Effect of age, period and birth–cohort on the frequency of glucose-6-phosphate dehydrogenase deficiency in Sardinian adults

Abstract

Background: Glucose-6-phosphate dehydrogenase (G6PD) deficiency is an inherited disorder common in Sardinia. In this study, the frequency variation of G6PD-deficiency across age groups and birth cohorts was investigated using Age–Period–Cohort analysis.

Methods: Data were collected from the clinical records of 11,252 patients (6975 women, age range 17–94 years) who underwent endoscopy between 2000 and 2016 at a teaching hospital (University of Sassari), Italy. G6PD status was assessed by enzymatic assay based on G6PD/6GPD ratio. A Poisson log-linear regression model was used to identify age and time trend in G6PD deficiency.

Results: Enzyme deficiency was detected in 11.4% of the entire cohort (men: 7.9%; women: 13.6%). Age–Period–Cohort analysis showed no inflection points across age groups, especially after age 80. The effects of time period and birth cohorts on G6PD deficiency were negligible (frequencies before and after 1950 were 11.0% and 11.8%, respectively).

Conclusions: These findings indicate that the frequency of G6PD deficiency does not vary significantly in oldest subjects. The lack of evidence for selection across the malaria eradication time may be explained by other factors, including somatic cell selection or misclassification of heterozygotes women as G6PD normal in the older birth cohorts. Additional molecular studies may help clarify these issues.

Key message

• The frequency of glucose-6-phosphate dehydrogenase deficiency is stable across age groups and does not vary in generations born before or after malaria eradication.

Key words:

• Birth cohort effect
• glucose-6-phosphate dehydrogenase
• malaria
• longevity

Introduction

d-glucose-6-phosphate dehydrogenase (G6PD, d-glucose-6-phosphate: NADP oxidoreductase; EC 1.1.1.49) is a “housekeeping” enzyme of the pentose phosphate pathway (PPP) which provides the extra-mitochondrial coenzyme nicotinamide-adenine dinucleotide phosphate (NADPH), as well as the ribose needed to synthesize DNA [1]. NADPH enables cells to neutralize the action of several oxidative agents through the regeneration of reduced glutathione (GSH) from the disulfide form (GSSG). Loss-of-function mutations in the G6PD gene lead to enzyme deficiency which is the most common inherited enzymatic disorder. Affected patients may experience drug- or infection-induced haemolytic anaemia since red blood cell antioxidant defense depends almost exclusively on GSH levels [2]. The G6PD gene is encoded on the X chromosome (Xq28 locus), therefore enzyme deficiency is inherited in an X-linked fashion [3]. Total enzyme deficiency is therefore observed in hemizygous men and homozygous women, whereas in heterozygous women variable levels of residual enzyme activity may be detected, as the result of partial compensation by the normal allele and random inactivation of one X chromosome [4]. Carriers of a G6PD-deficient allele, especially female heterozygotes, have a selective advantage against malaria caused by Plasmodium falciparum [5,6]. For this reason, in countries where malaria was endemic in the past, like the island of Sardinia, G6PD deficiency reaches a frequency ranging between 12 and 24% according to the area [7], with the majority of cases caused by a C→T transition at nucleotide 563 of the G6PD coding region [8].

A number of studies suggest that the frequency of G6PD deficiency varies across the adult lifespan, rising in advanced age owing to its protective effect against age-related diseases [9–12]. Interestingly, the frequency of G6PD deficiency was reported to be greater, nearly twofold, in Sardinian centenarians, especially in hemizygous males [12,13]. Conversely, the frequency of G6PD deficiency was reported to be lower in patients with coronary artery disease [14], a finding that was related to reduced atherogenicity of serum lipoprotein patterns via different metabolic mechanisms [15], to lower levels of pro-inflammatory cytokines released by G6PD-deficient leukocytes [16], and to cholesterol synthesis inhibition within NADPH-depleted cells [17]. Moreover, experimental evidence coming from in vivo and in vitro studies suggests that G6PD deficiency may offer protection against carcinogenesis [18,19]. Indeed, an inverse association between G6PD deficiency and the occurrence of some types of cancer such as colorectal [20] and hepatocellular [21] was recently reported. Nonetheless, the increased frequency of G6PD deficiency observed in previous studies among the oldest people living in Sardinia may just be an artefact created by the cohort effect, since the selective pressure of malaria ceased after it was eradicated in 1950, and in more recent generations G6PD deficiency has not conferred any selective advantage over normal enzyme activity.

In this study, the frequency distribution of G6PD deficiency was investigated in a large cohort of adult subjects from Northern Sardinia spanning several generations including those born before and after malaria eradication.

Patients and methods

Clinical records of patients scheduled for endoscopy at the Department of Internal Medicine, University of Sassari, Italy, from January 2000 to December 2016 were collected. Demographic data including gender, year of birth and age were entered in a digital database. In the case of multiple endoscopic procedures for the same patient within the given time period, only results from the first test were kept in the analysis. An additional group of subjects whose G6PD activity was measured at the local Hospital, were included in the analysis.

G6PD activity was determined in all patients by quantitative spectrophotometric method, based on the ratio between G6PD/6GPD in erythrocytes, as previously reported [22]. A ratio below 0.1 was used to define the presence of G6PD enzyme total deficiency and between 0.1 and 0.85 to define partial deficiency (Nurex, Italy). No molecular analysis was performed in G6PD-deficient patients.

An Institutional Review Board approval was obtained from the local Ethics Committee (Prot N° 2099/CE).

Statistical analysis

All patients were stratified by age and birth cohort in 10-year intervals. The frequency distribution of patients according to gender, age and birth cohort (at the time of endoscopy) was expressed as a percentage. G6PD activity in males was recorded as normal/deficient and in females as normal, partially deficient or totally deficient.

A log-linear Poisson regression model [23] was used to estimate the effects of age, time period and birth cohort on the frequency of G6PD deficiency. The analysis was conducted separately in males and females. Data for age and birth cohort were stratified in 10-year intervals, and for time periods in four intervals from 2000 to 2016. The number n_ijk of subjects cross-classified by i age groups (i = 1,…,8), j time periods (j = 1,…,4) and k birth cohorts (k = 1,…,7) were reported in specific tables. The frequency of G6PD deficiency was obtained by dividing the number of deficient subjects (d_ijk) by the number of patients (n_ijk) allocated to each cell and log-transformed according to the equation below: where f_ijk is the frequency of G6PD deficiency in the i-th age group, in the j-th year and within the k-th birth cohort; d_ijk represents the expected number of G6PD-deficient subjects and is assumed to be distributed as a Poisson variable; n_ijk indicates the population in each cell; μ is the intercept; α_i the effect of the i-th age group; β_j the effect of the j-th time period and γ_k the effect of k-th birth cohort. To avoid the problem of multicollinearity in the regression model, potentially created by the linear relationship between age, birth cohort and period (birth cohort = period − age), only second-order changes were considered (inflection points detectable in the plots of regression estimates) as reported in a previous study [24]. To evaluate the statistical significance of the coefficients, the log–likelihood ratio test was used. Significance level was set at 0.05 (two-sided tests). Analyses were performed using SPSS, version 16.0 (SPSS Inc., Chicago, IL).

Results

A total of 11,252 clinical records were available for the analysis. The proportion of women was 62% (6975). Enzyme deficiency was detected in 11.4% of the entire cohort, and more specifically in 7.9% of men and in 13.6% of women, in line with the frequency previously reported for Northern Sardinia [25]. The distribution of G6PD deficiency across all age categories remained remarkably stable up to the age of 90 to decline slightly thereafter in men (Mantel–Haenszel test of trend: χ²=0.066, p = .797 among men; χ²=0.005, p = .946 among women with total deficiency; χ²=0.031; p = .860 among women with partial deficiency (Table 1). Overall, the proportion of patients with G6PD deficiency in subjects younger and older than 90 was 8.0% and of 6.8% among men, and 13.6% and 12.1% among women, respectively, with no statistically significant differences (p = .780 and .702, Fisher's exact test). The statistical power given by our sample size was 98%, large enough to allow rejection of the null hypothesis of a 4% difference around a G6PD deficiency frequency of roughly 10%.

Table 1. Frequency of glucose-6-phosphate dehydrogenase (G6PD) deficiency by age decade and gender in 11,252 Sardinian patients undergoing endoscopic procedures.

	Males		Females			Total
Age decade	Total M	G6PD deficiency M^a	Total F	G6PD total deficiency F^a	G6PD partial deficiency F^a	Total M + F	G6PD deficiency M + F
<30	375	32 (8.5%)	896	16 (1.8%)	112 (12.5%)	1271	160 (12.6%)
30–39	560	44 (7.9%)	888	23 (2.6%)	98 (11.0%)	1448	165 (11.4%)
40–49	634	52 (8.2%)	1098	21 (1.9%)	119 (10.8%)	1732	192 (11.1%)
50–59	695	53 (7.6%)	1212	35 (2.9%)	145 (11.9%)	1907	233 (12.2%)
60–69	892	69 (7.7%)	1294	30 (2.3%)	152 (11.7%)	2186	251 (11.5%)
70–79	662	54 (8.2%)	941	16 (1.7%)	109 (11.6%)	1603	179 (11.2%)
80–89	415	33 (8.0%)	555	3 (0.5%)	58 (10.4%)	970	94 (9.7%)
≥90	44	3 (6.8%)	91	1 (1.1%)	10 (11.1%)	135	14 (10.4%)
Total	4277	340 (7.9%)	6975	145 (2.1%)	803 (11.5%)	11252	1288 (11.4%)

^a Mantel–Haenszel test of trend: males 0.066; p = .797. Females with total deficiency 0.005; p = .946; females with partial deficiency 0.031; p = .860.

Assuming a deficient allele frequency equal to 7.9% (corresponding to the total proportion of hemizygous males reported in Table 1) the expected frequency of heterozygous females is 2p(1–p) = 14.6%. This value is slightly higher than the observed figure of 11.5% for partially deficient females, a discrepancy that could be attributed to a minimal loss of specificity by the biochemical test.

The effect of birth cohort was also negligible (with an overall frequency of G6PD deficiency before and after 1950 of 11.0% and 11.8%, respectively) (Table 2). In particular, the proportion of females with partial G6PD deficiency, mostly heterozygotes, born before and after 1950 was very similar (11.53% vs 11.49%).

Table 2. Frequency of glucose-6-phosphate dehydrogenase (G6PD) deficiency according to birth cohort and gender in 11,252 Sardinian patients undergoing endoscopic procedures.

	Men		Women			Total
Birthcohort	Total M	G6PD deficiency M^a	Total F	G6PD total deficiency F^a	G6PD partial deficiency^a	Total M + F	G6PD deficiency M + F
<1930	491	45 (9.2%)	684	7 (1.0%)	77 (11.3%)	1175	129 (11.0%)
1930–39	776	65 (8.4%)	1095	12 (1.1%)	120 (11.0%)	1871	197 (10.5%)
1940–49	833	55 (6.6%)	1290	33 (2.6%)	157 (12.2%)	2123	245 (11.5%)
1950–59	699	58 (8.3%)	1175	30 (2.6%)	144 (12.3%)	1874	232 (12.4%)
1960–69	677	50 (7.4%)	1133	29 (2.6%)	108 (9.5%)	1810	187 (10.3%)
1970–79	521	47 (9.0%)	913	20 (2.2%)	110 (12.0%)	1434	177 (12.3%)
≥1980	280	20 (7.1%)	685	14 (2.0%)	87 (12.7%)	965	121 (12.5%)
Total	4277	340 (7.9%)	6975	145 (2.1%)	803 (11.5%)	11252	1288 (11.4%)

^a Mantel–Haenszel test of trend: males 0.167; p = .682. Females with total deficiency 1.195; p = .274; females with partial deficiency 3.160; p = .075.

Age–period–cohort analysis is reported in Figure 1. The regression coefficients for time period (calendar time) and birth cohort were estimated by the Poisson log-linear model. The age curve showed a remarkable steadiness up to advanced age in both males and females (Figure 1, left). Similarly, birth cohort showed no influence on the frequency of G6PD deficiency, as no definite inflection points were detectable in the curves for both males and females (Figure 1, middle). Also the effect of time period was not important, since the estimated prevalence remained basically unchanged throughout the interval 2000–2016 (Figure 1, right).

Figure 1. Variation of G6PD deficiency estimated by Poisson log-linear regression according to age, birth cohort and time period (calendar time) in 11,252 Sardinian patients stratified by gender. The categories chosen for reference are as follows: by age (left) ≥ 90 years; by birth cohort (middle) the generations born after 1980; by time period (right) the interval 2000–2016.

Discussion

Previous studies have suggested that G6PD activity might play a significant role in disease susceptibility, influencing survival fitness in a variety of animal species and also in humans [11]. However, results are controversial, given the remarkable species-dependent differences. Moreover, results obtained in animal models may not always be transferable to humans. In Caenorhabditis elegans, G6PD knockout obtained by RNA interference determines defective oogenesis and embryogenesis [26], whereas overexpression of the enzyme in Drosophila was associated with 30% lifespan extension [27]. In rodents, even a modest increase in G6PD activity has been shown to be associated with increased levels of NADPH, a general strenghtening of antioxidant defence and prolonged healthspan [28]. In the brain of aged mice, glucose-6-phosphate dehydrogenase is needed to protects against endogenous oxidative DNA damage and neurodegeneration [29], a finding consistent with a previous report that enzyme deficiency increases the mutation rate in the brain of mice [30]. In humans, the scenario is much more variegated, as evidence has been provided for both protective and harmful effects. A number of studies seem to mirror the results obtained in animal models, suggesting that G6PD deficiency is associated with increased risk of cardiovascular disease [10] and perhaps neurodegenerative diseases [31]. Enhanced oxidative stress and accelerated cellular senescence have been demonstrated in G6PD-deficient human fibroblasts [32]. Others, including some studies carried out on the Sardinian population, highlighted a protective effect against cardiovascular risk [9,14], as well as cancer [18–21]. Previous investigations in Sardinia, where the frequency of G6PD deficiency was higher in the past owing to endemic malaria, suggested that the frequency of carriers increased in advanced age. Schwartz et al. [13] reported a sharp increase in G6PD deficiency in centenarian men from Sardinia, suggesting that the enzymatic defect may protect elderly subjects from premature mortality, in contrast with our findings. The result reported by this author could have been determined by a cohort effect, since the oldest generations were more exposed to malarial risk, and further G6PD deficient carriers selection [33]. In our study, this hypothesis was tested using an age–period–cohort (APC) model. Our results show that the frequency of G6PD deficiency is constant over the whole age range (17–94 years), in both men and women. In addition, the APC model also excluded any birth cohort effect on the frequency of G6PD deficiency, since it remained substantially stable across the 1920–1980 generations. This rules out a major role of malaria selection since the older generations up until 1950 were exposed to malaria for nearly one-third of their lifetime. As a consequence, our findings seem to indicate that deficiency carriers are neither advantaged nor disadvantaged in terms of survival fitness. This is particularly true for heterozygote women who are the only genotype demonstrably protective against malaria [6]. In this respect, the discrepancy between the results of our investigation and those of the previously mentioned study could be explained, at least in part, by the great geographic variability of G6PD frequency within the island of Sardinia and, in particular, by the existence of an increasing gradient from North to South [34,35]. From this perspective, it cannot be ruled out that the centenarians surveyed in that study came from areas of higher frequency of G6PD deficiency compared with younger controls. The same could be said for epidemiological studies apparently showing a lower prevalence of cardiovascular disease in G6PD-deficient subjects [9,14]. Notably, our study was exempt from this bias, as the population was homogenous and living in the same area. Finally, although evidence obtained from large cohorts indicate a protective effect of G6PD deficiency for some type of cancers, but not for all, this may imply different pathways of carcinogenesis.

Genetic studies on human longevity have shown that the frequency of a list of allelic variants increases with age as they may confer a survival advantage (see for review [36]). Clearly, as our study illustrates, mutations causing G6PD deficiency cannot be included in such a list. How can it be explained, however, that the opposite effect (a significant decrease in G6PD deficiency at extreme ages) was not observed in our study, too? This could be attributed to a number of causes. It has been postulated that in cells other than erythrocytes, the intra-mitochondrial NADPH generation may be sufficient to compensate for the lack of its production in the pentose phosphate pathway [37], making G6PD deficiency in these cells clinically irrelevant. Alternatively, in the elderly a compensatory mechanism increasing the expression of residual enzyme activity is also possible. For example, in women who are heterozygous for G6PD deficiency it has been observed that X chromosome inactivation is not exactly random but tends to shift in favour of the normal cells, as if they were being gradually selected [38]. In any case, even though G6PD deficiency impairs defence against oxidative stress, this would hardly reach a threshold to constitute a significant mortality risk factor in the elderly.

Our study has a number of strengths, including the large age interval of participants (17–94 years) as well as the genetic and socio-cultural homogeneity of the population investigated. Furthermore, the overall frequency of enzyme deficiency was similar to previous studies [25]. On the other hand, two limitations should be taken into consideration: the lack of molecular genotyping and the possible problem of sample representativeness as the patients were referred to a gastrointestinal unit for dyspeptic symptoms. We think anyway that the homogeneity of the study population is great enough to minimize the effect of selection bias and residual confounders.

In conclusion, our results demonstrate that the frequency of G6PD deficiency does not change across age groups and specifically in generations preceding and following malaria eradication in Sardinia. The lack of evidence of selection across the malaria eradication time may be explained by other factors, including somatic cell selection or misclassification of heterozygotes women as G6PD normal in the older birth cohorts. Additional molecular studies may help clarify these issues. The positive association observed recently between G6PD deficiency and a lower risk of malignancies, although an interesting finding in itself, does not significantly affect the overall longevity of the population.

Acknowledgements

We wish to thank Mr. Giamaria Porcheddu and Dr. Alessandra Galistu for the help provided in retrieving G6PD data.

Disclosure statement

The authors report no conflicts of interest.