Expounding nested case control studies in the year of COVID-19: a scoping review and sub group meta-analysis

INTRODUCTION

The pandemic of COVID-19 or SARS-CoV-2.has created enough of fear psychosis and taken a heavy toll on human lives. Worldwide, 172,630,637 cases of COVID-19 by 6 June 2021, which were confirmed, and 3,718,683 deaths, reported to WHO. A total of 1,638,006,899 vaccine doses have been administered by 1^st June 2021.¹ While the route of this outbreak is difficult to foresee, effective response requires immediate action from the view of classic public health strategies to the timely development and implementation of successful counter measures.²

In trying to understand the disease progression, the vast cohort bank and the ever-changing nature of the virus, several preclinical and clinical trials for single drugs or combinations are in passage of advancement.³ It was essential to try and expound various nested case control studies being conducted under the auspices of COVID positivity. Nested case-control study (NCCS) is an amalgamation of design of cohort and case- control. NCCS is mainly a type of Cohort study due to its forward direction and temporality.⁴ This design is done only under two circumstances (a) disease must be new and rare in the population; (b) the diagnostic test must be fairly expensive. Thus, this study design was defined as ‘a small case control study nested within a big cohort study’.

NCCS is different from pure form case- control study where in cases and controls are drawn from the population in a fully catalogued cohort.⁵ The risk factor or exposure of interest is only measured among the cases and controls. NCCS is a more efficacious study as compared to a full cohort design.⁶ In a nested case-control study the cases are drawn from a well-defined catalogued population and the controls are also taken from the same enumerated population. Thus, it is 'nested' in an existing preset same source population, which is COVID-19 positive patients, with known sample size. The term 'cohort' is commonly referred to a group of sub research. There is no time prerequisite but defined by similarity in characteristics.⁷ In nested case-control design, only data of a sample of the full study population needs to be retrieved and analyzed without having to do a new diagnostic study from the beginning. In the field of biomarker research, NCCS sampling has been identified as a useful design option within an enumerated cohort study.⁸ The few advantages of NCCS are (a) problem of recall is eliminated; (b) abnormalities detected represents risk factors than the manifestations of early disease; (c) economical to conduct; and (d) greater comparability between cases and controls.⁹

The likely future emergence of pathogenic zoonotic corona viruses, its adverse economic impact as a pandemic and the and the lack of effective antiviral control plan of action have made it achingly clear that our preparedness to combat corona virus infections is extremely inadequate.¹⁰

METHODOLOGY

Search strategy and selection criteria

The present scoping review and meta-analysis is reported according to the Preferred reporting items such as (PRISMA) guidelines.⁶ We searched Medline for studies published between May 2020-May 2021, with curtailment pertaining to language as only English language articles were included. A thorough search of the PMC literature was done to identify journal articles which had nested case control study as the study design in patients affected with COVID-19. The electronic databases Medline, and Google search, where only free PMC articles were searched for reviews published between 2020 and 2021. We did not consider articles that were referred by other researcher or identified in through cross references in included studies. Eligible studies were only Nested case control studies and pure study designs such as purely randomized control trials, case-control or cohort studies, and case series were not included except for references. We had to exclude review articles, case reports editorials etc.

Data were extracted from eligible articles only: The number of primary studies included in the review included the NCCS study design study population(s), interventions, therapy used, biomarkers and risk factors.

Three investigators (LT, AS, MJ) independently sorted and screened the journal titles and abstracts, and two (LT, MJ) independently sorted and screened full text papers. A third reviewer settled the discrepancies between reviewers at the abstract and full text stage. The use of a sensitive screening methodology at the title or abstract level wherein selection by at least two reviewers was adequate for a study to undertake full text review. In the scoping review all studies under the NCCS label were reviewed and among them those having a measure for hypertension were segregated for meta-analyses.

The keyword search was done using specific terms such as Nested case control study and COVID-19. All those studies which were published and available for free were included and those that were under publication were not included. The inclusion and exclusion criteria allow studies of 2020 to 2021 to be considered. We used the Review manager 5.4 for meta-analyses.

Patient and public involvement

Patients or any public outside of the mentioned researchers were not involved in the evolution of the research question or its progressive development in the actualization of the research, or in the devising of the manuscript at any stage. All authors have taken their study population from COVID confirmed or COVID-19 convalescent population.

RESULTS

The factors studied were 4 for, biomarkers, 2 for therapy and 3 were on risk factors. Table 1 shows study results conclusion and unit of study. The number of population under the predetermined cohort was 2214807 (twenty two lakhs, fourteen thousand and eight hundred and seven. the total sub set of population derived for study in a nested case control study is 10,078 (0.5%) of the cohort population. of the total population studied in the cohort 1128 (12%) were cases and 8950 (88%) were controls.

The statistical guidelines followed were, varied for example Fallerini et al used Lasso Log Regression, Fishers exact and unpaired t test. The association between the presence of TLR7 rare variants and severe COVID-19 was significant. Gnavi et al made use of Conditional log regression risks of CVi associated with drug prescription of ACEIs or ARBs, both were considered together then separately (accounting for switch of medication from one to other), neatly expressed as odds ratios and 95% confidence intervals. Gu et al used Mann Whitney U test, Clopper Pearson and MLR was used by Lewin et al.

The t test by Li et al, and Nemars, Wilcoxin signed rank test by Li et al. Youden index, conditional logistic regression and Bayes theorem was used by Planquette et al. Shi et al and Choi et al used Chi square and multiple logistic regression. In Figure 1, articles selected using PRISMA guidelines from Pubmed PMC articles, Google Scholar and Medline. Records excluded: N=18 not NCCS; N=8 not between 15 months COVID time; N=1 not COVID by d-dimer in DVT. Journal articles included in review with N=9, and records of included studies N=9. Figure 2 depicts sub-group meta-analysis measures hypertension. The four studies included were Gnavi et al, Gu et al, Planquette et al and Shi et al. Three studies were from the year 2020 and one from 2021. Total events in the cases were 197 and 710 in controls. The effect size was calculated for dichotomous variables where in the events were calculated and the OR obtained. The confidence interval is 1.18 (0.54, 2.60). Line of no effect is crossed by two studies Gnavi et al and Gu et al weighted at 30.0% and 27.4 % respectively. Z=0.41 (p=0.68), I2=87%.

Figure 3 is funnel plot showing the variability of individual studies (standard error). It is called a funnel plot because as the study size increases the standard error approaches zero. The funnel plot is symmetrical in its distribution with equal distribution above the mean and below the mean. The scatter within the funnel is evenly distributed and toward the tip of the triangle.

DISCUSSION

Lasso regression is a fairly convenient balancing technique that reduces the model’s complexity.¹¹ Fallerini’s observation reported that screening in badly affected male patients with this specific marker for genetic disease to be started individual interferon treatment.¹² Similarly logistic LASSO regression model for diagnosing carcinoma supported radiologic findings and CDD was adopted by Kim et al.¹³

Conditional logistic regression or (CLR) is a special sort of logistic regression used when cases with specific conditions are each matched with n number of controls without the condition.¹⁴ However the common design that is followed if 1:1 and if 1:N then the N may be between one to five. De et al calculated the OR and 95% confidence interval, adjusted for age sex, cardiovascular conditions and risk factors using conditional logistic regression.¹⁵ No increased risk was observed with either angiotensin-converting enzyme inhibitors) or angiotensin-receptor blockers.

Multivariate logistic regression was used to find expressed ORs at 95%CI for determinants of COVID-19 associated hospitalization and mortality. Choi et al observed similar associations between risk of mortality and drugs used for respiratory diseases such as COPD, asthma etc. Oral cortico-steroids were independently associated with mortality.¹⁶ It was also observed that prolong use of systemic corticosteroids in the COVID pandemic was associated with increased odds of COVID-19 related hospitalization in arthritic disease.¹⁷

Multiple log regression has been employed by Lewin et al where fever or sensation of chills or fever and loss or reduction in taste or smell were independently related to sero-positivity.¹⁸ Shi et al used multivariate logistic analysis, and concluded that advanced age and therefore the higher level of lactase dehydrogenase were independently related to exacerbation in mild COVID-19 patients. Li et al also used MLR to point out that independent risk factors related to severeCOVID-19 were CRP, SpO₂, lymphocytes. Males had indicators showing higher levels in relation to myocardial injury, kidney injury or inflammations as compared in females.¹⁹ Logistic regression models have been used for the different outcomes that were associated with the exposures of interest.^11,20 Mann Whitney U test is used by Gu et al for continuous variables and χ² or Fisher’s exact test for binary variables.²¹

Clopper Pearson interval is a common method for calculating binomial confidence intervals.²² It is by far the foremost covered confidence interval, but it's too conservative especially at extreme p values. Crude seroprevalence rates and their respective 95% Clopper-Pearson CI have been calculated from the proportion of study participants who in the course developed antibodies against SARS-CoV-2.¹⁸ In another retrospective study by Szegdi et al with asymptomatic patients presenting for IVI the study included 279 cases and 319 controls. Of the many valid test done, only one SARS-CoV-2 positive patient was found, leading to a carrier rate of 0.36% with a 95% Clopper-Pearson CI of 0.01-1.99%.²³

T test Fisher exact and chi square were used. Continuous variables have been compared using t tests, and categorical variables were compared using the χ² test or the Fisher exact test. Nemars counts are compared directly for matched cases.²⁴ Categorical variables were expressed in percentages (%). Continuous variables were expressed as median and Inter-quartile range (IQR), even if the information on traditional distribution was not evident Wilcoxon signed-rank test and paired t test were used as appropriately supported whether The Wilcoxon signed-rank test performs better than the paired t-test, and is more powerful in various conditions. It is based upon the sample standard deviation which is affected by outliers.²⁰ This study reveals decreased SpO₂, decreased lymphocyte counts and increased CRP levels, which form likely indicators of severe COVID-19, in patients without comorbidities, advanced age and sex. Males were at higher risk of developing symptoms which were severe, of COVID-19 as compared to females.

Youden index is a abridged measure of the ROC curve. In this study Cases and controls have been matched in 1:2 ratio on the basis of centre of study, age and sex. It measures the effectiveness of a diagnostic marker and optimal threshold value (cut off point) for the marker.²⁵ Fluss et al (calculated as sensitivity+ specificity–1) was earmarked to get the adequate value of D-dimer peak levels. Planquette et al in their alternate calculation of the contingent probability is mentioned as Bayes Rule or Bayes Theorem.²⁶ The Positive and Negative predictive values (PPV and NPV) have been calculated with the Bayes’ theorem with the PE prevalence within the whole cohort (5.6%) and CTPA (21.2%) within the cohort. In another study mortality of patients with confirmed PE on admission to the hospital was not statistically significant as compared to hospital mortality without PE (p=0.09) in whole group as the sample size in the study was small.²⁷

For the sub group meta-analysis a forest plot revealed that the Random effects model was used due to the varying effect sizes crossing the no line of effect randomly. The pooled summary effect crosses the line of no effect numerically thus giving an eye ball effect of no significance at 95%CI. The summary effect is not statistically significant. Study heterogeneity is the measure in which the effect sizes vary in our study as they do not overlap with each other indicating high degree of heterogeneity.²⁸ The funnel plot depicting minimal bias in the studies chosen.²⁹ Though most common co morbidities that have categorically emerged are, hypertension and diabetes.³⁰ The relation of hypertension and COVID-19 patients could not be established in this review.

NCCS design

Fallerini et al studied 156 case who were hospitalised with endo-tracheal intubation and 77 control with no hospitalisation and oligo-symptomatic but the cohort of participants were compulsorily confirmed with RTPCR testing. Their limitation was low sample size and an preposterous conclusion of TLR7 screening for male patients to be started with interferon into the NCCS category.

Gnavi et al studied 316 and 171 cases of CVi, were matched with 1580 and 855 controls categorized for Hypertension and CDD respectively.³¹ They concluded that there was no reason to modify the current antihypertensive therapy in confirmation to recommendations already given by major cardiological societies.

Gu et al 94 deaths were designated as cases and 1:5 proportions of controls 354 who were survivors matched by age and sex. Increased mortality risk in COVID-19 was calculated. They concluded that there was strong evidence of association between existing CHD and mortality risk for patients with COVID-19 infections.

Li et al did a retrospective nested case control studies were in the 67 cases were severe COVID patients and 67 controls were patients with moderate COVID-19. Definitions for cases ad controls have been clearly defined. They concluded that the predictors that potentially emerge are increased CRP levels and decrease in lymphocyte counts and SPO₂ which was independent of co morbid conditions, advanced age or sex.

Plaquette et al from cohort of COVID-19 patients they selected cases with CTPA proven PE N=59 and compared them to two sub group of controls matched for age, sex and center who were excluded or not suspected of PE with group 1- N=118, negative CTPA and group 2- N=118 with un-enhanced CT. They concluded that PE is markedly increased in clinically suspected population and D-Dimer to be an interesting diagnostic tool.

Questionable NCCS design

Lewin et al studied COVID patients in which 173 positive for antibody against 176 who were antibody negative they concluded that seroprevalence was much higher than cumulative COVID-19 incidence was relatively low. The study design was more confirmed to a case control study pattern and not and NCCS design. Li L et al tried to observe the effect of Chinese medicine where in 64 cases were administered CM and 32 were controls with or without CM and also included Western medicine (WM).³² The distinction between case and controls is confusing though the cohort from whom the patients were drawn were COVID-19 positive. They concluded that intervention with Chinese medicine at the time of discharge may promote improvements in lung inflammation and COVID related symptoms

Shi et al observed in a cohort of mild patients with their aim to identify high risk COVID-19 patients going into severe illness. 124 patients, 16 progressed to severe condition and 69 stable mild patients were studied. They fund that high LDH levels and advanced age were independent risk factor for deterioration form mild to severe progression. Even though the inclusion exclusion criteria do have an overlap the study has a prospective approach like in a nested case cohort study design. Index dates of matching of cases with controls have not been mentioned.

Choi et al design had a cohort of COPD in 640 COVID-19 patients as cases and 2560 non-COVID-19 patients for matched controls.In another cohort of asthmatic patients with COVID-19 were 90 and controls were 360 without COVID. They concluded that that ICS does not increase the risk of COVID-19 or its related mortality or respiratory outcomes. The study design is purely case control and does not fit in a NCCS design.

CONCLUSION

The scoping review for Nested Case control studies was carried out for a period of one year May 2020 to May 2021 done worldwide. The research has made an effort to present a unique contribution, in a resource crunched setting, in trying to expound the study design implementations by lending critical viewing, theoretical revision and, new perspectives on use of nested case control studies in field settings. Although the sub group meta-analysis for hypertension done with on random effect due to the high heterogeneity of included studies, the data has been presented in the form of forest plots. Depending on the data type various authors used means for normally distributed data of continuous variables otherwise, the Mann-Whitney test and Chi squared test was used for categorical variables and Fishers exact test when the data was small or limited. The forest plot displaying Odds ratio for events occurring among the cohort, it was found that pooled summary did not show significance of hypertension in COVID-19 patients in the nested case control studies thus deeming it inconclusive. There is a need for larger collection of Nested case control studies to come to a definitive conclusion. Nested case control studies reviewed do not seem to follow the criteria laid by case control studies but seem inclusive of cross-sectional or nested observational in the presence of an umbrella provided by the cohort.

Funding: No funding sources

Conflict of interest: None declared

Ethical approval: Not required

REFERENCES

1. 1. WHO. Coronavirus (COVID-19) Dashboard, 2021. Available at: https://covid19.who.int. Accessed on 10 February 2022.
2. 2. Paules CI, Marston HD, Fauci AS. Coronavirus Infections-More Than Just the Common Cold. JAMA. 2020;323(8):707-8.
3. 3. Krishnan A, Hamilton JP, Alqahtani SA, Woreta T. A narrative review of coronavirus disease 2019 (COVID-19): clinical, epidemiological characteristics, and systemic manifestations. Intern Emerg Med. 2021;16(4):815-30.
4. 4. Jain V. PSM discusses the topic - Nested Case Control Study Design - in English, 2020. Available at: https://medicostimes.com/psm-vivek-jain-pdf/. Accessed on 4^th January 2022.
5. 5. Case-Control Studies. A Nested Case-Control Study, 2021. Available at: https://sphweb.bumc.bu.edu/otlt/mphmodules/ep/ep713_case-control/ep713_case-control3. Accessed on 10 February 2022.
6. 6. Cai T, Zheng Y. Evaluating prognostic accuracy of biomarkers in nested case-control studies. Biostatistics. 2012;13(1):89-100.
7. 7. Biesheuvel CJ, Vergouwe Y, Oudega R, Hoes AW, Grobbee DE, Moons KG. Advantages of the nested case-control design in diagnostic research. BMC Med Res Methodol. 2008;8(1):48.
8. 8. Rundle AG, Vineis P, Ahsan H. Design options for molecular epidemiology research within cohort studies. Cancer Epidemiol Biomarkers Prev. 2005;14(8):1899-907.
9. 9. Shankar R. Nested case control study, 2019. Available at: https://youtu.be/tTPbvCon1BU. Accessed on 4^th January 2022.
10. 10. Wilde AH, Snijder EJ, Kikkert M, Hemert MJ. Host Factors in Coronavirus Replication. In: Tripp RA, Tompkins SM, eds. Roles of Host Gene and Non-coding RNA Expression in Virus Infection. Cham: Springer International Publishing; 2017: 1-42.
11. 11. Appier. 5 Types of Regression Analysis And When To Use Them, 2021. Available at: https://www.appier.com/blog/5-types-of-regression-analysis-and-when-to-use-them/. Accessed on 10 February 2022.
12. 12. Fallerini C, Daga S, Mantovani S, Benetti E, Picchiotti N, Francisci D, et al. Association of Toll-like receptor 7 variants with life-threatening COVID-19 disease in males: findings from a nested case-control study. Elife. 2021;10:67569.
13. 13. Dandoy CE, Kim S, Chen M, Ahn KW, Ardura MI, Brown V, et al. Incidence, Risk Factors, and Outcomes of Patients Who Develop Mucosal Barrier Injury-Laboratory Confirmed Bloodstream Infections in the First 100 Days After Allogeneic Hematopoietic Stem Cell Transplant. JAMA Netw Open. 2020;3(1):1918668.
14. 14. NCSS. Conditional Logistic Regression, 2021. Available at: https://www.ncss.com/. Accessed on 10 February 2022.
15. 15. Abajo FJ, Martín S, Lerma V, Abril G, Aguilar M, Luque A, et al. Use of renin-angiotensin-aldosterone system inhibitors and risk of COVID-19 requiring admission to hospital: a case-population study. Lancet. 2020 May 30;395(10238):1705-14.
16. 16. Choi JC, Jung SY, Yoon UA, You SH, Kim MS, Baek MS, et al. Inhaled Corticosteroids and COVID-19 Risk and Mortality: A Nationwide Cohort Study. J Clin Med. 2020 Oct 23;9(11):3406.
17. 17. Shi J, Li Y, Zhou X, Zhang Q, Ye X, Wu Z, etal. Lactate dehydrogenase and susceptibility to deterioration of mild COVID-19 patients: a multicenter nested case-control study. BMC Med. 2020;18(1):168.
18. 18. Lewin A, Therrien R, Serres G, Grégoire Y, Perreault J, Drouin M, et al. SARS-CoV-2 seroprevalence among blood donors in Québec, and analysis of symptoms associated with seropositivity: a nested case-control study. Can J Public Health. 2021;112(4):576-86.
19. 19. Li X, Marmar T, Xu Q, Tu J, Yin Y, Tao Q, et al. Predictive indicators of severe COVID-19 independent of comorbidities and advanced age: a nested case-control study. Epidemiol Infect. 2020;148:255.
20. 20. Details for Non-Parametric Alternatives in Case C-Q, 2021. Available at: https://bolt.mph.ufl.edu/6050-6052/unit-4b/module-13/details-for-non-parametricalternatives/. Accessed on 10 February 20.
21. 21. Gu T, Chu Q, Yu Z, Fa B, Li A, Xu L, et al. History of coronary heart disease increased the mortality rate of patients with COVID-19: a nested case-control study. BMJ Open. 2020;10(9):e038976.
22. 22. Clopper C, Pearson E. The Use Of Confidence Or Fiducial Limits Illustrated In The Case Of The Binomial. Biometrika. 1934;26(4):404-13.
23. 23. Szegedi S, Huf W, Miháltz K, Marlovits PV. Prevalence of SARS-CoV-2 infection in patients presenting for intravitreal injection. Spektrum Augenheilkd. 2021;35(2):70-4.
24. 24. The Analysis Factor. The Difference Between a Chi-Square Test and a McNemar Test, 2014. Available at: https://www.theanalysisfactor.com/differencebetween-chi-square-test-and-mcnemar-test. Accessed on 10 February 2022.
25. 25. Fluss R, Faraggi D, Reiser B. Estimation of the Youden Index and its associated cutoff point. Biom J. 2005;47(4):458-72.
26. 26. Planquette B, Berre A, Khider L, Yannoutsos A, Gendron N, Torcy M, et al. Prevalence and characteristics of pulmonary embolism in 1042 COVID-19 patients with respiratory symptoms: A nested case-control study. Thromb Res. 2021;197:94-9.
27. 27. Brownlee J. A Gentle Introduction to Bayes Theorem for Machine Learning, 2019. Available at: https://machinelearningmastery.com/bayes-theorem-for-machine-learning. Accessed on 10 February 2022.
28. 28. Top Tip Bio. Forest Plot Interpretation - Clearly Explained, 2020. Available at: https://youtu.be/Pxs0gl3hRKE. Accessed on 10 February 2022.
29. 29. Dalby A. The Funnel Plot, 2020. Available at: https://youtu.be/Pxs0gl3hRKE. Accessed on 10 February 2022.
30. 30. Kridin K, Schonmann Y, Bitan D, Damiani G, Weinstein O, Cohen AD. The Burden of Coronavirus Disease 2019 and Its Complications in Patients With Atopic Dermatitis-A Nested Case-Control Study. Dermatitis. 2021;32(1):45-52.
31. 31. Gnavi R, Demaria M, Picariello R, Dalmasso M, Ricceri F, Costa G. Therapy with agents acting on the renin-angiotensin system and risk of severe acute respiratory syndrome coronavirus 2 infection. Clin Infect Dis. 2020;71(16):2291-3.
32. 32. Li L, Gou CY, Li XM, Song WY, Wang XJ, Li HY, et al. Effects of Chinese Medicine on Symptoms, Syndrome Evolution, and Lung Inflammation Absorption in COVID-19 Convalescent Patients during 84-Day Follow-up after Hospital Discharge: A Prospective Cohort and Nested Case-Control Study. Chin J Integr Med. 2021;27(4):245-51.