SciELO - Scientific Electronic Library Online

 
vol.69 issue3A comparison of viscoelastic behavior of engineering elastomers under different stress and temperatureEffect of transverse magnetic field on dose distributions of yttrium 90 skin patch source author indexsubject indexsearch form
Home Pagealphabetic serial listing  

Services on Demand

Journal

Article

Indicators

Related links

  • Have no similar articlesSimilars in SciELO

Share


Revista mexicana de física

Print version ISSN 0035-001X

Rev. mex. fis. vol.69 n.3 México May./Jun. 2023  Epub Sep 06, 2024

https://doi.org/10.31349/revmexfis.69.031101 

Medical physics

Analysis of factors that affect radiation dose level during interventional cardiology procedures using logistic regression

Edrees M. Harkia  b 

Z. S. Rashida 

a Physics Department, Education College, Salaheddin university-Erbil/ Kurdistan Region- Iraq. E-mail: edrees.nury@su.edu.krd

b Research center-Erbil Polytechnic University Kurdistan reign -Iraq.


Abstract

Interventional cardiology procedures (ICP) are considered some of the main medical procedures in which patients are exposed to high doses of radiation. The aim of this study was to examine how to control the level of radiation exposure and to analyze and study the factors affecting the increase in radiation exposure from the specified level using a regression method. The results correctly predicted that in 80.0 %, 90.5 %, and 95.2% of the cases, there were routine dose area product (DAP) levels, and in 64.3%, 33.3%, and 77.8% of cases, there were high levels of DAP, giving an overall percentage, correct prediction rate of 72.45%, 73.35%, and 90.0%, for coronary angiography (CA), percutaneous coronary intervention (PCI), and combined CA with PCI (CA/PCI), respectively. All the factors studied in this research, namely voltage (kV), current (mA). Fluoroscopy Time (FT) and Body Mass Index (BMI), have a significant relationship with the DAP level. We concluded that regression analysis is a reliable method for evaluating the user protocol in a center or hospital and identifying variables that have an effect on the dose area product level in interventional cardiology.

Keywords: Interventional cardiology; fluoroscopy; logistic regression

1. Introduction

Interventional cardiology procedures (ICP) are some of the major medical examination methods applied for the detection of cardiovascular diseases under fluoroscopic X-ray guidance to obtain images of the heart chambers, valves, and surrounding blood vessels [1]. The most frequently reported cardiac procedures by interventional fluoroscopy are coronary angiography (CA), percutaneous coronary intervention (PCI), and combined CA with PCI (CA/PCI) [2]. Due to the use of X-rays in interventional cardiac procedures, it is considered one of the main medical procedures in which patients are exposed to high doses of radiation. X-rays are ionizing radiation and pose a significant risk, with the main radiation-induced side effects being being skin injury (deterministic effect or tissue reaction) and increased cancer risk (stochastic effect) [3]. Therefore, more preventive measures and studies are needed to reduce the radiation dose. The radiation dose of the patient during interventional cardiology is influenced by three types of factors. First, the technical factors affecting the radiation dose (X-ray beam quality, X-ray geometry, X-ray beam limitation devices, and fluoroscopic and acquisition imaging dose rate settings). Second are procedure-related factors, which include the increase in the treatment of complex lesions, such as chronic total occlusions, because of improvements in techniques and PCI equipment. The third is the group of factors that are patient-related (body mass index (BMI), comorbidities, and seriousness of coronary artery disease) [4,5].

In general, for radiation protection and development of quality assurance programs, reference levels (RLs) were introduced by the International Commission on Radiological Protection. Establishing RLs for interventional cardiology is challenging because there are many factors influencing these procedures that lead to a wide dose distribution [6]. In interventional cardiology (IC), several research studies have focused on the dose area product for the establishment of reference levels and dose optimization [7,8]. However, literature data reveal that the most commonly studied parameters for cardiac interventions are BMI, fluoroscopy time (FT), peak skin dose, and dose area product (DAP) for each procedure [2-10].

The aim of this study was to examine how to control the level of radiation exposure, to analyze and study the factors affecting the increase in radiation exposure from the specified level, and to estimate the incidence of high radiation dose procedures using a logistic regression method.

2. Methods and materials

2.1. Logistic Regression

Logistic regression is a reliable method of identifying which variables have an impact on a topic of interest. The process of performing a regression allows one to confidently determine which factors matter most and which can be ignored. Binary logistic regression is used to estimate the association of one or more independent (predictor) variables with a binary dependent (outcome) variable. A binary (or dichotomous) variable is a categorical variable that can only take two different values or levels. The model usually has two types of objectives: predictive or explanatory. In a model with predictive objectives, we aim to establish a parsimonious model, i.e., a model involving the least number of variables that best explains the dependent variable. In the case of a model with explanatory objectives, we aim to study the causal relationship between a ‘cause’ variable and an ‘effect’ variable. Given a set of values of the independent variables, we wish to estimate the probability that the event of interest will occur and evaluate the influence each independent variable has upon the response in the form of an odds ratio (OR). The form for predicted probabilities is expressed as a natural logarithm (ln) of the odds ratio [11].

lnP(Y)1-P(Y)=β0+β1X1+β2X2++βkXk, (1)

P(Y)1-P(Y)=eβ0+β1X1+β2X2++βkXk, (2)

P(Y)=eβ0+β1X1+β2X2++βkXk1+eβ0+β1X1+β2X2++βkXk, (3)

where, lnP(Y)/(1-P(Y)) is the log (odds) of the outcomes, Y is the dichotomous outcome; X1,X2,.,Xk are the predictor variables, β1,β2,.,βk are the regression (model) coefficients, and β 0 is the intercept. In Eq. 3, the logistic regression model directly relates the probability of Y to the predictor variables. When an independent variable X k increases by one unit (X k + 1), with all other factors remaining constant, the odds of the dependent variable increase by a factor exp (β k ), which is called the OR and ranges from zero (0) to positive infinity. It indicates the relative amount by which the odds of the dependent variable increase (OR > 1) or decrease (OR < 1) when the value of the corresponding independent variable increases by one (1) unit [12]. The goodness-of-fit for the Logistic Regression (LR) model can be assessed in several ways. First, the overall model (rela-tionship between all of the independent variables and depen-dent variable) is assessed. Second, the significance of each of the independent variables also needs to be assessed. Third, the predictive accuracy or discriminating ability of the model need to be evaluated, and finally, the model needs to be vali- dated [13].

2.2. Data collection

The data from CA, PCI, and combined CA and PCI (CA/PCI) performed from 1 January to 31 August 2021 were collected from the Erbil Heart Center in the Kurdistan region in Iraq. For each procedure, the following data were collected: patient characteristics (age, sex, weight, and length to calculate BMI), exposure factors kV, mAs, and FT, and dosimetry indicators. Clinical data and technical factors were gathered from 29 coronary angiography (CA), 30 percutaneous transluminal intervention (PCI), and 30 double set-up (CA/PCI) procedures; all performed using the femoral approach. The data were gathered using a stratified random sampling method. This center has 10 cardiologists, 10 nurses, and 10 radiology technicians, with three active to angiography systems angiography rooms. In room 1, a GE Innova 2100 C-arm fluoroscope system 1316440G2283 model is set up, and room 2 is geared with Philips C-arm fluoroscope system 722064 185 and 105935 181 models.

2.3. Statistical analysis

Model construction: A binary logistic regression model (BLRM), a statistical approach to predict the presence of a DAP based on the available variables (Kv, mA, FT, and BMI), has been successfully used to predict the presence of a DAP level. It is known that DAP is related to the risk of exposure to radiation, which is widely used in the establishment of RL. DAP is the binary outcome variable used in the analysis. High DAP levels are assigned the value of 1, and routine DAP levels are assigned the value of 0. The BLRM has the following form:

Y=ln odds=βKv  Kv+βmA  mA+βFT  FT+βBMI  BMI (4)

In Eq. (4), the variable Y is the log (natural) of the odds of the event under consideration. In our case, the event will be the occurrence of a high DAP procedure. The βs are the coefficients of the regression calculated by the model of predictor variables kV, mA, FT, and BMI. The regression method was chosen with a free intercept. The justification for this is that the Automatic exposure control (AEC) compensates by keeping the quantity of radiation. The DAP values for 89 patients were dichotomized into three groups, which are CA, PCI, and CA/PCI, for each group divided into two subgroups; for CA, the DAP ≤ 35 Gy.cm2 and > 35 Gy.cm2, for PCI, the DAP ≤ 85 Gy.cm2, and > 85 Gy.cm2 for CA/PCI, the DAP ≤ 130 Gy.cm2 and > 130 Gy.cm2, respectively. The first subgroup is considered the routine radiation dose procedure, and the second is considered the high radiation dose procedure. The choice of level of DAP was based on the European Society of Vascular Interventional Radiology [7,10].

3. Results

Table I shows a summary of the 89 patients’ data in three groups of CA, PCI, and CA/PCI and their associated radiation dose metric with exposure factors. Figure 1 shows the scatter plots matrix, showing DAP as a function of kV, mA, FT, and BMI, respectively. Figure 2 demonstrates the scatter plots of DAP for CA, PCI, CA/PCI with the level of DRLs of the European Society of Interventional Cardiology. Figure 3 boxplots explain the four predicting variables kV, mA, FT, and BMI distribution for the two dependent variable categories of DAP: DAP > 35 Gy.cm2 and DAP ≤ 35 Gy.cm2 for CA, DAP > 85 Gy.cm2 and DAP ≤ 85 Gy.cm2 for PCI, and DAP > 130 Gy.cm2 and DAP ≤ 130 Gy.cm2 for CA/PCI.

Table I Sample size and mean, SD, Max. and Min. for patient characteristics for three group CA, PCI, and CA/PCI, and their associated radiation dose metric with exposure factors. 

Sample size BMI (kg/m2) kV mA FT DAP (Gy.cm2)
CA 29 Mean 26.741 80.482 13.975 3.500 44.274
S.D 3.746 17.795 3.362 2.341 38.534
Max. 33.306 120 18.7 12 148
Min. 15.241 53 5.9 1.08 4.510
PCI 30 Mean 28.038 78.900 15.926 10.182 69.779
S.D 3.417 16.649 3.049 5.931 30.257
Max. 34.131 120 19.5 24.800 148
Min. 15.241 45 8.6 2 4.510
CA/PCI 30 Mean 28.646 82.466 14.4 8.181 86.224
S.D 3.589 18.574 3.268 4.815 59.837
Max. 35.651 120 18.7 19.400 241.410
Min. 23.437 54 7 1.100 28.100

Figure 1 Scatter plots matrix showing DAP as function of Kv, mA, FT, and BMI, respectively, for CA, PCI, and CA/PCI. 

Figure 2 Scatter plots matrix showing DAP Level, the green is for the procedures with DAP >35Gy cm2 and DAP ≤35 Gy cm2. The blue is for the procedures with DAP >85 Gy cm2 and DAP≤80Gy.cm2. The yellow is for the procedures with DAP >130 Gy cm2 and DAP ≤130 Gy.cm2

Figure 3 Boxplots showing the four predicting variables kV, mA, FT and BMI distribution for the two dependent variable categories of DAP; DPA >35Gy cm2 and DAP ≤35 Gy cm2 for CA,DPA >85 Gy cm2 and DAP≤85Gy for PCI, and DPA >130 Gy cm2 and DAP ≤35 Gy.cm2

The binary regression analysis was conducted to investigate whether kV, mA, FT, and BMI factors predict DAP levels. The Hosmer-Lemeshow goodness-of-fit was not significant (P > 0.05), indicating that the model was correctly specified Table II. The model correctly predicted 80.0%, 90.5%, and 95.2% of cases where there were routine DAP levels and 64.3%, 33.3%, and 77.8% of cases where there was a high level of DAP, giving an overall percentage correct prediction rate of 72.45%, 73.35%, and 90.0% for CA, PCI, and CA/PCI, respectively. The model’s results showed that independent variables (kV, mA, FT, and BMI) were found to be significant for all three groups: CA, PCI, and CA/PCI. As shown in Table III, the obtained LRM with four predictors (variables) is given by Eq. (5) below:

for CA

Y=lnodds=0.020  kV-0.50  mA-0.125FT-0.011BMI

for PCI

Y=lnodds=0.016kV-0.211mA+0.084  FT+0.007  BMI

for CA/PCI

Y=lnodds=0.0109  kV-0.021  mA+0.989  FT-0.188  BMI. (5)

Table II Classification table for CA, PCI, and CA/PCI.  

Observed Predicted
DAP.Classes Percentage Correct
≤ 35 > 35
CA DAP.Classes ≤ 35 14 2 87.5
> 35 8 5 38.5
Overall Percentage 65.5
DAP.Classes
≤ 85 > 85
PCI DAP.Classes ≤ 85 20 2 90.9
> 85 5 3 37.5
Overall Percentage 76.7
DAP.Classes
≤ 130 > 130
CA/PCI DAP Classes ≤ 130 20 1 95.2
> 130 2 7 77.8
Overall Percentage 90.0

Table III Results of the logistic regression analysis for all the variables that may be related to the occurrence of high dose area product for CA, PCI, and CA/PCI procedures. 

β S.E. Wald Exp(β) 95% C.I. for EXP(β)
Lower Upper
CA
Step 1 a Kv 0.020 0.024 0.501 1.021 0.962 1.061
FT 0.500 0.296 2.859 1.649 0.966 4.303
mA -0.125 0.110 1.297 0.882 0.686 1.077
BMI (kg/m2) -0.011 0.098 0.012 0.989 0.789 1.167
PCI
Step 1 a Kv 0.016 0.026 0.358 1.016 0.963 1.060
FT 0.084 0.073 1.328 1.088 0.935 1.227
mA -0.211 0.132 2.533 0.810 0.672 1.078
BMI (kg/m2) 0.007 0.106 0.004 1.007 0.832 1.219
CA/PCI
Step 1 a Kv 0.0109 0.081 0.638 1.011 0.799 1.099
FT 0.030 0.028 0.511 1.030 0.593 1.791
mA -0.021 0.087 0.327 0.979 1.035 6.978
BMI (kg/m2) -0.188 0.217 0.751 0.829 0.542 1.268

a. Variable(s) entered on step 1: Kv, mA, FT, BMI (kg/m2).

When a logistic regression is calculated, as in Eq. (5) and Table III for CA, the regression coefficients (βKv=  0.02,    βmA= -0.125,    βFT=0.500, and βBMI=-0.011) are the estimated increase in the log odds of the DAP per unit increase in the value of the kV, and FT; also the estimated increase in the log odds of the DAP per unit, decrease in the value of the mA and BMI. And we can say that the increase of one unit of kV, mA, FT, and BMI will be reflected in the DAP increase by 0.2%, -12.5%, 50.0%, and -1.1%, respectively. In other words, the exponential function of the regression coefficient (eβKv=1.021,  eβmA=  0.882,  eβFT=  1.649, and eβBMI=  0.989) are the odds ratios associated with a one-unit increase in the Kv, mA, FT, and BMI, respectively. However, for PCI, the regression coefficients (βKv=  0.016,    βmA=  -0.211,βFT=0.084 and βBMI=0.007) are the estimated increase in the log odds of the DAP per unit increase in the value of the kV, FT, and BMI, and were decreased in the value of the mA. Furthermore, we can say that an increase of one unit of kV, mA, FT, and BMI will be reflected in the DAP increase by 1.6%, -21.1%, 8.4%, and 0.7%, respectively.

In other words, the exponential function of the regression coefficient (eβKv=1.016,eβmA=  0.810,  eβFT=    1.088, and eβBMI=  1.007), are the odds ratios associated with a one unit increase in the kV, mA, FT, and BMI, respectively. Finally, for CA/PCI, the regression coefficients βKv=  0.0109,    βmA=  -0.021,  βFT=0.030, and βBMI=-0.188) are the estimated increases in the log odds of the DAP per unit increases in the value of the kV, mA, FT, and BMI, respectively. We can say that the increase in one unit of kV, mA, FT, and BMI will be reflected in the DAP increase by 1.09%, -2.1%, 3.0%, and -18.8%, respectively. In other words, the exponential function of the regression coefficient (eβKv=1.011,  eβmA=0.979,  eβFT=1.030, and eβBMI=0.829) are the odds ratios associated with a one unit increase in the kV, mA, FT, and BMI, respectively.

4. Discussion

The scatter plots matrix in Fig. 1; showing DAP as a function of kV, mA, FT, and BMI, respectively. We note that the scatter plot for kV and FT are strong, positive association because as kV and FT increases, so the DAP increased, but we note that the scatter plot for mA and BMI are strong, negative association, because in general, as mA and BMI increase, their DAP decreases. According to the results of the binary regression model, all the factors studied in this research, namely kV, mA, FT, and BMI, have a significant relationship with the DAP levels and confidently determine this result, which agrees with those published by others [2,4,5,11]. However, when reviewing the various studies on patient dosimetry, there appears to be a considerable variation in the radiation doses received by patients, as shown in Table IV. As there are a number of possible explanations, it is necessary to establish such a model to explain the radiation doses received by patients during interventional cardiology procedures and to detect the reason why the patient received a high dose and higher than the permissible dose level. When coefficient β of the variable is positive, we obtain OR > 1, and it therefore corresponds to a risk factor. If the value β is negative, OR will be < 1, and the variable therefore corresponds to a protective factor [12,13]. According to this, we concluded that for CA, the odds of (patients  that  DAP35/patients  that  DAP>35) increase if kV and FT increase by one unit and decrease if mA and BMI decrease by one unit. This means that the number of patients with DAP > 35 Gy.cm2 can be reduced when kV and FT decrease and mA increases. However, we concluded for PCI that the odds of (patients  that  DAP85/patients  that  DAP>85) increases if kV and FT increased by one unit, and decreased if mA and BMI decreased by one unit. This means that the number of patients with DAP > 85 Gy.cm2 can be reduced when kV and FT decrease and mA increase. For CA/PCI, the odds of (patients  that  DAP130/patients  that  DAP>130) increases if kV and FT increased by one unit, and decreased if mA and BMI decreased by one unit. It means that the number of patients with DAP > 130 Gy.cm2 can be reduced when kV and FT decrease and mA increases. Finally, we can conclude that increased kV and FT are the two factors that help to increase the risk, and according to the model, decreased mA is a factor that helps in radiation protection. From this, we concluded that we can identify variables that have an effect on the DAP level in interventional cardiology.

Table IV DAP (Gy.cm2) comparisons with values published in the literature. 

DAP Gy.cm2 Ref.
CA PCI CA/PCA
39.9 78.3 109.3 [2]
87 91 [6]
35 85 130 [8]
83 193 199 [14]
45 86 96 [15]
43.72 38.77 [16]
44.274 69.779 86.224 This study

5. Conclusion

We can conclude that according to the model, increased kV and FT are the two factors that increase the risk, while decreased mA is a factor that helps in radiation protection. From this, we can identify variables that have an effect on the DAP level in interventional cardiology. Also, we state that regression analysis is a reliable method for evaluating user protocols in a center or hospital. We can also identify the most critical factors and the factors that can be disregarded. Thus, we conclude that the regression analysis method can be used in quality assurance and driving diagnostic reference levels and dose optimization.

Acknowledgments

The authors would like to thank Sallahddin University for their support in the current work.And the authors give thanks to the Surgeries Hospital Specialist - Heart Center and express sincere thanks to Prof. Dr. Wasfi Kahwachi, Tishk International University for encouragement and helpful suggestions during this research work.

References

1. J. Jankowski, et al., System for Personal Dosimetry in Interventional Radiology, Radiation Protection Dosimetry 101 (2002) 221, https://doi.org/10.1093/oxfordjournals.rpd.a005971. [ Links ]

2. I. Pantos, et al., Patient radiation doses in interventional cardiology procedures, Current cardiology reviews 5 (2009) 1, https://doi.org/10.2174/157340309787048059. [ Links ]

3. C. Cousins, et al., ICRP PUBLICATION 120: Radiological protection in cardiology, Annals of the ICRP 42 (2013) 1, https://doi.org/10.1016/j.icrp.2012.09.001. [ Links ]

4. K. A. Fetterly, et al., Clinical determinants of radiation dose in percutaneous coronary interventional procedures: influence of patient size, procedure complexity, and performing physician, JACC: Cardiovascular Interventions 4 (2011) 336, https://doi.org/10.1016/j.jcin.2010.10.014. [ Links ]

5. J.-L. George, et al., Time-course reduction in patient exposure to radiation from coronary interventional procedures: the Greater Paris Area Percutaneous Coronary Intervention Registry, Circulation: Cardiovascular Interventions 10 (2017) e005268, https://doi.org/10.1161/CIRCINTERVENTIONS.117.005268. [ Links ]

6. E. Samara, et al., An audit of diagnostic reference levels in interventional cardiology and radiology: are there differences between academic and non-academic centres?, Radiation protection dosimetry 148 (2012) 74, https://doi.org/10.1093/rpd/ncq600. [ Links ]

7. T. Siiskonen, et al., Establishing the European diagnostic reference levels for interventional cardiology, Physica Medica 54 (2018) 42, https://doi.org/10.1016/j.ejmp.2018.09.012. [ Links ]

8. D. Alves da Silva, et al., Overview of occupational radiation exposure in medical workers in Santa Catarina, Brazil, between 2014 and 2017, Radiation protection dosimetry 183 (2019) 355, https://doi.org/10.1093/rpd/ncy125. [ Links ]

9. A. Shah, et al., Radiation dose during coronary angiogram: relation to body mass index, Heart, Lung and Circulation 24 (2015) 21, https://doi.org/10.1016/j.hlc.2014.05.018. [ Links ]

10. A. Sulieman, et al., Occupational and patient exposure in coronary angiography procedures, Radiation Physics and Chemistry 104 (2014) 68, https://doi.org/10.1016/j.radphyschem.2013.12.028. [ Links ]

11. S. Domínguez-Almendros, N. Benítez-Parejo, and A. R. González-Ramírez, Logistic regression models, Allergologia et immunopathologia 39 (2011) 295, https://doi.org/10.1016/j.aller.2011.05.002. [ Links ]

12. K. S. Fatah and Z. R. A. Alkaki, Application of Binary Logistic Regression Model to Cancer Patients: a case study of data from Erbil in Kurdistan region of Iraq., Zanco Journal of Pure and Applied Sciences 33 (2021) 117, http://dx.doi.org/10.21271/ZJPAS.33.4.12. [ Links ]

13. Q. M. Abdulqader, Applying the binary logistic regression analysis on the medical data, Science Journal of University of Zakho 5 (2017) 330, https://doi.org/10.25271/2017.5.4.388. [ Links ]

14. D. L. Miller, C. M. Hilohi, and D. C. Spelic, Patient radiation doses in interventional cardiology in the US: advisory data sets and possible initial values for US reference levels, Medical physics 39 (2012) 6276, https://doi.org/10.1118/1.4754300. [ Links ]

15. J.-L. Georges, et al., Patient exposure to X-rays during coronary angiography and percutaneous transluminal coronary intervention: Results of a multicenter national survey, Catheterization and Cardiovascular Interventions 83 (2014) 729, https://doi.org/10.1002/ccd.25327. [ Links ]

16. M. Abuzaid, et al., Radiation dose to the paediatric undergoing diagnostic coronary angiography and percutaneous intervention procedures, Radiation Physics and Chemistry 167 (2020) 108265, https://doi.org/10.1016/j.radphyschem.2019.04.015. [ Links ]

Received: July 28, 2022; Accepted: September 06, 2022

Creative Commons License This is an open-access article distributed under the terms of the Creative Commons Attribution License