Study Design and Methods

Question:

Discuss about the Public Health for Esophageal Adenocarcinoma Development.

Terry et al., (2000) briefly presented the research study clearly in an informative manner of the study designing, study findings that help readers to decide whether the paper is suitable for referring and reading purposes.

The issue addressed in the paper is the association between gastroesophageal reflux (GER) and risk for esophageal adenocarcinoma. In people with long-term GER illness, the risk for carcinoma increases by 40 folds in gastric cardia and esophageal cancer. According to Karimi et al., (2014) the incidence of gastric cardia and esophagus adenocarcinoma has risen in Western Europe and United States including Sweden in recent years. Among all reasons, the study highlighted the role of GER in esophageal adenocarcinoma development. There is paucity of data that illustrates association between GER and risk for esophageal adenocarcinoma as a previous study showed weaker associations, however, medical records showed occurrence of GER (Pohl et al., 2013). Moreover, several foods cause temporary reflux symptoms where relaxation of lower esophageal sphincter (LES) takes place. Therefore, the present study was aimed at determining the association between these foods with risk for esophageal adenocarcinoma or gastric cardia.

The participants who were newly diagnosed with gastric cardia (n=313) or esophagus carcinoma (n=216) were eligible for the study. From Swedish population, randomly selection was done for case control subjects after the inclusion or exclusion criteria, 258 and 185 cases of cardia and esophageal adenocarcinoma respectively and 815 controls. All the participants were subjected to one-to-one interviews with interviewers and a structured food frequency questionnaire was used. The questionnaire comprised of 63 beverage and food items that was used for the evaluation of dietary habits 20 years before the interview. It comprised of frequency of consumption, last daily meal, and average meal size of the seven common Swedish meals. Apart from this dietary assessment, the other exposures include reflux tertiles consisting of chocolate and time of last meal, fruit juice and reflux symptoms studied for regurgitation, heartburn or absent at the time of assessment, severity and frequency. The dietary assessment compared the case group with control group and studied how frequently LES relaxing foods gives rise to chronic reflux symptoms among the sufferers as compared to controls.

The outcome of the exposure was that there was no association found between dietary factors that cause LES relaxation and adenocarcinoma risk of esophageal cancer and gastric cardia. This might be possible as sufferers that cause LES relaxation might avoid these foods (Sethi & Richter, 2017). The outcome of exposure showed no association between LES relaxation caused by dietary factors and risk for esophageal malignancy.

Results

The study design is nationwide population-based case control study in Sweden that was studied with 250 and 185 gastric cardia and esophageal malignancy cases respectively and 815 controls. This design is a type of epidemiological observational study where two groups (case and control) groups are identified and comparison is done to identify the factors that contribute to a particular medical condition (Breslow, 2014). In the given article, case group (newly diagnosed adenocarcinoma of esophagus and gastric cardia) and control group were compared for the identification of dietary factors that contribute to risk for gastric cardia and adenocarcinoma of esophagus.

The study population was a nationwide Swedish case-control study of age below 80 years who lived between 1 December 1994 and 31 December 1997. The case group participants for the study comprised of newly diagnosed patients with gastric cardia and esophageal adenocarcinoma. The control group was selected randomly from the Swedish population register that matched gender and age among the cases. Therefore, after the exclusion criteria, 258 and 185 gastric and esophageal adenocarcinoma respectively and controls (815) were taken for the study.

The main finding of the paper was that there was no association between LES-relaxing foods and chronic reflux symptoms, however, it might have happened due to avoidance of these reflux causing foods among the sufferers. Moreover, there was no association found between dietary factors associated with risk for gastric or esophagus cardia. This finding indicates that LES relaxation in dietary factors and transient GER are not associated with risk of gastric and esophageal adenocarcinoma.

There is no such future study implication, although previous study showed eating habits and consumption of foods with relaxed LES or temporary GER were not related with adenocarcinoma risk of gastric cardia and esophagus. However, the paper findings indicated that dietary factors associated with transient GER and LES relaxation are not related with risk of gastric cardia or esophageal malignancy.

Yes, the author has studied the confounding variables in this study that might have an outside influence changing the effect of the independent and dependent variables. The independent variable is the dietary factors and variable of reflux symptoms like heartburn or regurgitation are dependent on it. The results of study conducted by Terry et al., (2000) are likely to be affected by confounding variables. LES-relaxing foods associated with dietary factors were studied in the sample population; however, the subject characteristics suggested that other factors also influenced the study findings. Men were 86% and 83% of the cases and controls respectively. Among the case group, tobacco smoking was higher with gastric cardia or adenocarcinoma of oesophagus as compared to controls that had lowest history of tobacco smoking. In addition, median BMI was higher among gastric cardia or oesophageal adenocarcinoma in case group as compared to control group having low median BMI. Alcohol drinking was also more prevalent among gastric cardia or oesophageal adenocarcinoma patients as compared to controls. This illustrates that these confounding variables are likely to affect the result findings, as dietary factors are not only the variable that increases the risk for adenocarcinoma of oesophagus or gastric cardia.

Discussion

Researchers in their study have mentioned these confounding variables in their study that may have damaged the internal validity of the experiment (Szklo & Nieto, 2014). These third variables were no controlled or eliminated by the researchers in the study. In the study, dietary factors that affect LES-relaxation increase the risk for adenocarcinoma of oesophagus and gastric cardia. In this dietary factors is independent variable and increased adenocarcinoma risk is the dependent variable, however, other variables like alcohol drinking, tobacco smoking and high median BMI are other variables that affected the dependent variable. This confounding variable may have an effect on the risk of adenocarcinoma rather than only dietary factors. The author has explicitly explained the confounding variables in their study explaining the characteristics of subjects in the results.

The results are also affected by bias, as there was potential measurement error in exposure as the questionnaire comprised of questions associated with dietary habits of the participants in last two decades prior to interview (Vandenbroucke et al., 2014). The result findings were weakened by non-differential measurement error of exposure. The dietary factors were only studied in the study that fulfilled the research aim and accordingly, the questionnaire comprised of questions about reflux symptoms two decades before interview that assessed the factor with a possible latency period before cancer occurrence. There is possibly limitation of the data collection due to potential measurement error in the exposure. However, this bias was used for gathering data that is relevant to GER that was useful to detect increased risk with reflux symptoms based on severity, presence and duration (Levy & Lemeshow, 2013).

Looking into the issues and bias witnessed in case-control studies, there is selection, incident, sources of cases and selection of controls. The selection of cases in the study showed that there were more males than females comprised of 83% controls and 86% cases being baseline subject characteristics. Selection bias is a major problem in case control studies that depicts non-comparability between controls and cases. Cases or controls for the study are selected based on some characteristic that is exhibited related to exposure to risk factor (Fithian & Hastie, 2014).  In the given research study, the cases are selected who are unrepresentative of the general population producing cases. However, case control study is designed to select controls that represent population producing cases. Incident case is another bias observed in this case-control study during a particular period as the cases comprised of newly diagnosed patients with 313 and 216 gastric cardia and adenocarcinoma of oesophagus respectively studied between 1 December 1994 and 31 December 1997. The incident cases use is preferential for the researchers as recall of exposure is accurate among the newly diagnosed adenocarcinoma cases for esophagus and gastric cardia (Geneletti et al., 2013). However, it is easy to assess the temporal sequence of intervention or exposure and disease among the newly diagnosed or incident cases.

Conclusion

Sources of cases are another issue observed in this study as the sample recruitment is also biased. In the present study, the sample is recruited from Swedish register being population based. The research study is population based nationwide case control study and it has major limitations, as it is difficult to conduct and is more expensive. The selection of controls is also biased as it is an inherent issue in case-control studies (van Rein et al., 2014). The controls used in this kind of study design are subjected to estimate the prevalence of exposure in population that give rise to cases. Hence, the control group is selected randomly from the general population that gives rise to cases. However, in practical, this condition is not possible, as the control group, selection should also be subjected to same exposure status as cases in absence of exposure and disease association (Sedgwick, 2015). For minimizing bias, controls need to be selected from a representative sample among the population from which cases are produced. The controls must have been selected from Swedish population register only as the cases.

According to Bradford-Hill criteria, the stronger the association between risk factor and outcome, the relationship is likely to become causal (Boniface, Scannell & Marlow, 2017). In the given study, there is no relationship between dietary factors causing LES relaxation and risk for oesophagus adenocarcinoma risk and gastric cardia. Another criterion that can be suggested is that risk occurred before exposure that cannot be plausible suggesting there is no causal association between exposure and outcome (VanderWeele et al., 2016). Moreover, there is no consistency in the findings, as the same findings are not observed among different populations. Previous findings suggested strong relationship between foods that cause LES-relaxation (dietary factors) and risk for oesophageal adenocarcinoma and gastric cardia, however, this study showed no association between dietary factors and adenocarcinoma risk. There is also lack of specificity of association where there is no one to one relationship between cause and outcome.  The temporal sequence of association is also not observed, as exposure did not precede outcome where the sample were already exposed to oesophageal adenocarcinoma and gastric cardia before studying the outcome (Marshall & Galea, 2014). The study did not suggest any biological gradient association as the changes in adenocarcinoma rates did not follow corresponding exposure changes that are dietary factors being dose dependent (Vandenbroucke, Broadbent & Pearce, 2016). The biological plausibility is also not explained in the study, as there was no explanation of any potential biological mechanism. The results showed no association between dietary factors causing LES relaxation associated with gastric cardia or oesophageal adenocarcinoma. Therefore, the findings suggested no association between LES relaxation and risk for gastric cardia or oesophageal malignancy.

Limitations

External validity is another great challenge in case-control studies as it occurs due to systematic error. It is defined as the ability to generalize results to a wider population setting (Allodi & Massacci, 2014). The inferences made in the cause-effect relationships from specific study that are externally valid can be generalized to the wider population (Woolcock, 2013). The results obtained in the present study cannot be generalized to other population as the samples are large obtained from single geographic location (Sweden). This is the reason that the conclusions drawn from the cause-effect relationship cannot be applied to other geographical locations. The food avoidance among the sufferers is a possible consequence that narrowed the exposure range. The reflux-promoting foods like garlic, onion and citrus may have anti-carcinogenic properties negating harmful effects of reflex symptoms. Moreover, the previous studies examined the diet factors and risk for cancers, however, the present study only comprised of few dietary aspects that caused reflux symptoms. This can be one reason that the conclusions drawn from the study cannot be generalized to the wider population. However, the present study results do not contradict the pervious results that there is a strong relationship between severity and frequency of reflux associated with risk of gastric cardia and oesophageal adenocarcinoma. Furthermore, the results suggest that in the general population, reflux symptoms are not a public health issue, rather a consequence of the adenocarcinoma condition.  

The response rates of controls and cases are related to LES-relaxing foods that is called recall bias or differential misclassification that explain negative findings of the study. The hypotheses regarding mint, chocolate, coffee, portion size and time of last daily meal are unknown among the public and that might have elicited differential recall between control and case subjects in the study. Although, the response rates were high as compared to previous findings of dietary factors associated with oesophageal adenocarcinoma, a high degree of differential participation in exposure could have cause bias results away or towards unity.

The discussion section of the paper is explained beautifully depicting results of the study article lacking external validity, as the sample is not representative in the general population. The sample consisted of population from Sweden with self-reported recurrent reflux symptoms before interview and no considerations regarding low mean basal pressure or dyspepsia. The association between reflux producing foods with respect to gastric cardia and oesophageal adenocarcinoma is not addressed properly in the study and this is the reason, it cannot be generalized into other population settings. In addition, the time of last meal before bedtime, average portion size of meal were not related to adenocarcinoma risk of oesophagus and gastric cardia. This issue is witnessed in the nationwide population case study in Sweden that are addressed in the paper. Proper explanations are given for the result findings showing no association with adenocarcinoma risk with good discussion of the strengths of the paper, however, the limitations are not addressed explicitly.

References

Allodi, L., & Massacci, F. (2014). Comparing vulnerability severity and exploits using case-control studies. ACM Transactions on Information and System Security (TISSEC), 17(1), 1.

Boniface, S., Scannell, J. W., & Marlow, S. (2017). Evidence for the effectiveness of minimum pricing of alcohol: a systematic review and assessment using the Bradford Hill criteria for causality. BMJ open, 7(5), e013497.

Breslow, N. E. (2014). Case-control studies. In Handbook of epidemiology (pp. 293-323). Springer New York.

Fithian, W., & Hastie, T. (2014). Local case-control sampling: Efficient subsampling in imbalanced data sets. Annals of statistics, 42(5), 1693.

Geneletti, S., Best, N., Toledano, M. B., Elliott, P., & Richardson, S. (2013). Uncovering selection bias in case–control studies using Bayesian post?stratification. Statistics in medicine, 32(15), 2555-2570.

Karimi, P., Islami, F., Anandasabapathy, S., Freedman, N. D., & Kamangar, F. (2014). Gastric cancer: descriptive epidemiology, risk factors, screening, and prevention. Cancer Epidemiology and Prevention Biomarkers, 23(5), 700-713.

Levy, P. S., & Lemeshow, S. (2013). Sampling of populations: methods and applications. (pp 380-389) John Wiley & Sons.

Marshall, B. D., & Galea, S. (2014). Formalizing the role of agent-based modeling in causal inference and epidemiology. American journal of epidemiology, 181(2), 92-99.

Pohl, H., Wrobel, K., Bojarski, C., Voderholzer, W., Sonnenberg, A., Rösch, T., & Baumgart, D. C. (2013). Risk factors in the development of esophageal adenocarcinoma. The American journal of gastroenterology, 108(2), 200-207.

Sedgwick, P. (2015). Bias in observational study designs: case-control studies. BMJ: British Medical Journal (Online), 350.

Sethi, S., & Richter, J. E. (2017). Diet and gastroesophageal reflux disease: role in pathogenesis and management. Current opinion in gastroenterology, 33(2), 107-111.

Szklo, M., & Nieto, J. (2014). Epidemiology. (pp. 350-358) Jones & Bartlett Publishers.

Terry, P., Lagergren, J., Wolk, A., & Nyrén, O. (2000). Reflux-Inducing Dietary Factors and Risk ofAdenocarcinoma of the Esophagus and Gastric Cardia. Nutrition and cancer, 38(2), 186-191.

van Rein, N., Cannegieter, S. C., Rosendaal, F. R., Reitsma, P. H., & Lijfering, W. M. (2014). Suspected survivor bias in case–control studies: stratify on survival time and use a negative control. Journal of clinical epidemiology, 67(2), 232-235.

Vandenbroucke, J. P., Broadbent, A., & Pearce, N. (2016). Causality and causal inference in epidemiology: the need for a pluralistic approach. International journal of epidemiology, 45(6), 1776-1786.

Vandenbroucke, J. P., von Elm, E., Altman, D. G., Gøtzsche, P. C., Mulrow, C. D., Pocock, S. J., … & STROBE Initiative. (2014). Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. International journal of surgery, 12(12), 1500-1524.

VanderWeele, T. J., Hernán, M. A., Tchetgen Tchetgen, E. J., & Robins, J. M. (2016). Re: Causality and causal inference in epidemiology: the need for a pluralistic approach. International journal of epidemiology, 45(6), 2199-2200.

Woolcock, M. (2013). Using case studies to explore the external validity of ‘complex’development interventions. Evaluation, 19(3), 229-248.