sorting was based on the individual difference between the z scores of the respective age indicator. The coefficient of variation was similar for both ChronAge and fBioAge in the resulting two difference groups.

Statistical Analyses

The analyses were performed in SPSS (version 21). Alpha levels were set at .05. Hierarchical regression models were used to observe whether fBioAge explains additional cognitive variance beyond that explained by ChronAge, or whether ChronAge explains cognitive variance beyond that of fBioAge. In the first model (see Table 2), performed on the two ChronAge groups (60-69.5 years and 69.5+), ChronAge was entered in the first step and fBioAge was entered in the second step. In the second model (see Table 3), the order was switched so that fBioAge instead was entered first and ChronAge second. The third model (see Table 4) included diagnoses and blood markers in the first step, ChronAge in the second step, and fBioAge in the third and final step. Only diagnoses and blood markers that were significantly correlated with cognitive abilities and age were used in the model. This model was performed on the two groups: fast and slow aging. The fourth model (see Table 5) was the same as the third model except that the order was reversed between fBioAge and ChronAge. The first hypothesis in the present study was connected to the regression models reported in Tables 2 and 3, the second hypothesis to the regression models on the total sample reported in Tables 4 and 5, and, finally, the third hypothesis to the regression models performed on the fast and slow aging group reported in Tables 4 and 5.

Table 2.

Age-Related Cognitive Variance in the Two ChronAge Groups (With ChronAge Entered First and fBioAge Second).

Order of entered variables	60-69.5 years, n = 228					>69.5 years, n = 132					Total sample
Order of entered variables	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p
Verbal fluency
1. ChronAge	.00	.00	0.06	−.02	.803	.05	.05	6.50	−.22	.012	.03	.03	10.24	−.017	.001
2. fBioAge	.15	.15	39.90	−.39	.000	.23	.18	30.56	−.44	.000	.19	.16	70.50	−.42	.000
Processing speed
1. ChronAge	.00	.00	0.00	.00	.975	.05	.05	6.02	−.21	.015	.01	.01	5.23	−.12	.023
2. fBioAge	.20	.20	54.25	−.45	.000	.26	.22	37.60	−.48	.000	.22	.20	91.72	−.47	.000
Semantic knowledge
1. ChronAge	.00	.00	0.00	.00	.953	.03	.03	4.48	−.18	.036	.01	.01	3.61	−.10	.058
2. fBioAge	.13	.13	33.28	−.37	.000	.24	.21	34.51	−.46	.000	.17	.16	68.49	−.42	.000
Recall
1. ChronAge	.00	.00	0.41	−.04	.521	.05	.05	6.08	−.21	.015	.05	.05	18.10	−.22	.000
2. fBioAge	.10	.10	24.51	−.32	.000	.19	.14	22.12	−.38	.000	.16	.11	46.20	−.35	.000
Recognition
1. ChronAge	.00	.00	0.01	.01	.930	.02	.02	2.42	−.14	.122	.02	.02	8.36	−.15	.004
2. fBioAge	.05	.05	11.91	−.23	.001	.10	.08	11.50	−.29	.001	.08	.06	23.53	−.26	.000

Table 3.

Age-Related Cognitive Variance in the Two ChronAge Groups (With fBioAge Entered First and ChronAge Second).

Order of entered variables	60-69.5 years, n = 228					>69.5 years, n = 132					Total sample
Order of entered variables	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p
Verbal fluency
1. fBioAge	.15	.15	39.41	−.39	.000	.21	.21	35.19	−.46	.000	.19	.19	81.69	−.043	.000
2. ChronAge	.15	.00	0.63	.05	.428	.23	.02	2.85	−.13	.094	.19	.00	1.04	−.05	.309
Processing speed
1. fBioAge	.19	.19	52.71	−.44	.000	.25	.25	42.57	−.50	.000	.22	.22	98.50	−.47	.000
2. ChronAge	.20	.01	1.44	.07	.231	.26	.01	2.33	−.12	.130	.22	.00	0.04	.01	.834
Semantic knowledge
1. fBioAge	.13	.13	32.21	−.35	.000	.23	.23	38.64	−.48	.000	.17	.17	72.86	−.41	.000
2. ChronAge	.13	.00	1.06	.07	.304	.24	.01	1.39	−.09	.241	.17	.00	0.10	.02	.750
Recall
1. fBioAge	.10	.10	25.05	−.32	.000	.17	.17	25.97	−.41	.000	.14	.14	59.80	−.38	.000
2. ChronAge	.10	.00	0.03	.01	.870	.19	.02	2.85	−.14	.094	.16	.01	5.97	−.12	.015
Recognition
1. fBioAge	.05	.05	11.50	−.22	.001	.09	.09	13.29	−.31	.000	.08	.08	30.03	−.28	.000
2. ChronAge	.05	.00	0.45	.04	.505	.10	.01	0.85	−.08	.359	.08	.01	2.28	−.08	.132

Table 4.

Explained Cognitive Variance in the Slow and Fast Aging Group (With ChronAge Entered Second and fBioAge Third).

Order of entered variables	Slow, n = 159					Fast, n = 201					Total sample
Order of entered variables	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p
Verbal fluency
1. Diagnoses/blood markers	.39	.39	7.77		.000	.19	.19	3.53		.000	.28	.28	11.36		.000
2. ChronAge	.44	.05	11.44	−.24	.001	.21	.03	6.60	−.18	.011	.30	.02	7.89	−.13	.005
3. fBioAge	.46	.02	5.44	−.20	.021	.23	.02	4.59	−.19	.034	.35	.05	27.35	−.27	.000
Processing speed
1. Diagnoses/blood markers	.42	.42	9.52		.000	.24	.24	5.40		.000	.36	.36	17.48		.000
2. ChronAge	.48	.06	16.04	−.28	.000	.26	.02	3.66	−.13	.057	.37	.01	4.61	−.10	.033
3. fBioAge	.49	.02	4.63	−.18	.033	.30	.04	10.83	−.28	.001	.43	.06	38.84	−.30	.000
Semantic knowledge
1. Diagnoses/blood markers	.47	.47	9.65		.000	.28	.28	5.44		.000	.38	.38	16.20		.000
2. ChronAge	.50	.03	9.07	−.20	.003	.30	.02	5.75	−.16	.017	.39	.01	4.42	−.09	.036
3. fBioAge	.51	.01	2.40	−.13	.124	.33	.03	7.40	−.23	.007	.43	.04	25.44	−.24	.000
Recall
1. Diagnoses/blood markers	.25	.25	5.90		.000	.10	.10	2.66		.009	.16	.16	8.13		.000
2. ChronAge	.33	.08	16.73	−.31	.000	.13	.03	5.62	−.17	.019	.19	.03	10.24	−.16	.002
3. fBioAge	.34	.01	2.37	−.17	.126	.15	.02	4.29	−.19	.040	.23	.04	18.08	−.24	.000
Recognition
1. Diagnoses/blood markers	.16	.16	3.97		.001	.12	.12	3.56		.001	.11	.11	6.40		.000
2. ChronAge	.21	.05	9.06	−.25	.033	.12	.01	1.87	−.10	.173	.13	.01	4.61	−.11	.033
3. fBioAge	.21	.01	1.78	−.14	.184	.13	.01	1.32	−.11	.252	.15	.02	8.98	−.17	.003

Table 5.

Explained Cognitive Variance in the Slow and Fast Aging Group (With fBioAge Entered Second and ChronAge Third).

Order of entered variables	Slow, n = 159					Fast, n = 201					Total sample
Order of entered variables	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p	Cumulative R²	Change in R²	F for change	β	p
Verbal fluency
1. Diagnoses/blood markers	.39	.39	7.77		.000	.19	.19	3.53		.000	.28	.28	11.36		.000
2. fBioAge	.45	.05	14.09	−.28	.000	.23	.04	10.73	−.23	.001	.35	.06	32.86	−.29	.000
3. ChronAge	.46	.01	2.95	−.14	.088	.23	.00	0.60	−.07	.438	.35	.01	2.80	−.08	.095
Processing speed
1. Diagnoses/blood markers	.42	.42	9.52		.000	.24	.24	5.40		.000	.36	.36	17.48		.000
2. fBioAge	.47	.06	15.41	−.29	.000	.30	.06	14.55	−.26	.000	.43	.07	43.32	−.31	.000
3. ChronAge	.49	.02	5.21	−.18	.024	.30	.00	0.20	.04	.655	.43	.00	0.68	−.04	.412
Semantic knowledge
1. Diagnoses/blood markers	.47	.47	9.65		.000	.28	.28	5.44		.000	.38	.38	16.20		.000
2. fBioAge	.50	.03	8.58	−.21	.004	.32	.05	13.35	−.24	.000	.43	.05	29.31	−.25	.000
3. ChronAge	.51	.01	2.87	−.14	.093	.33	.00	0.06	−.02	.800	.43	.00	0.87	−.04	.351
Recall
1. Diagnoses/blood markers	.25	.25	5.90		.000	.10	.10	2.66		.009	.16	.16	8.13		.000
2. fBioAge	.31	.07	13.87	−.32	.000	.15	.05	9.75	−.23	.002	.22	.06	24.22	−.27	.000
3. ChronAge	.34	.02	5.00	−.22	.027	.15	.00	0.30	−.05	.584	.23	.01	4.38	−.11	.037
Recognition
1. Diagnoses/blood markers	.16	.16	3.97		.001	.12	.12	3.56		.001	.11	.11	6.40		.000
2. fBioAge	.20	.04	7.45	−.23	.007	.13	.01	3.09	−.13	.080	.14	.03	11.57	−.19	.001
3. ChronAge	.21	.02	3.33	−.18	.070	.13	.00	0.12	−.03	.733	.15	.01	2.08	−.08	.150

Results

Two Age Groups

To examine whether the associations between age measures and cognitive abilities are different across the examined age range, we constructed two age groups. A previous study from our group (Finkel et al., 2017) found that fBioAge starts to accelerate around 75 years of age, and therefore, it was of interest to examine whether the associations between fBioAge and cognition differed in these two age groups. The ChronAge variable was z transformed, and the two age groups were constructed based on the participants’ z values, one group with z values ≤0 (n = 228) and one group with z values >0 (n = 132). The z value zero represented 69.5 years of age. The coefficient of variation (measured as the standard deviation divided by the mean) was similar for both ChronAge and fBioAge in the two groups. The two age groups were only used for selection, both age measures (ChronAge and fBioAge) in the calculations were still used as continuous variables.

Explained variance

fBioAge accounted for additional variance in all the studied cognitive abilities over and above that explained by ChronAge in both age groups (see Table 2). All significance levels for fBioAge in Step 2 were p < .0005. The largest difference between the two age measures was in the ChronAge group 60 to 69.5 years, where ChronAge did not explain any cognitive variance. When fBioAge instead was entered first and ChronAge second, ChronAge did not account for any of the remaining cognitive variance in any of the two age groups (see Table 3). Also in the total sample, fBioAge significantly accounted for additional variance beyond that explained by ChronAge in all five cognitive abilities (see Table 2). When diagnoses and blood markers were controlled for, fBioAge still contributed with additional variance over and above that explained by ChronAge in all five cognitive abilities in the total sample (see Table 4).

The fast and slow aging group

The slow aging group, that is, the group that had not aged as much as their indicated ChronAge, comprised significantly more men and more literate participants than the other group (see Table 1). Participants in the slow aging group were significantly older chronologically and had more years of education than the fast aging group. Participants in the slow aging group were also significantly less likely diagnosed with nine of the 26 diagnoses, that is, depression, back pain, leucorea, obstructive pulmonary symptoms, respiratory tract infection, impaired vision, hearing impairment, oral conditions, and severe headache, but more likely diagnosed with one diagnosis, cataract. Significant differences between the two groups were found in two of the 13 studied blood markers, in hemoglobin and homocysteine. Participants in the slow aging group demonstrated higher levels in both. The participants in the slow aging group performed also significantly better in verbal fluency, recall, semantic knowledge, and processing speed.

Explained variance

In the slow aging group, fBioAge contributed with additional variance beyond that explained by diagnoses, blood markers, and ChronAge in verbal fluency and processing speed (see Table 4). In the fast aging group, fBioAge contributed with additional variance in verbal fluency, episodic recall, processing speed, and semantic knowledge. When the opposite order between ChronAge and fBioAge was used, ChronAge accounted for variance beyond that of fBioAge only in processing speed and recall in the slow aging group and not at all in the fast aging group (see Table 5).

Discussion

This is the first study about fBioAge and cognition in a low-income country. Our aim was to compare ChronAge and fBioAge, and to examine their associations with some cognitive abilities in the old age range (60+). fBioAge was overall a stronger predictor of cognition than ChronAge. This was also true when controlling for diagnoses and blood markers.

fBioAge accounted for additional variance beyond that explained by ChronAge in all five studied cognitive abilities. This was in accordance with our first hypothesis. In participants between 60 and 69.5 years of ChronAge, ChronAge explained practically no cognitive variance. One reason that fBioAge was a stronger predictor of cognition than ChronAge might be that the effects of the aging processes differ in the older age group (>69.5 years of age) compared with the younger age group (≤69.5 years of age), and that the fBioAge variable captures this better. According to longitudinal results from our own group (Finkel et al., 2017), fBioAge does not seem to increase much, in general, before age 75 and instead starts to accelerate after that age. Cognitive performance normally does not change so much in middle age, chronologically measured (Rönnlund et al., 2005), whereas ChronAge instead increases steadily with 1 year per year and probably does not reflect the cognitive aging effects well during this time period. The aging processes are individual to a high degree (Gunn et al., 2009) and fBioAge can better capture that (Finkel et al., 2017). As the sample was divided according to the participants’ ChronAge, the results could have been affected by the restriction of range in ChronAge. However, also in the total sample, fBioAge accounted for additional cognitive variance beyond that explained by ChronAge (see Table 1). These results were rather stable in the total sample, and stayed significant even after controlling for medical diagnoses and blood markers, which is in accordance with our second hypothesis. Of the five tested cognitive measures, processing speed showed the strongest associations with the two age measures, which is in line with what the processing speed account states (Salthouse, 1996; Verhaeghen & Salthouse, 1997). As fBioAge was significantly associated with all the tested cognitive abilities, it seems that fBioAge is a valid predictor of both fluid and crystallized cognitive abilities. This is clearly important if fBioAge is going to be used as an alternative to ChronAge in cognitive aging studies including different types of cognitive abilities.

We also compared two groups that had aged functionally faster and slower than their ChronAge. These groups are interesting to examine, especially the slow aging group where the participants were healthier in general (see Table 1). Do the participants in the slow aging group have some specific characteristics that differ from the other group that help them to slow down their functional aging? In the present study, we found that they were healthier (they had significantly less chronic diseases) and they also performed significantly better in the studied cognitive abilities. Surprisingly, these persons were on average chronologically older than the participants in the other group (72 compared with 66 years of age). They were, however, more educated and more literate. More education is, in general, positive for health (Cohen & Syme, 2013) and cognitive abilities (Ritchie, Bates, & Deary, 2015). Levels of hemoglobin and homocysteine were significantly higher in the slow aging group. Hemoglobin is mainly involved in transporting oxygen from the lungs to the rest of the body, which provides energy to the cells in organs and tissues for better functioning. Probably this is somehow related to the biological aging process. Homocysteine is an amino acid that is a by-product in the body’s digestion of protein (Ganguly & Alam, 2015). This amino acid has received increased attention in recent years. High levels can be harmful to the body, as homocysteine is considered a strong and independent risk factor for circulatory diseases (Ganguly & Alam, 2015). The finding that the slow aging group had higher levels is, therefore, interesting. The common recommendation to reduce high homocysteine levels (Ganguly & Alam, 2015) may need further examination.

fBioAge accounted for variance beyond that explained by diagnoses, blood markers and ChronAge in verbal fluency and processing speed in both the slow and fast aging group, and in semantic knowledge and recall in the fast aging group. ChronAge, on the contrary, accounted for additional variance (over and above that explained by diagnoses, blood markers and fBioAge) only in the two cognitive abilities, processing speed and episodic recall in the slow aging group. These results are not in line with our third hypothesis, as we expected no differences in relationships between the age measures and cognition in the two groups. However, this might be an effect of too small sample sizes (see limitations below). This deserves further examination and will also be a natural continuation of the present study.

Compared with ChronAge, the aging indicator fBioAge is more in line with the aging processes. It allows, for example, individual differences in the aging trajectories and for different types of aging trajectories during different periods in life, features that ChronAge does not have. ChronAge is, however, much easier to assess than fBioAge. It demands more work to measure body performance and to construct an fBioAge indicator. Even so, we believe that fBioAge is a promising alternative to ChronAge in both low- and high-income countries, and because of its possibility to indicate individual differences in aging.

This study had several strong features, but there were also methodological limitations. First, the total final sample size included 360 participants. With a larger sample, probably more relationships would have turned out to be significant in the subgroups. However, the age-related cognitive variance would most likely not have been stronger. Dividing the participants into more groups regarding the difference between fBioAge and ChronAge, or gender-separated analyses would have been preferable to get a fuller picture, but as we removed 5% of the participants at each ChronAge end, the resulting groups would have been too small and lacked necessary power. Second, this was a cross-sectional study, which made it impossible to draw any conclusions about causality (see Nilsson, Sternäng, Rönnlund, & Nyberg, 2009; Sternäng, Wahlin, & Nilsson, 2008). A cross-sectional approach explores age differences between individuals, whereas a longitudinal approach can assess age-related changes in cognitive functioning within individuals. These results, therefore, require replication by longitudinal studies, which could more reliably assess whether fBioAge can predict cognitive decline as people age. Finally, different statistical analyses methods have their strengths and weaknesses. In this study, the fBioAge variable was created as a composite of the four included standardized measures. However, a factor analysis of the four included measures (one factor extracted) resulted in very similar results (not shown).

Conclusion

This is the first study on fBioAge and cognition in a low-income country. Compared with ChronAge, fBioAge was a more sensitive predictor of cognition across a broad range of the old adult span. fBioAge seems, based on both the results and the fact that it allows for individual differences in the speed of aging, to have the potential of becoming an important aging indicator in future aging studies.

Footnotes

Authors’ Note

Åke Wahlin passed away in December 2016. He contributed much to the design of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a grant from the Swedish Research Council for research in the Humanities and Social Sciences (Grant 421-2011-1621). The Poverty and Health in Aging (PHA) project was funded by Department for International Development (DfID), UK, and Swedish Agency for Research Cooperation/Swedish International Development Agency, and the Swedish Research Council.

References

Allain

T. J.

Wilson

A. O.

Gomo

Z. A. R.

Mushangi

Senzanje

Adamchak

D. J.

Matenga

J. A.

(1997). Morbidity and disability in elderly Zimbabweans. Age and Ageing, 26, 115-121. doi:10.1093/ageing/26.2.115

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Anstey

K. J.

(2008). Cognitive aging and functional biomarkers: What do we know, and where to from here? In Hofer

Alwin

(Eds.), Handbook of cognitive aging: Interdisciplinary perspectives (pp. 327-339). Thousand Oaks, CA: Sage.

Anstey

K. J.

Christensen

(2000). Education, activity, health, blood pressure and Apolipoprotein E as predictors of cognitive change in old age: A review. The Gerontology, 46, 163-177. doi:10.1159/000022153

Anstey

K. J.

Lord

S. R.

Smith

G. A.

(1996). Measuring human functional age: A review of empirical findings. Experimental Aging Research, 22, 245-266. doi:10.1080/03610739608254010

Anstey

K. J.

Smith

G. A.

(1999). Interrelationships among biological markers of aging, health, activity, acculturation, and cognitive performance in late adulthood. Psychology and Aging, 14, 605-618. doi:10.1037/0882-7974.14.4.605

Bäckman

Jones

Berger

A.-K.

Laukka

E. J.

Small

B. J.

(2005). Cognitive impairment in preclinical Alzheimer’s disease: A meta-analysis. Neuropsychology, 19, 520-531.

Bäckman

Small

B. J.

Fratiglioni

(2004). Cognitive deficits in preclinical Alzheimer’s disease: Current knowledge and future directions. In Dixon

R. A.

Bäckman

Nilsson

L.-G.

(Eds.), New frontiers in cognitive aging (pp. 161-177). Oxford, UK: Oxford University Press.

Bäckman

Small

B. J.

Wahlin

Å.

Larsson

(2000). Cognitive functioning in very old age. In Craik

F. I. M.

Salthouse

T. A.

(Eds.), The handbook of aging and cognition (2nd ed., pp. 499-558). Mahwah, NJ: Lawrence Erlbaum.

10.

Baltes

P. B.

Lindenberger

(1997). Emergence of a powerful connection between sensory and cognitive functions across the adult life span: A new window to the study of cognitive aging? Psychology and Aging, 12, 12-21.

11.

Belsky

D. W.

Caspi

Houts

Cohen

H. J.

Corcoran

D. L.

Danese

. . . Moffitt

T. E.

(2015). Quantification of biological aging in young adults. Proceedings of the National Academy of Sciences of the United States of America, 112, E4104-E4110. doi:10.1073/pnas.1506264112

12.

Birren

J. E.

(1959). Principles of research on aging. In Birren

J. E.

(Ed.), Handbook of aging and the individual: Psychological and biological aspects (pp. 3-42). Chicago, IL: The University of Chicago Press.

13.

Cohen

A. K.

Syme

S. L.

(2013). Education: A missed opportunity for public health intervention. American Journal of Public Health, 103, 997-1001. doi:10.2105/AJPH.2012.300993

14.

Emery

C. F.

Finkel

Pedersen

N. L.

(2012). Pulmonary function as a cause of cognitive aging. Psychological Science, 23, 1024-1032. doi:10.1177/0956797612439422

15.

Ferdous

(2009). Determinants and functional impact of nutritional status among older persons in rural Bangladesh. Uppsala, Sweden: Uppsala University.

16.

Finkel

Sternäng

Wahlin

. (2017). Genetic and environmental influences on longitudinal trajectories of functional biological age. Behavior Genetics, 47, 375-382. doi:10.1007/s10519-017-9851-5

17.

Ganguly

Alam

S. F.

(2015). Role of homocysteine in the development of cardiovascular disease. Nutrition Journal, 14, Article 6. doi:10.1186/1475-2891-14-6

18.

Gunn

D. A.

Rexbye

Griffiths

C. E.

Murray

P. G.

Fereday

Catt

S. D.

. . . Christensen

(2009). Why some women look young for their age. PLoS ONE, 4(12), e8021. doi:10.1371/journal.pone.0008021

19.

Jylhävä

Pedersen

N. L.

Hägg

(2017). Biological age predictors. Ebiomedicine, 21, 29-36. doi:10.1016/j.ebiom.2017.03.046

20.

Kabir

Z. N.

Ferdous

Cederholm

Khanam

M. A.

Streatfied

Wahlin

. (2006). Mini Nutritional Assessment of rural elderly people in Bangladesh: The impact of demographic, socio-economic and health factors. Public Health Nutrition, 9, 968-974. doi:10.1017/PHN2006990

21.

Kramer

A. F.

Bherer

Colcombe

S. J.

Dong

Greenough

W. T.

(2004). Environmental influences on cognitive and brain plasticity during aging. The Journals of Gerontology, Series A: Biological Sciences & Medical Sciences, 59, 940-957.

22.

Krøll

Saxtrup

(2000). On the use of regression analysis for the estimation of human biological age. Biogerontology, 1, 363-368. doi:10.1023/A:1026594602252

23.

Levine

M. E.

(2013). Modeling the rate of senescence: Can estimated biological age predict mortality more accurately than chronological age? The Journals of Gerontology, Series A: Biological Sciences & Medical Sciences, 68, 667-674. doi:10.1093/gerona/gls233

24.

Lindenberger

Baltes

P. B.

(1994). Sensory functioning and intelligence in old age: A strong connection. Psychology and Aging, 9, 339-355.

25.

Lobo

Launer

L. J.

Fratiglioni

Andersen

Di Carlo

Breteler

M. M. B.

, . . . Neurologic Diseases in the Elderly Research Group. (2000). Prevalence of dementia and major subtypes in Europe: A collaborative study of population-based cohorts. Neurology, 54(11, Suppl. 5), S4-S9.

26.

MacDonald

S. W. S.

Dixon

R. A.

Cohen

A. L.

Hazlitt

J. E.

(2004). Biological age and 12-year cognitive change in older adults: Findings from the Victoria Longitudinal Study. The Gerontology, 50, 64-81. doi:10.1159/000075557

27.

McGraw

Winn

Whitaker

(1995). Reliability of the Snellen chart. British Medical Journal, 310, 1481-1482. doi:10.1136/bmj.310.6993.1481

28.

Nilsson

L.-G.

Nyberg

Bäckman

(2002). Genetic variation in memory functioning. Neuroscience & Biobehavioral Reviews, 26, 841-848.

29.

Nilsson

L.-G.

Sternäng

Rönnlund

Nyberg

(2009). Challenging the notion of an early-onset of cognitive decline. Neurobiology of Aging, 30, 521-524. doi:10.1016/j.neurobiolaging.2008.11.013

30.

Palmer

Kabir

Z. N.

Ahmed

Hamadani

J. D.

Cornelius

Kivipelto

Wahlin

. (2014). Prevalence of dementia and factors associated with dementia in rural Bangladesh: Data from a cross-sectional, population-based study. International Psychogeriatrics, 26, 1905-1915. doi:10.1017/S1041610214001392

31.

Raz

Lindenberger

Rodrigue

K. M.

Kennedy

K. M.

Head

Williamson

. . . Acker

J. D.

(2005). Regional brain changes in aging healthy adults: General trends, individual differences and modifiers. Cerebral Cortex, 15, 1676-1689. doi:10.1093/cercor/bhi044

32.

Ritchie

S. J.

Bates

T. C.

Deary

I. J.

(2015). Is education associated with improvements in general cognitive ability, or in specific skills? Developmental Psychology, 51, 573-582. doi:10.1037/a0038981

33.

Rönnlund

Nyberg

Bäckman

Nilsson

L.-G.

(2005). Stability, growth, and decline in adult life span development of declarative memory: Cross-sectional and longitudinal data from a population-based study. Psychology and Aging, 20, 3-18. doi:10.1037/0882-7974.20.1.3

34.

Salthouse

T. A.

(1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403-428. doi:10.1037/0033-295X.103.3.403

35.

Salthouse

T. A.

Kausler

D. H.

Saults

J. S.

(1990). Age, self-assessed health status, and cognition. The Journals of Gerontology, Series B: Psychological Sciences & Social Sciences, 45, 156-160. doi:10.1093/geronj/45.4.P156

36.

Shao

Janse

Visser

Meyer

A. S.

(2014). What do verbal fluency tasks measure? Predictors of verbal fluency performance in older adults. Frontiers in Psychology, 5, Article 772. doi:10.3389/Fpsyg.2014.00772

37.

Spiro

III . (2007). The relevance of a lifespan developmental approach to health. In Aldwin

C. M.

Park

C. L.

Spiro

III (Eds.), Handbook of health psychology and aging (pp. 75-93). New York, NY: Guilford Press.

38.

Sternäng

Lövdén

Kabir

Z. N.

Hamadani

J. D.

Wahlin

. (2016). Different context but similar cognitive structures: Older adults in rural Bangladesh. Journal of Cross-Cultural Gerontology, 31, 143-156. doi:10.1007/s10823-016-9284-2

39.

Sternäng

Reynolds

C. A.

Finkel

Ernsth-Bravell

Pedersen

N. L.

Aslan

A. K. D.

(2016). Grip strength and cognitive abilities: Associations in old age. The Journals of Gerontology, Series B: Psychological Sciences & Social Sciences, 71, 841-848. doi:10.1093/geronb/gbv017

40.

Sternäng

Wahlin

Å.

Nilsson

L.-G.

(2008). Examination of the processing speed account in a population-based longitudinal study with narrow age cohort design. Scandinavian Journal of Psychology, 49, 419-428. doi:10.1111/j.1467-9450.2008.00663.x

41.

United Nations Department of Economic and Social Affairs: Population Division. (2015). World population ageing. New York, NY: United Nations.

42.

Verhaeghen

Salthouse

T. A.

(1997). Meta-analyses of age-cognition relations in adulthood: Estimates of linear and nonlinear age effects and structural models. Psychological Bulletin, 122, 231-249. doi:10.1037/0033-2909.122.3.231

43.

Wahlin

. (2004). Health, disease, and cognitive functioning in old age. In Dixon

R. A.

Bäckman

Nilsson

L.-G.

(Eds.), New frontiers in cognitive aging (pp. 279-302). Oxford, UK: Oxford University Press.

44.

Wahlin

Å.

MacDonald

S. W. S.

deFrias

C. M.

Nilsson

L.-G.

Dixon

R. A

. (2006). How do health and biological age influence chronological age and sex differences in cognitive aging: Moderating, mediating, or both? Psychology and Aging, 21, 318-332. doi:10.1037/0882-7974.21.2.318