Abstract
There are some discrepancies about the superiority of the off-pump coronary artery bypass grafting (CABG) surgery over the conventional cardiopulmonary bypass (on-pump). The aim of this study was estimating risk ratio of mortality in the off-pump coronary bypass compared with the on-pump using a causal model known as collaborative targeted maximum likelihood estimation (C-TMLE). The data of the Tehran Heart Cohort study from 2007 to 2020 was used. A collaborative targeted maximum likelihood estimation and targeted maximum likelihood estimation, and propensity score (PS) adjustment methods were used to estimate causal risk ratio adjusting for the minimum sufficient set of confounders, and the results were compared. Among 24,883 participants (73.6% male), 5566 patients died during an average of 8.2 years of follow-up. The risk ratio estimates (95% confidence intervals) by unadjusted log-binomial regression model, PS adjustment, TMLE, and C-TMLE methods were 0.86 (0.78–0.95), 0.88 (0.80–0.97), 0.88 (0.80–0.97), and 0.87(0.85–0.89), respectively. This study provides evidence for a protective effect of off-pump surgery on mortality risk for up to 8 years in diabetic and non-diabetic patients.
Similar content being viewed by others
Introduction
Atherosclerosis is a progressive chronic inflammatory disease of the vessels, which causes the thickness of the arterial wall to increase1. The myeloid cells destabilize the lipid-rich plaques in the artery wall and cause myocardial infarction2.
Coronary artery bypass grafting (CABG) was introduced in 1968 and rapidly accepted as the standard treatment for patients with atherosclerosis. Percutaneous coronary intervention (PCI) was introduced in 1977 to the medical community as well. Applying these treatments has reduced morbidity and mortality3,4.
CABG has been generally performed with the use of cardiopulmonary bypass (on-pump) machine and operating on a beating heart (off-pump) techniques5. Studies of these two types of CABG and a comparison of their effect on mortality and complications such as myocardial infarction, and stroke revealed no short-term differences between the approaches, while some other studies showed a more promising effect in long-term outcomes for the off-pump technique2,4. In contrast, some studies advocate the no advantages of off-pump CABG for 10-year death or revascularization outcome and manifest lower time to composite endpoint in the off-pump group than in the on-pump group6,7. There are many controversies and debates about the risks and benefits of CABG with cardiopulmonary bypass8. Moreover, the effect of CABG with and without cardiopulmonary bypass on mortality among diabetes and non-diabetes patients need more research9.
Confounding can be a major reason for this ambiguity about off-pump consequences. Conventional methods rely on correct specification of outcome model and can lead to biased and confounded results. There are numerous approaches for addressing this problem e.g., propensity score (PS) methods, g-formula or double-robust methods such as targeted maximum likelihood estimation (TMLE)10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31. PS methods and g-formula rely on exposure and outcome modelling, respectively, and TMLE uses both of these models32. Failure to specify the correct model in PS methods and the bias-variance trade-off issue in TMLE leads to biased estimates for the parameter of interest and high variance when analysing a large number of variables, respectively33,34. Since the estimates from sparse data, regression of exposure given confounders is close to zero or one within one or more strata of covariates, can be biased, double robust estimators are not immune to light violations of the positivity assumption in sparse data. The collaborative targeted maximum-likelihood estimation (C-TMLE) addresses biased estimates from sparse data by applying the nuisance parameter estimation and decreasing mean square error (MSE) in the parameter estimate .35 C-TMLE method does not require that either the estimate of exposure or outcome be consistent, rather it focuses on the reduction of the distance between the estimate of exposure mechanism and its parameter and outcome estimate and its parameter, therefore this approach provides an unbiased and precise estimate (low standard error or narrower confidence interval) of the parameter34. Since C-TMLE produces stable estimates of borderline identifiable parameters and is super-efficient and also because of ambiguities about the effect of off-pump we aimed to use a stronger and more optimized double robust method, C-TMLE, for address a selected- least set of covariates and estimate the effect of the off-pump technique surgery on mortality.
Methods
Study oversight and setting
We examined the effect of off-pump CABG on death by a prospective analysis of Tehran Heart Center’s Surgery data-registry with a median follow-up of 8 years. From 2007 to 2016, patients having undergone isolated CABG were involved in the cohort study by protocol and then were followed in 4, 6, and 12 postoperative months and annually afterward. Follow-ups were done by inpatient visits and face-to-face interviews or telephone contacts for patients who could not attend clinic. The data are stored in the CABG Follow-up databank, which contains classified variables such as demographics, atherosclerosis risk factors, the status of their control, ECG and routine laboratory findings, exercise test results, major events such as acute coronary syndrome, mortality, etc. The Tehran Heart Center (THC)is a major medical treatment facility in Iran that serves both local patients and those referred from other cities. In THC, 28,945 subjects with isolated first-time CABG procedures and with at least 1 graft that has been followed between March 2007 and March 202036.Of these, 4062 patients with incomplete registry data, loss to complete follow-up or CABG with valve surgery were excluded. Finally, 24,883 participants were included in the analysis.
To ensure obscurity, personal identifiers such as name, phone number, and residence address were removed from the raw data, and personal identification number was converted into a serial number. Moreover, this study was approved in terms of the study protocol by the Tehran Heart Center (THC) Review Board and the ethical committee of the Tehran University of Medical Sciences (IR.TUMS.SPH.REC.1402.004). At the time of hospitalization, all patients gave written and informed consent, which included prolonged data collection in the follow-up. Using data in the present study were according to the Helsinki Declaration.
Definition of variables
We defined blood hypertension as systolic blood pressure ≥ 140 mm Hg or diastolic blood pressure ≥ 90 mm Hg or a history of prescribed antihypertensive medications. Additionally, we labeled diabetes mellitus (DM) as any one of fasting plasma sugar ≥ 126 mg/dl; or HbA1c ≥ 6.5%; taking oral hypoglycemic drugs; or injecting insulin; or random plasma glucose ≥ 200 mg/dl, any type of diabetes was considered as diabetes (9941, 39.9%). Ejection fraction was classified as normal (heart failure ≥ 55 in men and ≥ 60 in women), mildly reduced, and reduced (heart failure < 40%)13. Each patient with one of the following disorders was involved in the dyslipidemia group; total cholesterol level ≥ 240 mg/dl; LDL-C level of more than 160 mg/dl; triglyceride serum ≥ 200 mg/dl; and HDL-C of less than 40 mg/dl and less than 50 mg/dl in men and women, respectively; or a history of taking lipid medications based on the National Cholesterol Education Program (NCEP) Adult Treatment Plan (ATP) III37,38. According to the patient's self-response, we classified cigarette smoking status as never smoker and before or current smoker. Moreover, patients were labeled as opium-addicted if they reported current consumption of opium, either inhalation or oral or drinking (with water or tea). The Body Mass Index (BMI) was calculated as the individual's weight in kilograms divided by the square of the height in meters. COPD was diagnosed as incompletely reversible airflow obstruction which was confirmed by spirometry with a ratio of post-bronchodilator forced expiratory volume in one second to forced vital capacity (FEV1/FVC), 0.7, in the lack of bronchiectasis or tuberculosis to otherwise. Additionally, we consider a family history of CAD as acute myocardial infarction or recorded CAD, diagnosed by either computed tomography or invasive coronary angiography, in first-degree relatives.
Exposure
Patients with CABG on the beating heart without cardiopulmonary bypass, and off-pump coronary artery bypass (OPCAB), were defined as the exposure group. In contrast, performed CABG on patients using the on-pump technique, with cardiopulmonary bypass39, was classified in the unexposed group.
Outcome
Our outcome was all deaths following surgery, evaluated as both dichotomous (Death = 0 and Death = 1) and time-to-event endpoints, from the date of surgery up to the last follow-up date. In addition, subjects with any information about the outcome were considered as right censored.
Causal diagram
As shown in Fig. 1, the technique of surgery (Off-pump/on-pump) and death were considered as the main exposure and outcome of the study, respectively. Then the causal directed acyclic graph (cDAG) or causal diagram illustrating the causal relationships between exposure, outcome, and covariates, used for selecting a minimally sufficient set of confounders40,41,42,43,44,45,46,47,48. The presence of an arrow from a particular variable to another variable indicates that there is a causal direct effect. The variables age, COPD, pre or current myocardial infarction, smoking, and consumption of opium were considered as the minimum sufficient set which was included in the adjustment model. Smoking and opium are associated with cardiac function decline. For patients with ventricular dysfunction, off-pump CABG has better outcomes49. The unmeasured variables in our study were indicated as U and directed by a dotted arrow (U1 is lifestyle). Additionally, to better contrast variables were depicted by different colors (Fig. 1).
Analysis
Normally distributed and non-Normally distributed variables were presented as mean (SD) and median (IQR), respectively. Normality was assessed by the Kolmogorov–Smirnov test and Q-Q plot50. Moreover, log-binomial regression model was used for estimating unadjusted/adjusted risk ratio. Independent Student's t-test was performed for comparison of continuous variables with Normal distribution and Mann–Whitney U-test was performed for not Normally distributed. Additionally, the Pearson χ2 test was conducted for categorical variables. Markov Chain Monte Carlo-based method was used to impute the missing data of BMI (80 missing values, 0.30%), smoking status (49 missing values, 0.19%), MI (92 missing values, 0.37%), COPD (80 missing values, 0.30%), and family history (31 missing values, 0.12%). All variables in Table 1 were involved in the imputation process.
In the first step, TMLE in general form (event = 0 and event = 1) models the outcome (death) based on exposure (technique of surgery) and confounders (age, COPD, pre or current myocardial infarction, smoking, and consumption of opium) using super learner (initial estimate). Then, using a regression model of exposure on confounders (super leaner), PS is estimated and a clever covariate, known as H, is generated (step 2). In this model, the technique of surgery (exposure) is regressed on age, COPD, pre or current myocardial infarction, smoking, and consumption of opium. The predicted probability of the fitted model is PS. In the last step, the outcome model is updated (step 1) by refitting step 1 and including H in model; in the updated model, initial estimate is considered as an offset. Finally, risk ratio is estimated using standardized mean in the exposed and un-exposed groups based on the updated outcome model.
C-TMLE is extension of TMLE which attempts to avoid increasing variance by sequentially updating outcome regressions based on PS estimates that incorporate an increasing number of confounders51. The general C-TMLE procedure is based on the following steps;
In the first step, an estimate of outcome model (\({\widehat{Q}}^{0}\)) on exposure (technique of surgery) and confounders (all confounders) is obtained using super learner.
In the second step, a sequence of regression model of exposure (\(\widehat{g}\)) is generated, so that the confounders are included in the exposure model in a stepwise manner based on the cross-validated penalized likelihood of the outcome model .(\({\widehat{g}}^{1}\) : age , \({\widehat{g}}^{2}:age and COPD, {\widehat{g}}^{3}:age,COPD,\) and pre or current myocardial infarction, \({\widehat{g}}^{4}\):\(age, COPD,\) pre or current myocardial infarction, and smoking,\({\widehat{g}}^{5}: age,COPD,\) pre or current myocardial infarction, smoking, and consumption of opium.). The first clever covariate is constructed by adding a confounder, the one which maximizes the penalized likelihood of the outcome model, to the exposure model.(\({\widehat{g}}^{1}\) : age).
In the third step, the outcome regression model, fitted in the first step, is updated based on the first clever covariate, with initial estimate of step 1 as offset. Next, exposure model is updated by including an additional confounder (\({\widehat{g}}^{2}:age and COPD\)). This process continues until the log-likelihood of the outcome model does not increase (until \({\widehat{g}}^{5}\) in our analysis). In the Final step, risk ratio is estimated using standardized mean in the exposed and un-exposed groups based on the last updated outcome model. identifying TMLE (\({\widehat{Q}}^{*}\)), the candidate TMLE in the sequence that minimizes the cross-validated risk and finally estimating effect measure by using outcome mean in exposure and un-exposure groups34.
For comparison, we estimated the PS for each patient using multivariable logistic regression model. To compare models, we calculated Akaike information criterion (AIC). A minimally sufficient set of confounders based on Fig. 1 was included as predictors and the variable surgery technique (off-pump/on-pump) was involved as the response variable. Using the PS and exposure variable in the log-binomial model, we estimated the PS-adjusted effect. Additionally, adjusted Cox proportional hazards model was performed. Standard errors for TMLE and C-TMLE were estimated using influence function, and all effect estimates were reported as risk ratio with 95% CIs52,53,54,55,56,57. All analyses were repeated in diabetes and non-diabetes subgroups. Statistical analysis was performed using R, version 3.5.0 (R Foundation for Statistical Computing, Vienna, Austria).
Results
During the follow-up, with the median (Q1, Q3) of 8 (5.6, 11.1) years, 24,883 patients 18,318 male and 6565 female underwent isolated coronary artery surgery and were included in the Tehran Heart Center’s CABG data registry. The mean (SD) age of the study participants was 65 (10.0) years, ranged from 19 to 95 years (more than 96% of patients were older than 30 years). A total of 1881 patients underwent off-pump surgery 367 (19.5%), and from 22,922 patients with on-pump, 5176 (22.6%) died during the study follow-up. The on-pump group had a longer follow-up time (8.3 (5.8, 11.2) and 6.3(5.0, 7.3), respectively). Moreover, participants with off-pumps were more likely to have a BMI higher than 30 kg/m2 (26.5% vs. 24.0%). On-pump coronary artery bypass graft surgery was associated with having a myocardial infarction, family history of CVD, and dyslipidemia. In contrast, participants who experienced a lower ejection fraction were more likely to have off-pump surgery (Table 1).
Moreover, we assessed the association between study variables and surgery technique in both diabetic and non-diabetic patients. As shown in Table 2, both diabetic and non-diabetic participants who underwent on-pump CABG were more likely to die than the off-pump group. In addition, the high body mass index was associated with taking off-pump surgery in non-diabetic patients. Moreover, Table 2 indicates that patients with diabetes who have hypertension had an increased probability of undergoing off-pump CABG.
To assess the potential violation of the positivity assumption, the distribution of PS for both on-pump and off-pump groups has been depicted in Fig. 2. The median (IQR) for both groups are 0.081(0.059, 0.084) and 0.082 (0.060, 0.084), respectively. The minimum value (0.054) and maximum value (0.1) were almost equal. In general, both distributions are very similar so that PS values were mostly lower than 0.1 and bimodal nearly in 0.06 and 0.08.
The effect estimates of surgery technique on the risk of mortality along with 95% CIs have been estimated by four statistical methods, presented in Table 3: unadjusted, PS-adjusted, TMLE and C-TMLE, for subgroups of diabetes and non-diabetes patients as well as Total sample.
The Cox proportional hazards model shows a higher hazard of mortality for the off-pump technique, with HR = 1.31, 95% CI 1.10–1.55 in diabetes patients, HR = 1.21, 95% CI 1.02–1.43 in non-diabetes patients, and HR = 1.24, 95% CI 1.10–1.40 in the total sample. Proportional hazards assumptions were confirmed using scaled Schoenfeld residuals (The p-values were 0.13, 0.73, and 0.26 in the total, diabetes, and non-diabetes, respectively).
Considering 95% CI, PS-adjusted and TMLE revealed evidence for weak protective effect of off-pump CABG, consistently in diabetic and non-diabetic groups as well Total sample. However, this evidence of effect was stronger based on C-TMLE. As shown in Table 3, double robust methods (TMLE and C-TMLE) exhibit a better fit for the data (lower AIC). Additionally, it is important to note that the most precise effect estimates were provided by C-TMLE.
Discussion
In this large prospective registry-based cohort study, 24,883 patients (1881 off-pump and 22,922 on-pumps) underwent CABG surgery and were followed up for an average of 8.2 years. According to the C-TMLE, there was an evidence of protective effect of the off-pump surgery on mortality: in the Total sample risk was decreased at least about 10%, consistently in diabetic and non-diabetic patients. This evidence was less clear based on PS adjusted and TMLE analyses.
Conventional analysis revealed that off-pump surgery was associated with lower mortality and higher body mass index. These findings were in agreement with some previous researches claiming off-pump CABG is associated with higher survival, including the Oxford database study58 and the Emory Healthcare Hospitals data analysis59. Our study demonstrated that most patients who had previous or current myocardial infarction or a history of CVD were more likely to have on-pump CABG. It should be noted that the choice between off-pump and on-pump seems to be surgeon-dependent60. Our findings were in the same direction with the study that reported off-pump CABG as a superior technique in reducing mortality over the on-pump in patients with diabetes61. Totally, the unadjusted analysis suggested a protective effect which is similar to the findings of previous observational studies62. We assumed that these results may be affected by several potential confounders such as age and smoking that were not controlled, correctly. To overcome the bias induced, the adjusted risk ratio was calculated using PS-adjusted and TMLE, both of which indicated a protective effect for off-pump CABG, Also, C-TMLE provided a stronger evidence regarding a protective effect of off-pump vs on-pump surgery. C-TMLE can address potential biased estimates from light sparse data by applying the nuisance parameter estimation and also, decrease mean square error (MSE) in parameter estimate. Accordingly, the estimated risk ratio by C-TMLE is more confident8,35,63.
On the contrary, the Cox proportional hazard model shows a higher hazard ratio of mortality for off-pump procedures. This could be due to the fact that the hazard ratio of mortality varies over time. By definition, hazard ratio is conditional on survival and so is inherently subject to selection bias if there is an unmeasured risk factor of mortality, even independent of the exposure at the start of follow-up64.
Substantial strengths of our study were applying a precise double-robust method for efficient adjustment of confounders, low bias estimation, large sample size, and low proportion of missing covariates (< 0. 1%). Our study has the following limitations. First, time-to-death data were not assessed. Second, there was a possibility of risk fluctuation as there were some time-dependent covariates65,66. Third, our minimal sufficient confounders were self-reported; specifically opium consumption and cigarette smoking, so there was a possibility of residual confounding due to measurement error67,68. Fourth, the confounding effect of drugs was missed. Last, limitations to the generalizability of the results because of the quality of training for pump staff, type of hospital (governmental, private, and teaching) and availability and cost of equipment required for on-pump surgery may be different among countries.
Comments
This study provides evidence for a protective effect of off-pump on mortality risk for up to 8 years. Further research on survival time using C-TMLE is needed.
Data availability
The data used and analysed during the present study are accessible from the corresponding author in STATA and SPSS format if required.
References
Gaetti-Jardim, E. Jr., Marcelino, S. L., Feitosa, A. C., Romito, G. A. & Avila-Campos, M. J. Quantitative detection of periodontopathic bacteria in atherosclerotic plaques from coronary arteries. J. Med. Microbiol. 58(12), 1568–1575 (2009).
Sun, L., Zhou, M., Ji, Y., Wang, X. & Wang, X. Off-pump versus on-pump coronary artery bypass grafting for octogenarians: A meta-analysis involving 146 372 patients. Clin. Cardiol. 45(4), 331–341 (2022).
Favaloro, R. G. Saphenous vein autograft replacement of severe segmental coronary artery occlusion: Operative technique. Annals Thoracic Surg. 5(4), 334–339 (1968).
Newman, M. F. Neurological outcome research group and the cardiothoracic anesthesiology research endeavors investigators; longitudinal assessment of neurocognitive function after coronary-artery bypass surgery. N. Engl. J. Med. 344(6), 395–402 (2001).
Stroobant, N., Van Nooten, G., Belleghem, Y. V. & Vingerhoets, G. Short-term and long-term neurocognitive outcome in on-pump versus off-pump CABG. Eur. J. Cardio-thoracic Surg. 22(4), 559–564 (2002).
Quin, J. A. et al. Ten-year outcomes of off-pump vs on-pump coronary artery bypass grafting in the Department of Veterans Affairs: A randomized clinical trial. JAMA Surg. 157(4), 303–310 (2022).
Sheikhy, A.F, Forouzannia, K. et al. Off-pump versus on-pump coronary artery bypass graft surgery outcomes in patients with severe left ventricle dysfunction: inverse probability weighted study. BMC Cardiovasc Disord. 22, 488. (2022).
Lamy, A. et al. Five-year outcomes after off-pump or on-pump coronary-artery bypass grafting. N Engl. J. Med. 375(24), 2359–2368 (2016).
Magee, M. J. D. T. et al. Influence of diabetes on mortality and morbidity: Off-pump coronary artery bypass grafting versus coronary artery bypass grafting with cardiopulmonary bypass. Annals Thorac. Surg. 72(3), 776–781 (2001).
Almasi-Hashiani, A., Nedjat, S. & Mansournia, M. A. Causal methods for observational research: A primer. Archives Iranian Med. 21(4), 164–169 (2018).
Almasi-Hashiani, A. et al. The causal effect and impact of reproductive factors on breast cancer using super learner and targeted maximum likelihood estimation: A case-control study in Fars Province Iran. BMC Public Health. 21(1), 1–8 (2021).
Koohi, F., Khalili, D., Soori, H., Nazemipour, M. & Mansournia, M. A. Longitudinal effects of lipid indices on incident cardiovascular diseases adjusting for time-varying confounding using marginal structural models: 25 years follow-up of two US cohort studies. Global Epidemiol. 4, 100075 (2022).
Smith, M. J. et al. Introduction to computational causal inference using reproducible Stata, R, and Python code: A tutorial. Statist. Med. 41(2), 407–432 (2022).
Elduma, A. H. et al. The Targeted Maximum Likelihood estimation to estimate the causal effects of the previous tuberculosis treatment in multidrug-resistant tuberculosis in Sudan. PloS one 18(1), 0279976 (2023).
Mansournia, M.A. et al. Effect of physical activity on functional performance and knee pain in patients with osteoarthritis: Analysis with marginal structural models. Epidemiology. 631–40. (2012).
Mansournia, M. A., Naimi, A. I. & Greenland, S. The implications of using lagged and baseline exposure terms in longitudinal causal and regression models. Am. J. Epidemiol. 188(4), 753–759 (2019).
Mokhayeri, Y., Nazemipour, M., Mansournia, M. A., Naimi, A. I. & Kaufman, J. S. Does weight mediate the effect of smoking on coronary heart disease? Parametric mediational g-formula analysis. Plos One 17(1), e0262403 (2022).
Abdollahpour, I., Nedjat, S., Mansournia, M. A. & Schuster, T. Estimation of the marginal effect of regular drug use on multiple sclerosis in the Iranian population. PloS One 13(4), e0196244 (2018).
Abdollahpour, I., Nedjat, S., Mansournia, M. A., Sahraian, M. A. & Kaufman, J. S. Estimating the marginal causal effect of fish consumption during adolescence on multiple sclerosis: A population-based incident case-control study. Neuroepidemiology 50(3–4), 111–118 (2018).
Khodamoradi, F. et al. The effects of smoking on metabolic syndrome and its components using causal methods in the Iranian population. Int. J. Prev. Med. 12, 118 (2021).
Khodamoradi, F. et al. The effect of smoking on latent hazard classes of metabolic syndrome using latent class causal analysis method in the Iranian population. BMC Public Health. 23(1), 2058 (2023).
Almasi-Hashiani, A., Mansournia, M. A., Rezaeifard, A. & Mohammad, K. Causal effect of donor source on survival of renal transplantation using marginal structural models. Iranian J. Public Health. 47(5), 706 (2018).
Mozafar Saadati, H., Mehrabi, Y., Sabour, S., Mansournia, M. A. & Hashemi Nazari, S. S. Estimating the effects of body mass index and central obesity on stroke in diabetics and non-diabetics using targeted maximum likelihood estimation: Atherosclerosis risk in communities study. Obesity Sci. Practice. 6(6), 628–637 (2020).
Saadati, H. M., Sabour, S., Mansournia, M. A., Mehrabi, Y. & Nazari, S. S. H. The direct effect of body mass index on cardiovascular outcomes among participants without central obesity by targeted maximum likelihood estimation. Arquivos Brasileiros de Cardiologia. 116, 879–886 (2021).
Saadati, H. M., Sabour, S., Mansournia, M. A., Mehrabi, Y. & Nazari, S. S. H. Effect modification of general and central obesity by sex and age on cardiovascular outcomes: Targeted maximum likelihood estimation in the atherosclerosis risk in communities study. Diabetes Metabolic Syndrome Clin. Res. Rev. 15(2), 479–485 (2021).
Aryaie, M. et al. Longitudinal causal effect of modified creatinine index on all-cause mortality in patients with end-stage renal disease: accounting for time-varying confounders using G-estimation. Plos one. 17(8), e0272212 (2022).
Aryaie, M., Sharifi, H., Saber, A., Nazemipour, M. & Mansournia, M. A. Longitudinal causal effects of normalized protein catabolic rate on all-cause mortality in patients with end-stage renal disease: Adjusting for time-varying confounders using the G-estimation method. Am. J. Epidemiol. 190(6), 1133–1141 (2021).
Shakiba, M., Nazemipour, M., Mansournia, N. & Mansournia, M. A. Protective effect of intensive glucose lowering therapy on all-cause mortality, adjusted for treatment switching using G-estimation method, the ACCORD trial. Sci. Rep. 13(1), 5833 (2023).
Mohammadi, N. et al. The marginal causal effect of opium consumption on the upper gastrointestinal cancer death using parametric g-formula: An analysis of 49,946 cases in the golestan cohort study, Iran. Plos One 16(1), e0246004 (2021).
Abdollahpour, I. et al. Estimating the Marginal Causal Effect and Potential Impact of Waterpipe Smoking on Risk of Multiple Sclerosis Using the Targeted Maximum Likelihood Estimation Method: A Large, Population-Based Incident Case-Control Study. Am. J. Epidemiol. 190(7), 1332–1340 (2021).
Mokhayeri, Y. et al. Effects of Hypothetical Interventions on Ischemic Stroke Using Parametric G-Formula. Stroke 50(11), 3286–3288 (2019).
Robins, J. A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect. Math. Model. 7(9–12), 1393–1512 (1986).
Vansteelandt, S. & Keiding, N. Invited commentary: G-computation–lost in translation?. Am. J. Epidemiol. 173(7), 739–742 (2011).
van der Laan, M.J., Rose, S., Sekhon, J.S., Gruber, S., Porter, K.E., & van der Laan, M.J. Propensity-score-based estimators and C-TMLE. Targeted learning: causal inference for observational experimental data. 343–64 (2011).
Gruber, S., van der Laan, M., editors. Collaborative targeted maximum likelihood estimation to assess causal effects in observational studies. Biopharmaceutical Applied Statistics Symposium: Volume 2 Biostatistical Analysis of Clinical Trials; Springer. (2018)
Poorhosseini, H. A. S. The Tehran heart center. Eur. Heart J. 39(29), 2695–2696 (2018).
Rezaianzadeh, A., Jafari, F., Sadeghi, S.E., & Rahimikazerooni, S. The prevalence and predictors of pre-hypertension and hypertension in Kherameh cohort study: A population based study on 10,663 persons in south of Iran. J. Human Hypertens. 35(3):257–64. (2021).
Grundy, S. M. et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: A report of the American college of cardiology/American heart association task force on clinical practice guidelines. Circulation 139(25), e1082–e1143 (2019).
Yusuf, S. et al. Effect of coronary artery bypass graft surgery on survival: overview of 10-year results from randomised trials by the coronary artery bypass graft surgery trialists collaboration. Lancet 344(8922), 563–70 (1994).
Etminan, M., Collins, G. S. & Mansournia, M. A. Using causal diagrams to improve the design and interpretation of medical research. Chest 158(1), S21–S28 (2020).
Kyriacou, D. N., Greenland, P. & Mansournia, M. A. Using causal diagrams for biomedical research. Annals Emergency Med. 81(5), 606–613 (2023).
Etminan, M., Brophy, J. M., Collins, G., Nazemipour, M. & Mansournia, M. A. To adjust or not to adjust: the role of different covariates in cardiovascular observational studies. Am. Heart J. 237, 62–67 (2021).
Etminan, M., Nazemipour, M. & Mansournia, M. A. Potential biases in studies of acid-suppressing drugs and COVID-19 infection. Gastroenterology 160(5), 1443–1446 (2021).
Mansournia, M. A., Nazemipour, M. & Etminan, M. Causal diagrams for immortal time bias. Int. J. Epidemiol. 50(5), 1405–1409 (2021).
Mansournia, M. A., Nazemipour, M. & Etminan, M. Time-fixed vs time-varying causal diagrams for immortal time bias. Int. J. Epidemiol. 51(3), 1030–1031 (2022).
Mansournia, M. A., Nazemipour, M. & Etminan, M. A practical guide to handling competing events in etiologic time-to-event studies. Global Epidemiol. 4, 100080 (2022).
Mansournia, M. A., Nazemipour, M. & Etminan, M. Interaction contrasts and collider bias. Am. J. Epidemiol. 191(10), 1813–1819 (2022).
Soodejani, M. T., Tabatabaei, S. M., Lotfi, M. H., Nazemipour, M. & Mansournia, M. A. Adjustment for collider bias in the hospitalized Covid-19 setting. Global Epidemiol. 6, 100120 (2023).
Sheikhy, A. F. A. et al. Off-pump versus on-pump coronary artery bypass graft surgery outcomes in patients with severe left ventricle dysfunction: Inverse probability weighted study. BMC Cardiovasc. Disorders. 22(1), 488 (2022).
Demir, S. Comparison of normality tests in terms of sample sizes under different skewness and Kurtosis coefficients. Int. J. Assess. Tool. Educ. 9(2), 397–409 (2022).
Lendle, S. D., Fireman, B. V. D. L. & Mark, J. Targeted maximum likelihood estimation in safety analysis. J. Clin. Epidemiol. 66(8), S91–S98 (2013).
Mansournia, M. A. et al. A CHecklist for statistical assessment of medical papers (the CHAMP statement): Explanation and elaboration. British J. Sports Med. 55(18), 1009–1017 (2021).
Mansournia, M. A., Nazemipour, M., Naimi, A. I., Collins, G. S. & Campbell, M. J. Reflection on modern methods: Demystifying robust standard errors for epidemiologists. Int. J. Epidemiol. 50(1), 346–351 (2021).
Mansournia, M. A., Nazemipour, M. & Etminan, M. P-value, compatibility, and S-value. Global Epidemiol. 4, 100085 (2022).
Greenland, S., Mansournia, M.A., & Joffe, M. To curb research misreporting, replace significance and confidence by compatibility: a preventive medicine golden jubilee article. Prevent. Med. 107127. (2022)
Mansournia MA, Nazemipour M. Recommendations for accurate reporting in medical research statistics. Lancet (London, England). 403(10427), 611–2 (2024).
Mansournia, M. A. et al. Checklist for statistical Assessment of Medical Papers: the CHAMP statement. Br. J. Sports Med. 55(18), 1002–1003 (2021).
Raja, S. G. et al. On-pump and off-pump coronary artery bypass grafting for patients needing at least two grafts: comparative outcomes at 20 years. Eur. J. Cardio-Thoracic Surg. 57(3), 512–519 (2019).
Keeling, W. B. et al. Off-pump coronary artery bypass grafting attenuates morbidity and mortality for patients with low and high body mass index. J. Thoracic Cardiovasc. Surg. 146(6), 1442–1448 (2013).
Taggart, D.P. et al. Ten-year outcomes after off-pump versus on-pump coronary artery bypass grafting: Insights from the Arterial Revascularization Trial. J. Thoracic Cardiovasc. Surg. 162(2):591–599e8. (2021).
Renner, A. et al. Coronary revascularization in diabetic patients: Off-pump versus on-pump surgery. Annals Thoracic Surg. 96(2), 528–534 (2013).
Filardo, G. et al. Efficacy and effectiveness of on-versus off-pump coronary artery bypass grafting: A meta-analysis of mortality and survival. J. Thoracic Cardiovasc. Surg. 155(1):172–9. e5. (2018).
Benedetto, U. et al. Off-pump versus on-pump coronary artery bypass grafting: Insights from the Arterial Revascularization Trial. J. Thoracic Cardiovasc. Surgery. 155(4):1545–53.e7. (2018).
Hernán, M. The hazards of hazard ratios. Epidemiology 21(1), 13–15 (2010).
Mansournia, M.A. et al. Handling time varying confounding in observational research. bmj 359. (2017).
Doosti-Irani, A., Mansournia, M. A. & Collins, G. Use of G-methods for handling time-varying confounding in observational research. Lancet Global Health. 7(1), e35 (2019).
Pakzad, R. et al. Effect of Smoking on Breast Cancer by Adjusting for Smoking Misclassification Bias and Confounders Using a Probabilistic Bias Analysis Method. Clin. Epidemiol. 12, 557–568. https://doi.org/10.2147/CLEP.S252025 (2020).
Pakzad, R. et al. Effect of alcohol consumption on breast cancer: Probabilistic bias analysis for adjustment of exposure misclassification bias and confounders. BMC Med. Res. Methodol. 23(1), 157 (2023).
Acknowledgements
We thank Mina Pashang and her colleagues at the Tehran Heart Center for their efforts in collecting data and following up with participants. This study is the result of a doctoral thesis.
Funding
This study was funded by Tehran University of Medical Sciences (TUMS), 9811128003.
Author information
Authors and Affiliations
Contributions
Mansournia MA and Adineh HA were responsible for designing the study, receiving and treating data, analyzing, interpreting results, and drafting the manuscript. Hosseini K and Zareban I participated in data treatment and critical appraisal. Also, Jalali A and Nazemipour M participated in revising.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Adineh, H.A., Hoseini, K., Zareban, I. et al. Comparison of outcomes between off-pump and on-pump coronary artery bypass graft surgery using collaborative targeted maximum likelihood estimation. Sci Rep 14, 11373 (2024). https://doi.org/10.1038/s41598-024-61846-1
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-024-61846-1
- Springer Nature Limited