Mechanisms for the Sex-Specific Effect of H. Pylori on Risk of Gastroesophageal Reflux Disease and Barrett's Esophagus

Barrett's esophagus is a precancerous lesion of esophageal adenocarcinoma. Gastroesophageal reflux disease (GERD) is a major risk factor for both (1, 2). Over the past few decades, the incidence of Barrett's esophagus and esophageal adenocarcinoma has been rising in western populations (3–6), to which the declining prevalence of Helicobacter pylori infection might contribute (7). Colonization by H. pylori can induce a spectrum of changes in gastric physiology, which could lead to either gastric acid hypersecretion or hyposecretion, depending on the location and severity of gastritis (8, 9). In the case where H. pylori infection leads to acid hyposecretion, risk of developing GERD, Barrett's esophagus, and esophageal adenocarcinoma could be reduced because of less acidic reflux in the distal esophagus (10–12).

Meta-analyses of case–control studies of H. pylori seropositivity and GERD or Barrett's esophagus have reported inverse associations [odds ratio (OR), 0.60; 95% confidence interval (CI), 0.47–0.78 (ref. 13) and OR, 0.68; 95% CI, 0.58–0.79 (ref. 10), respectively]. Conversely, a meta-analysis of seven randomised-controlled trials and five cohort studies found inconclusive evidence that eradication of H. pylori was associated with a subsequent increased risk of GERD (14). For Barrett's esophagus, many studies included in the meta-analysis compared H. pylori prevalence in cases with prevalence in endoscopy controls (10). This could overestimate the inverse association between H. pylori and Barrett's esophagus, because controls who had an indication for endoscopy but did not have Barrett's esophagus could have had higher H. pylori prevalence than the general population.

Erosive esophagitis and Barrett's esophagus are nearly twice as common in men than women and esophageal adenocarcinoma is up to nine times more common in men (15, 16). A steeper rise in Barrett's esophagus (3, 4) and esophageal adenocarcinoma (17) incidence over time for men suggests that the decreasing prevalence of H. pylori might have a greater effect on risk for men than for women. Previous case–control studies have also observed possible stronger inverse associations between H. pylori and risk of GERD (18) and Barrett's esophagus (19) for men compared with women.

Despite H. pylori's potential beneficial effect on GERD and Barrett's esophagus, H. pylori infection is a major cause of gastric intestinal metaplasia and non-cardia gastric cancer (20). H. pylori eradication is thus recommended. It is therefore important to understand the mechanisms by which H. pylori infection affects risk of GERD and Barrett's esophagus, and to identify mediators that could be intervened on. To assess the potential effects of H. pylori infection, and possible differences by sex, on the risk of GERD and Barrett's esophagus, we first estimated the ORs between being H. pylori seronegative and having GERD and Barrett's esophagus for men and for women, using a case–control study nested in the Melbourne Collaborative Cohort Study. Then, to investigate the potential for sex differences in H. pylori-induced change in gastric physiology, we compared the prevalence of H. pylori-associated non-atrophic and atrophic (antral, corporal, or multifocal) gastritis in men and women using an algorithm based on serum pepsinogen-I and gastrin-17 data. Finally, we quantified how much of the effect of H. pylori seronegativity on Barrett's esophagus was mediated by GERD using causal mediation analysis (21–23).

The case–control study was nested in the Melbourne Collaborative Cohort Study, which comprises 41,513 participants (59% women) ages 40 to 69 years recruited between 1990 and 1994 (24). Potential participants were predominantly identified via the Victorian Electoral Roll and Melbourne telephone directory (24). People born in Southern Europe were targeted to extend the range of lifestyle exposures. The study protocol was approved by the Human Research Ethics Committee at Cancer Council Victoria (CCV IEC 9001). Written consent to participate was obtained on recruitment.

Participants of age 40 to 59 years at baseline, who had no prior cancer diagnosis, and were not part of a random sub-cohort for an earlier case–cohort study (few of whom had blood samples remaining) were eligible for this case–control study. To minimize selection bias, participants of age >59 years at baseline or born in Greece were excluded because of low participation at the follow-up survey when information on GERD and Barrett's esophagus was collected.

Between 2007 and 2010, information on GERD and Barrett's esophagus was collected via computer-assisted telephone interviews using a systematic questionnaire. Relevant endoscopy and pathology reports and correspondence from gastroenterologists were obtained if participants reported they have been diagnosed with Barrett's esophagus or were unsure if they had. Of the 20,146 eligible participants, 1,054 (5.2%) reported daily heartburn or acid regurgitation with onset at least two years after baseline, 1598 (7.9%) reported symptoms 1 to 6 d/wk, and 10,699 (53%) reported never having had symptoms ≥1 d/wk or having symptoms during pregnancy only (Fig. 1).

For the study of GERD, participants were eligible cases if they reported ever having had daily symptoms. Participants were eligible GERD controls if they reported never having had symptoms ≥1 d/wk or having symptoms during pregnancy only. Participants with intermediate symptoms were excluded from the analysis investigating the association between H. pylori and GERD to minimize misclassification of cases and controls. A sample size calculation was performed a priori to guide the numbers required. For the case–control study of GERD, 600 cases and 600 controls were required to detect an OR of 0.67 or less with 90% power given an estimated prevalence of H. pylori antibodies of 40% (25). The final number of cases was lower (Fig. 1) due to changes in the definition of GERD after the sample was selected; 622 controls were selected.

For the study of Barrett's esophagus, a case was defined as having either endoscopically confirmed columnar-lined esophagus, an esophageal biopsy showing columnar epithelium, or correspondence from a gastroenterologist stating endoscopically diagnosed Barrett's esophagus after baseline. We used endoscopically confirmed columnar epithelium as our Barrett's esophagus case definition to prevent misclassifying intestinal metaplasia identified from biopsy taken from regular or irregular Z line as Barrett's esophagus. Diagnosis of Barrett's esophagus was confirmed by a gastroenterologist (B.J. Kendall). In a sensitivity analysis, we restricted Barrett's esophagus cases to those with intestinal metaplasia. Participants were eligible Barrett's esophagus controls if they had not been diagnosed with Barrett's esophagus. The Barrett's esophagus control group was made up of GERD controls as well as additional participants with GERD symptoms. The inclusion of participants with GERD symptoms in the Barrett's esophagus control group ensures the prevalence and distribution of GERD symptoms in the Barrett's esophagus controls is the same as the eligible cohort. A total of 685 Barrett's esophagus controls were sampled. All Barrett's esophagus cases were included (Fig. 1). No a priori sample size calculation was performed for Barrett's esophagus.

For causal mediation analysis, we defined the mediator GERD as a binary variable [ever having daily symptoms (exposed) or less than daily symptoms (unexposed)].

H. pylori serology, serum pepsinogens, and fasting gastrin-17 were measured in baseline plasma samples stored in liquid nitrogen. Laboratory staff was blinded to the case–control status.

Presence of H. pylori IgG antibodies was assessed using ELISA (Diagnostic Technology) and Western Blot assay (MPD Helico Blot 2.1, MP Biomedicals). An ELISA value greater than 30 U/mL was considered seropositive. For samples with ELISA results around the cutoff value or with extreme values, the Western Blot assay was conducted. According to the manufacturer's instructions, H. pylori seropositivity was defined as either (i) 116kD (CagA) positive, where CagA should be present with one or more of the following bands: 89kD (VacA), 37kD, 35kD, 30kD (UreA) and 19.5kD together, or with current infection marker; (ii) the presence of any one band at 89kD, 37kD or 35kD, with or without current infection marker; or (iii) presence of both 30kD and 19.5kD with or without current infection marker. ELISA (sensitivity 96.4%; specificity 92.7%) and Western Blot (sensitivity 97.4%; specificity 87.7%) assays have demonstrated high test validity (26, 27). The inclusion of current infection marker in the Western Blot assay improved test validity for active infection (28, 29). Samples that were seropositive based on ELISA and CagA criteria but seronegative based on Western Blot criteria were classified as seropositive. Otherwise, where ELISA and Western Blot results were discordant, the results from Western Blot were used for its higher sensitivity.

In the absence of information on biopsy-proven gastritis, we classified gastritis types using the algorithm of Väänänen and colleagues (30), which is based on H. pylori serology, pepsinogens, and gastrin-17. For seropositive individuals, those with gastrin-17 <2.5 pmol/L were classified as either multifocal (corporal/antral) atrophic gastritis (pepsinogen-1 <25 μg/L) or atrophic antral gastritis (pepsinogen-1 ≥25 μg/L); those with gastrin-17 2.5–5 pmol/L were classified as either multifocal atrophic gastritis (pepsinogen-1 <50 μg/L) or non-atrophic gastritis (pepsinogen-1 ≥50 μg/L); and those with gastrin-17 ≥5 pmol/L were classified as either atrophic corporal gastritis (pepsinogen-1 <25 μg/L) or non-atrophic gastritis (pepsinogen-1 ≥25 μg/L). For seronegative individuals, those with pepsinogen <25 μg/L were classified as multifocal atrophic gastritis (gastrin-17 <5 pmol/L) or gastric corporal gastritis (gastrin-17 ≥5 pmol/L); and those with pepsinogen-1 ≥25 μg/L were classified as normal gastric mucosa. Fasting plasma pepsinogens and gastrin-17 were measured using ELISA kits (Biohit).

Sex-specific OR and 95% CI for GERD and Barrett's esophagus comparing H. pylori seronegative with seropositive (reference) participants were estimated using conditional logistic regression with an interaction term between H. pylori and sex. We chose seropositive participants as the reference group for statistical analysis for a more intuitive interpretation of the causal mediation analysis results. We used conditional logistic regression with country of birth (Italy vs. others) as the grouping variable, as H. pylori prevalence was substantially higher for participants born in Italy (Supplementary Table S1), but there were few such participants in the analysis sample (Table 1). The interaction with sex was tested using a likelihood ratio test, comparing models with and without the interaction term. Analysis models included confounders that were identified from the literature and included in a causal diagram (Supplementary Fig. S1 and Supplementary Materials and Methods). These were age, quintiles of an area-based measure of socioeconomic position calculated from census data and based on indicators for disadvantage such as proportion of households or residents with low annual income, low skilled occupations or unemployed [Index of Relative Socioeconomic Disadvantage from the Socio-economic Indexes for Areas (SEIFA; ref. 31)], educational attainment, cigarette smoking, and physical activity level (quartiles of a score based on self-reported combined weekly frequency of vigorous exercise, non-vigorous exercise, and walking over the six months before the interview; the score weights vigorous exercise frequency two times that of non-vigorous exercise and walking).

The Fisher's exact test was performed to assess evidence for difference in gastritis types by sex for H. pylori seropositive individuals. The test was also performed among seronegative individuals as reference. Complete case analysis was performed for gastritis types.

We used a Monte Carlo simulation-based g-computation approach to estimate the role of GERD in mediating the effect of H. pylori seronegativity on Barrett's esophagus risk (21–23). This approach allowed as to also evaluate the confounding effect of adiposity, measured by body mass index (BMI), which was identified as a post-exposure (post H. pylori) confounder for the association between GERD and Barrett's esophagus (23). Participants born in Italy were excluded from mediation analysis due to sparse data. Details on the mediation analysis are available in Supplementary Materials and Methods.

The mediating effect through GERD (i.e., the interventional indirect effect) can be interpreted as the excess Barrett's esophagus risk that could be prevented in a seronegative population by intervening to change the distribution of GERD from its prevalence for seronegative individuals to the prevalence for seropositive individuals, whereas the distribution of BMI remains as it would be for seronegative individuals. The part of the effect not mediated through GERD (i.e., the interventional direct effect) can be interpreted as the excess Barrett's esophagus risk in the seronegative population compared with those seropositive, with the distribution of BMI changing from what it is in the seronegative population compared with those seropositive, whereas the distribution of GERD remains at what it is for seropositive individuals.

For the Barrett's esophagus study, we had missing data for H. pylori, the GERD mediator, and SEIFA (Table 1). Missing data were multiply imputed using chained equations (32). Thirty imputations were performed. The imputation models were stratified by Barrett's esophagus status and sex and included all variables in the analysis models. The interaction terms between exposure and BMI and GERD were generated from multiply imputed variables (33). Analyses were performed within each imputed dataset and results combined using Rubin's rules (34). Multiple imputations were not used for the GERD analysis as <5% of data were missing (35). All analyses were performed using Stata version 15 (36).

The data analyzed in this study were accessed from the Melbourne Collaborative Cohort Study under license and are not publicly available due to consent and ethical restrictions. Codes used for statistical analysis are available upon request from the corresponding author.

Of the 20,146 eligible cohort participants, 14,726 (73%) provided complete GERD data and 15,446 (77%) provided complete Barrett's esophagus data. The final sample included 425 GERD cases with 622 controls and 169 Barrett's esophagus cases with 685 controls (Fig. 1). Age distribution was similar for cases and controls. GERD cases were more likely to be women, more socioeconomically disadvantaged, have lower educational attainment, be current smokers, less physically active, and have higher BMI at baseline than GERD controls. Barrett's esophagus cases were more likely to be men, born outside Italy, less socioeconomically disadvantaged, and have higher BMI at baseline. Barrett's esophagus cases were also substantially more likely to have had daily GERD symptoms at follow-up (Table 1).

There was a statistical interaction between H. pylori seronegativity status and sex on their effect on GERD and Barrett's esophagus (P = 0.011 and P = 0.016, respectively). For men, H. pylori seronegativity was associated with higher risk of GERD (OR, 1.69; 95% CI, 1.03– 2.75) and Barrett's esophagus (OR, 2.28; 95% CI, 1.27–4.12; Table 2). To compare with the conventionally reported ORs (i.e., with seronegativity as reference), H. pylori infection is associated with lower risk of GERD (OR, 0.59; 95% CI, 0.36–0.97) and Barrett's esophagus (OR, 0.45; 95% CI, 0.25–0.81) for men. No associations were observed for women. We therefore performed mediation analysis for men only. When Barrett's esophagus cases were restricted to those with confirmed intestinal metaplasia (n = 112), the association for men was stronger (OR, 3.30; 95% CI, 1.64–6.64). Results from complete-case analysis were similar (Supplementary Table S2).

Fasting serum pepsinogen-1 and gastrin-17 data were available for 107 (77%) seropositive men, 194 (83%) seropositive women, 153 (93%) seronegative men, and 318 (96%) seronegative women. Median pepsinogen-1 and gastrin-17 were higher for seropositive individuals compared with those seronegative (Table 3). Median pepsinogen-1 was higher for seropositive men compared with seropositive women, whereas median gastrin-17 was higher for seropositive women. A similar pattern but smaller difference was observed for seronegative men and women.

On the basis of the gastritis classification algorithm (30), most seropositive individuals had non-atrophic gastritis (n = 115) or atrophic antral gastritis (n = 182), and few had atrophic corporal gastritis (n = 3) or multifocal atrophic gastritis (n = 1). Among those who were seropositive, men had a lower prevalence of non-atrophic gastritis (29% vs. 43%) and a higher prevalence of atrophic antral gastritis (68% vs. 56%) than women (P = 0.015; Table 3). The prevalence of atrophic antral gastritis in seropositive men was 14 per 100 (95% CI, 2–25) higher than that of seropositive women. Among those who were seronegative (n = 471), all had normal gastric mucosa except for two with atrophic corporal gastritis.

For the mediation analysis, the estimated total effect of H. pylori seronegativity on Barrett's esophagus was 15 per 1,000 (95% CI, 3–26), or 2.50-fold (95% CI, 1.02–6.12) higher for seronegative compared with seropositive men (Table 4). 32% of this effect was mediated through GERD: The estimated indirect effect was 5 (95% CI, −4–14) per 1,000. The remaining 68% was unexplained: The estimated direct effect not mediated through GERD was 10 (95% CI, 0–20) per 1,000. Effects estimated from analysis models that included all interaction terms (H. pylori-BMI, H. pylori-GERD, and BMI-GERD interaction in the Barrett's esophagus model) were similar, with wider confidence intervals (Supplementary Table S3). Effects estimated from complete case analysis were similar (Supplementary Table S4).

Men who were H. pylori seronegative had higher GERD and Barrett's esophagus risk than those testing positive, whereas no association was observed for women. Prevalence of atrophic antral gastritis was higher in seropositive men than in seropositive women. For men, approximately one third of the excess Barrett's esophagus risk from being seronegative was mediated by GERD.

We explored potential mechanisms of H. pylori infection on GERD and Barrett's esophagus risk using novel approaches, including examining serologically defined types of gastritis in seropositive individuals and quantifying the effect mediated by GERD in seronegative individuals, as potential explanations for differences in associations by sex. The nested case–control study design ensured the plasma samples for H. pylori and biomarker assays were collected before GERD and Barrett's esophagus cases were diagnosed, minimizing reverse causation. Selection bias was minimized by sampling controls from the same cohort as cases. As opposed to using endoscopy controls, recruitment for the Melbourne Collaborative Cohort Study was unlikely to be influenced by H. pylori serostatus. However, there is potential selection bias from loss to follow-up. For example, we previously reported that the median BMI was 25.5 kg/m² (IQR, 23.2–28.3 kg/m²) for those who provided GERD data and 26.4 kg/m² (IQR, 23.8–29.3 kg/m²) for those lost to follow-up (37). This could have potentially biased the association between H. pylori and risk of GERD and Barrett's esophagus away from the null.

H. pylori serological status was determined using a combination of immunoblotting and enzyme immunoassay. Misclassification of H. pylori status is possible but would be non-differential by outcome status, and thus could have underestimated the true effect of H. pylori on risk of GERD and Barrett's esophagus. Gastritis was sub-classified on the basis of a published algorithm (30) that uses H. pylori serology, serum pepsinogen-1, and gastrin-17 levels. Our lack of endoscopic and histological diagnosis of gastritis and gastric atrophy increased the likelihood of misclassification. The algorithm had previously demonstrated 79% (95% CI, 69%–89%) sensitivity and 91% (95% CI, 88%– 94%) specificity against endoscopic diagnosis (30). However, as the algorithm was validated in Finnish patients with dyspepsia (30), its validity in our study population might be different given the differences in demographic and clinical characteristics. Biomarker measurements were unlikely to be influenced by acid suppressants as the use of these medicines was rare at baseline. Only six men and six women included in the analysis reported use of proton pump inhibitors or H2 antagonists at cohort recruitment. Collection and storage conditions for serum samples have been highly standardized (24). Liquid nitrogen is commonly used for long-term plasma storage in prospective studies (38, 39). Overall, misclassification of gastritis type using H. pylori serology, plasma pepsinogen-1 and gastrin-17 levels due to the lack of endoscopic and histological diagnosis, use of acid suppressant, and degradation of biomarkers is possible, but unlikely to explain the sex difference in gastritis pattern observed.

Consistent with previous studies, H. pylori seronegativity was associated with a similar magnitude of increased risk of GERD and Barrett's esophagus, with a stronger association observed for men than for women (18, 19). One possible explanation for part of the stronger association observed for men is a potential sex difference in gastric physiological change following H. pylori infection. At one end of the spectrum, non-atrophic antral-predominant gastritis can lead to acid hypersecretion and is characteristic of patients who developed duodenal ulcer (8). At the other end of the spectrum, atrophic pangastritis or corpus-predominant gastritis can lead to acid hyposecretion and are characteristic of patients with gastric cancer (9). Most individuals with H. pylori infection exhibit change in gastric physiology and gastric acid secretion that fall somewhere between the two extremes (8). Our results suggest that following H. pylori infection, men were more likely to fall somewhere toward the acid hyposecretion end of the spectrum, with low gastrin-17 and normal or low pepsinogen-1 characteristic of atrophic antral gastritis. Increased severity of gastric antral atrophy has been associated with increased colonization of H. pylori in the corpus mucosa (40, 41), which could lead to impaired secretory function of parietal cells and subsequently higher gastric pH (40, 42). In comparison, women were more likely to fall closer toward the acid hypersecretion end of the spectrum, with normal or high gastrin-17 and pepsinogen-1 characteristic of non-atrophic gastritis (8).

The sex difference in pattern of gastric inflammation and atrophy induced by H. pylori infection has been observed in mice (43). Gastric tissue of male mice responded more rapidly and aggressively following H. pylori infection, characterized by more severe gastritis and higher risk of progression to gastric cancer (43). This is consistent with the higher gastric cancer risk in men observed worldwide (44). H. pylori infection is also more likely to persist into adulthood in males than in females (45–47), which could further contribute to the sex difference in risk. Although there is currently inconclusive evidence that eradication of H. pylori is associated with a subsequent increased risk of GERD, pooled estimates from a meta-analysis of randomised controlled trials and cohort studies suggest increased risk is possible (OR, 1.17; 95% CI, 0.94–1.45; ref. 14).

The mediating role of GERD in the effect of H. pylori on Barrett's esophagus has been previously suggested (48), and supports the proposed mechanism that H. pylori affects GERD and Barrett's esophagus risk by changing gastric physiology and acid secretion (10–12). To our knowledge, this is the first study that has quantified the effect mediated by GERD. We estimated interventional effects, which offer advantages over traditional methods (21, 49). We were able to account for the confounding effect of adiposity by modeling BMI as a post-exposure mediator-outcome confounder. Moreover, estimating interventional effects allowed us to evaluate how much Barrett's esophagus risk could be reduced with a clinical intervention that would reduce occurrence of daily GERD symptoms (22). GERD mediated approximately 32% of the excess Barrett's esophagus risk for being H. pylori seronegative. This can be interpreted as 32% of the excess Barrett's esophagus risk in those seronegative being preventable by an intervention that changed GERD symptom frequency from daily to less than daily. The proportion of indirect interventional effect through GERD could be higher in clinical practice with an intervention that changed symptom frequency from daily to none.

It is possible that some participants with GERD but without symptoms were misclassified as non-cases. This misclassification could lead to underestimation of the true effect of H. pylori on GERD. Moreover, the use of acid suppressant during follow-up by participants with reflux symptoms could have lowered their subsequent Barrett's esophagus risk. In both cases, the indirect effect of H. pylori on Barrett's esophagus mediated by GERD could have been underestimated.

The exclusion of Italian-born participants from the causal mediation analysis due to sparse data could limit our results’ generalizability. A meta-analysis found that the effect of H. pylori on GERD could differ by geographic locations, potentially due to differences in H. pylori prevalence (13). Italian-born participants had much higher H. pylori prevalence than those born in Australia, New Zealand, or Northern Europe (Supplementary Table S1).

The sex difference in H. pylori-induced change in gastric physiology and acid secretion is a potential mechanism for the sex-specific effect of H. pylori infection on GERD and Barrett's esophagus. This proposed mechanism is supported by evidence of mediation by GERD in H. pylori's effect on Barrett's esophagus, and warrants investigation in future studies.

A.M. Hodge reports grants from National Health and Medical Research Council during the conduct of the study. N. Castaño-Rodríguez reports other support from Cancer Institute NSW and UNSW Sydney during the conduct of the study. R.L. Milne reports grants from NHMRC during the conduct of the study. D.R. English reports grants from Australian National Health and Medical Research Council during the conduct of the study. No disclosures were reported by the other authors.

S.E. Wang: Conceptualization, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. S.G. Dashti: Conceptualization, supervision, investigation, methodology, writing–review and editing. A.M. Hodge: Conceptualization, supervision, investigation, writing–review and editing. S.C. Dixon-Suen: Conceptualization, supervision, investigation, writing–review and editing. N. Castaño-Rodríguez: Data curation, investigation, writing–review and editing. R.J.S. Thomas: Conceptualization, funding acquisition, writing–review and editing. G.G. Giles: Conceptualization, funding acquisition, writing–review and editing. R.L. Milne: Conceptualization, funding acquisition, writing–review and editing. A. Boussioutas: Conceptualization, funding acquisition, writing–review and editing. B.J. Kendall: Conceptualization, supervision, investigation, methodology, writing–review and editing. D.R. English: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing.

We thank David Whiteman for assistance with developing the questionnaire. The Melbourne Collaborative Cohort Study cohort recruitment was funded by VicHealth and Cancer Council Victoria. The Melbourne Collaborative Cohort Study was further augmented by Australian National Health and Medical Research Council grants 209057, 396414, and 1074383 and by infrastructure provided by Cancer Council Victoria. Cancer cases and vital status of participants were ascertained through the Victorian Cancer Registry and the Australian Institute of Health and Welfare, including the National Death Index and the Australian Cancer Database. The study of gastroesophageal reflux and Barrett's Esophagus was funded by an NHMRC project grant (504708). S.E. Wang is supported by an Australian Government Research Training Program Scholarship. S.C. Dixon-Suen is supported by an Alfred Deakin Postdoctoral Research Fellowship from Deakin University. N. Castaño-Rodríguez is supported by a Cancer Institute NSW Early Career Fellowship (2019/ECF1082) and a UNSW Scientia Fellowship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.