Critical Appraisal Of Case-Control Studies: Validity And Reliability Of Findings

CASP Tool for Case Control Study

Discuss About The Critical Appraisal Check Control Studies?

The current paper seeks to critically analyze the study by Hassan, Bondy, Wolff, Abbruzzese, Vauthey, Pisters, Evans, Khan, Chou, Lenzi, and Li (2007).  By doing so, the validity and usefulness of the findings will be assessed. Since Hassan, et al (2007)’s research was a case-control study, the tool provided by the CASP (Critical Appraisal Skills Programme) will be used (case control study) to guide the appraisal. The paper will also determine the extent to which there is a causal association between pancreatic cancer-which is the main outcome- and consumption of alcohol, smoking of cigarette, pancreatic cancer family history, diabetes mellitus, and history of pancreatitis. Overall, Hassan, et al (2007)’s study is found to be of good quality since its methodological strengths surpass its weaknesses, and it has internal and external validity.

The critical appraisal tool for evaluating case control studies is composed of three wide matters that should be put into consideration, validity of the findings, what the findings are, and whether the findings will help locally (CASP, 2017).  The three issues are then broken down into a total of 11 questions, which will be used to guide evaluation of the chosen article.

The authors addressed a clearly focused issue since the objective of carrying out the research was evident with a clear outline of the population and risk factors being investigated. A case control was appropriate to answer the question since pancreatic cancer is a rare condition. As Song and Chung (2010) outlines, case controls are suited well when investigating rare outcomes. Also, case control studies allow for more than one risk factor to be evaluated for a just one outcome (Song and Chung, 2010). In Hassan, et al (2007)’s study, multiple risk factors including, heavy intake alcohol intake, pancreatic cancer cigarette smoking, and pancreatitis family history,  and diabetes mellitus were all examined for  pancreatic cancer.

The authors used hospital-based case-control. The cases were incident as they were selected from patients that had recently been diagnosed with pancreatic cancer. As CLIO (2004) outlines, using incident cases as opposed to prevalent increases confidence that exposures occurrence were before the outcome disease’s onset. Additionally, using incident cases ensured no cases of over-representation of long duration.  The number of cases selected was significantly large (808 participants) and were varying with age, ethnicity, sex, and social classes, ensuring that they were a representative of a defined population. The exclusion and inclusion criterion was clear ensuring consistency in the characteristics of cases. All the cases had been diagnosed with pancreatic ductal adenocarcinoma, resided in the U.S. and would communicate in English. Patients with other types of cancers were excluded together with those with past history of cancer. As such, the cases were selected in an acceptable way.

Validity of the Results

When recruiting controls, the authors were keen to avoid bias. Controls were chosen from healthy friends as well as genetically unrelated members of the patients’ family with other cancer types other than pancreatic cancer. This would help minimize selection bias by excluding first degree and relatives that are not related by genetics but with pancreatic cancer as controls since the former may have genetic factors related to the outcome while thelatter may share the same lifestyle factors that may predispose them to the disease such as dietary habits. Therefore, by doing so the study would determine the true relationship between pancreatic cancer and some risk factors such as family history of the disease, cigarette smoking, and environmental factors without the link being altered.

The controls were in many ways similar to the cases. The authors used matching where the cases and controls were frequently matched by factors such as age, ethnicity, and sex. As Rose and Laan (2009)  suggests, matching helps to increase the efficiency of the study by allowing similar distributions across confounding variables between case and controls. Although some scholars such as Pearce (2016) argue that matching does not eliminate confounding but instead may introduce it by the matching factors, it is more feasible that matched sampling results to balancing controls and cases across the chosen matching variable levels, thus reducing variance and improving statistical efficiency (Rose and Laan, 2009). As such, Hassan, et al (2007) improved the efficiency of their study through matching.

A study’s validity is also determined by the rate of non-response. According to Groves (2006) high non-response can lead to non-response bias. In Hassan, et al (2007)’s study, the non-response rate was 19.4% with the reasons for failure to participate varying. However, the authors justified that statistical analysis showed that there was no significant differences between missed and selected patients based on sex, race/ethnicity, age, residency state, and educational level. As such, the study was free from non-response bias. The quality of case control studies is contributed by the number of cases and controls selected. In the current study under appraisal, the authors recruited the same number of cases and controls. Selecting equal number increases the efficiency of a study as BMJ (n.d.) states. However, the cases comprised older individuals and had lower level of education compared to controls, an aspect that may have interfered with the study results.

Ascertainment of exposure may also impact on the validity and reliability of a study. Just like most case-control studies, Hassan, et al (2017)’s study established exposure from personal recall by means of self administered questionnaire and structured interviews. As BMJ (n.d.) states, the validity of information from personal recall depend significantly on the subject matter. Therefore, it may have been difficult for some participants to remember their past habits, reducing the reliability of the findings. For instance, one of the risk factors sought to be investigated by Hassan, et al (2007) was dietary habits. Recall bias may have resulted since it may be problematic for individuals to remember their past nutritional habits. Additionally, it is more likely that cases may remember past exposures than controls since they may have figured the potential causes of their conditions. As Carlson and Morrison (2009) states, bias may result if controls and cases recall differently past experiences.

What the Results Are

Confounding factors may affect the findings of a study as they may distort the true association between variables. They may falsely demonstrate an evident or mask an association between a risk factor and an outcome when there is no existence of any relationship (Skelly, Dettroli, Brodt, 2012). Hassan, et al (2007) addressed various confounding such as exposure to tobacco, use of alcohol, and chronic illnesses. Other important confounding factors accounted for by the authors are genetic and lifestyle considerations where they excluded first degree relatives and spouses respectively. These factors may increase the risk of developing pancreatic cancer. Therefore, by the authors recruiting controls that were considered free from exposure to them they decreased the effect of confounding on the study.

Hassan, et al (2007)’s findings revealed that diabetes mellitus, heavy consumption of alcohol, history of pancreatitis, pancreatic cancer family history, and cigarette smoking were significant pancreatic cancer  risk factors. The study also revealed synergistic interactions between family pancreatic cancer family history, cigarette smoking, and type 2 diabetes in women. The results were adjusted for confounding and the associations were most likely insignificantly affected by these factors.

The precision of a research finding is an important determinant of the quality of the study. Precision can be indicated by sample size and study’s efficiency (Carlson and Morrison, 2009). Overall, Hassan, et al (2007)’s study provided precise results since they included balanced groups of non-exposed, exposed, without outcome, and with outcome. Additionally, the adjusted odds ratio had narrow confidence intervals, indicating high precision in estimating the associations.

Hassan, et al (2007) incorporated a significant number of participants in their study. Additionally, participants were composed of people from different geographic areas, ages, and ethnic groups, making it representative. Therefore, the results can be generalized to a more universal population. Therefore, the study has external validity

Conclusion

The study by Hassan, et al (2017) can be deemed of good quality based on the current appraisal. Although it was found to have some weaknesses such as recall bias, its precision and internal and external validity were not compromised. The authors addressed most of the issues that may lower the quality of the study such as confounding factors, selection of cases and controls, addressed a clearly focused question, and chose controls effectively. As such, the study can be applied to the general population and can be used to support evidence-based healthcare coupled with other evidences from other research studies.

References

BMJ. (n.d.). chapter 8: case-control and cross sectional studies. The BMJ. Retrieved from https://www.bmj.com/about-bmj/resources-readers/publications/epidemiology-uninitiated/8-case-control-and-cross-sectional

Carlson, M.D.A., & Morrison, R.S. (2009). Study design, precision, and validity in observational studies. Journal of Palliative Medicine. 12(1), 77-82.

CLIO. (2004). Incident vs. Prevalent cases. CLIO Learning Modules. Retrieved from https://cliomods.stanford.edu/trailmaps/selection/case-control/incident-vs-prevalent-cases/index.html

Critical Appraisal Skills Program, (CASP). (2017). Critical appraisal checklist, case control studies. Retrieved from https://www.casp-uk.net/#!casp-international/c1zsi

Groves, R.M. (2006). Nonresponse rates and nonresponse bias in houseghold surveys. Public Opinion Quarterrly. 70(5), 646-675.

Hassan, M.M., Bondy, M.L., Wolff, R.A., Abbruzzese, J.L., Vauthey, J., Pisters, P.W., Evans, B., Khan, R., Chou, T., Lenzi, R., &Li,  D. (2007). Risk factors for pancreatic cancer: case-control study. Am J Gastroenterol, 102(12), 2696-2707.

Pearce, N. (2016). Analysis of matched case-control studies. BMJ. 352

Rose, S. & Laan, M.J. (2009). Why match? Investigating matched case-control study designs with causal effect estimation. The International Journal of Biostatistics. 5(1)

Skelly, C.A., Dettroli, J.R., & Brodt, D.E. (2012). Assessing bias: the importance of considering confounding. Evidence-Based Spine-Care Journal. 3(1), 9-12.

Song, J.W. & Chung, K.C. (2010). Observational studies: cohort and case-control studies. Plast Reconstr Surg. 126(6), 2234-2242.

Young, J.M. & Solomon, J.M. (2009). How to criticaaly appraise an article. Nature Reviews Gastroenterology and Hepatology. 6 (2), 82-91.