Document rByEyxb6EY8XkNqm5m90YX9dE

t OBSERVATIONS ON THE VINYL CHLORIDE STUDY FROM A STATISTICAL VIEWPOINT The Objective and Scope of the Study The vinyl chloride study was undertaken to define an acceptable career exposure level to vinyl and vinylidene chloride in light of data collected from a human population exposed over a period of years to these chemicals. The study was retrospective in that the data analyzed was assembled from informa tion already available. Ninety-seven employees working within the area of concern were declared in-scope of the study on the basis of having experienced no serious acute exposures and having had little or no contact with other toxic chemicals. For many of these employees, complete medical information was not available. Physical exams had been given on a voluntary basis in 1958, 1959-60, 1961, and 1965-66. A summary of lab procedures under investigation and the number of employees examined is shown in Table A. The non-response rate is high for some tests and must be considered in the final interpretation of results. The procedures appearing in Table A were selected from a broader list of lab tests available from the medical records because these showed promise of correlation with exposure variables based on a preliminary analysis of the more complete set of data. The Industrial Hygiene section of the Bio-Chem Department was responsible for determining exposure levels experienced by plant workers in the study population. The time-weighted average concentrations of VC1 and VC12 were measured for each job classification on a yearly basis. It was necessary in several cases to project backwards in order to estimate the exposure levels in years prior to installation of the monitoring devices. Thus, the exposure data for the ith individual consists of a 1 X n^ matrix of yearly TWA exposure levels for the n^ years preceding the date of medical exam. The details of measuring these yearly time-weighted averages will not be discussed here. The Method of Analysis The question posed by this study and the methods of data collection made it rather difficult to specify a statistical model for which all assumptions of the model could be met. Since definite medical diagnoses were not available. R&S 033608 2- - the strategy of analysis was to attempt to explain variation in the observed clinical variables by first removing effects resulting from non-environmental factors such as age, and then to see if the unexplained variation could be further reduced by the introduction of variables representing some function of the exposure levels to YC1 and VC12. This led to the use of multiple regression analysis. Obviously, there was no way of arbitrarily selecting the levels of the independent variables. As a result .very few observations occurred at the higher exposure levels which were of most interest. The assumptions that the errors were independently normally distributed with constant variance was not rigorously tested. However, these assumptions appear to be reason ably satisfied. Findings (1) Six lab procedures of the eleven under study showed correlations with one or more exposure variables for at least one periodic exam series at the . 05 level. These were BSP, B Protein, Icterus Index, Hemoglobin, Systolic and Diastolic Blood Pressure. (2) All six of these lab tests showed correlations during the 1965-66 period, while only hemoglobin showed a significant correlation at the . 05 level prior to this date; that being in 1958. However, two tests, B Protein and Icterus Index, were administered only in 1965-66 and one, the BSP, was obtained only on a limited basis prior to 1965. (3) In the final analysis much weight is placed on the readings ob served for those few employees who experienced high exposure levels, especially in the 1949-58 period of their careers. For example, the average exposure level in 1950 was 153 ppm, whereas the average exposure level for 1965 was 29 ppm. (4) The BSP test showed the most pronounced correlation with one o more exposure variables. (5) Three individuals in the study had Time-Weighted-Averages of 595 ppm in 1953. Two of these were given physicals in 1965. The weight of their relatively high exposure levels in the early 50's is evident from Figure I where they represent the two highest time-weighted averages despite average exposure levels in the 60's. R&S 033609 R&s 033610 -3- (6) The correlation between age and TWA concentration of chemical exposure increased as exam dates became more recent with 1965-66 showing the highest correlations. Year Correlation 1958 1959-60 1961 1965-66 .048 .125 .153 .361 (7) The average career length as of the 1965-66 physicals was 14.4 years. (8) A number of interrelated exposure measures were examined, such as dose, number of years of TWA greater than 50, 100, or 200 ppm and several interactions between exposure measures and age or V weight. However, TWA was relied upon where possible because of the ease in presenting and inter preting results. (9) Although no significant correlations in the 1958-61 BSP readings were detected after removing the weight affect, the mean BSP value for the 16 observations was a high 4. 06 and the average exposure level was 120 ppm. (Tentative) Conclusions There is strong evidence that several clinical results were influenced by the amount of exposure to YC1 and VC12 experienced by study subjects. It is not known if the effects are the result of accumulative action over the entire career of the individual or the result of the higher exposure levels experienced prior to 1958. The latter possibility seems more plausible at this time. Effects were detected in 1958 and 1965-66; none were observed in 1959-60 or 1961. A delayed response would account for the effects detected in 1965-66 after the over-all exposure levels had been lowered con siderably. Assuming this explanation to be correct, the upper exposure limit could then be set by following back on the careers of those subjects showing effect in 1965 to their exposure levels prior to 1958. M. Gerald Ott Corporate Medical Department 4/22/68 TABLE A. Clinical Measurements Obtained on Members of the Study Population by Year of Periodic Exam R&S 033611 Clinical Variable Number of In-Scope Employees Weight Height Blood Pressure, Sys. Blood Pressure, Dia. BSP Hemoglobin Hematoc rit Thymol Turbidity SGOT Prothrombin B Protein Icterus Index Aik. Phosphatase Last PE Prior to VC1 Exp Number of Employees 1958 1959 1961 97 87 89 88 96 69 80 72 96 69 80 72 96 69 80 69 96 69 80 69 11 4 1 35 60 74 59 44 73 71 1965 92 70 70 69 69 61 44 66 66 66 58 16 APPENDIX I ANOYA TABLES FOR LAB PROCEDURES SHOWING SIGNIFICANT CORRELATIONS WITH ONE OR MORE EXPOSURE VARIABLES Table 1. ANOYA for BSP readings obtained on 61 employees during 1965-66 Source Total Regression Residuals Lack of Fit Pure Error d, f. 60 2 58 30 28 SS 202.7 108. 1 94.6 61.5 33. 1 MS 54.05 1.63 2.05 1.18 F-Value 33.1** 1.7 Coefficients No. XI X2 Variable Dose prior to 1954 (Yrs of 200+ Exp)2 Mean 321.6 3. 95 Original Correlation .70 .69 Coeff/S. E. 2. 57* 2. 24* Prediction Equation BSP = 2. 28 + . 0014 XI + . 0636 X2 R&S 033612 Table 2. Alternate ANOVA for BSP readings obtained on 61 employees during 1965--66. Source T otal Regression Residuals d.f. 60 2 58 SS 202. 7 110. 1 92. 6 MS 55.05 1.59 F-Value 34.4** Coefficients No. XI X2 Variable ^Weight (TWA/200)4 Mean - 4. 59 . 069 Original Correlation . 20 .71 Coeff/S. E. 1.58 8. 00** Prediction Equation BSP = 2. 56 + .011 XI + 7.09 X2 Significant at . 05 level. 'I' #T" Significant at . 01 level. Table 3, ANOVA for B Protein readings obtained on 58 employees during 1965-66. Source Total Regression Residuals d.f. 57 2 55 SS 445. 0 63.7 381.3 MS 31.85 6. 93 F-Value 4.60=!= Coefficients No. Variable XI V Weight X2 (TWA/200)2 Mean - 5.07 . 145 Original Correlation . 237 . 322 Coeff/S.E 1.60 2. 36* Prediction Equation B Prot = 14. 22 + . 026 XI + 3. 82 X2 R&S 033613 Table 4. ANOVA for Icterus Index readings obtained on 16 employees during 1965-66. Source Total Regre s sion Residuals d.f. 15 3 12 SS 50.7 30. 5 20. 2 MS 10. 17 1.68 F-Value 6.03** Coefficients No. Variable XI Age X2 (TWA)2 X3 ( V Weight)2 Mean 38. 5 5601.7 398. 2 Original Correlation .378 .460 - .452 Coeff/S.E 2. 92* 2.44* 1.99 Prediction Equation Icterus Index = 12.74 - .143X1 + .00009X2 - .0013X3 R&S 033614 Table 5. ANOVA for Hemoglobin readings obtained on 35 employees during 1958. Source Total Regression Residuals d.f. 34 4 30 SS 34. 20 15. 97 18. 23 MS 3.99 . 608 F -Value 6.57** Coe fficients No. Variable XI Age X2 History of Jaundice X3 V Weight X4 (TWA) Mean 37. 2 . 20 -6. 66 108.7 Original Correlation -.484 . 303 . 07 0 -.368 Coeff/S. E. 2. 27* 2.75 * 1.91 3.05** Prediction Equation Hemoglobin = 16.44 - . 038 XI + 1. 1 X2 + . 015 X3 - . 0048 X4 Table 6. ANOVA for Hemoglobin readings obtained on 44 employees during 1965-66. Source Total Regression Residuals d.f. 43 2 41 SS 51.44 8. 24 43. 20 MS 4. 12 1.06 F -Value 3. 89* Coefficients No. Variable XI History of Jaundice X2 TWA Mean . 09 61.8 Original Correlation . 176 -.27 0 Coeff/S 2. 10* 2. 55* Prediction Equation Hemoglobin = 14.98 + 1.2 X1 -. 0088 X2 Table 7. ANOVA for Systolic Blood Pressure readings on 69 employees during 1965-66. Source Total Regression Residuals d.f. 68 3 65 SS 13983. 5 5002.I. 8981.4 MS 1667.4 138. 2 F -Value 12. 1** No. XI X2 X3 Variable Age V Weight Age X(Yrs of 100+ Exp) Mean 43. 2 - 2. 1 134. 3 Original Correlation . 474 . 205 .489 Coeff/S.E. 2. 97++ 2. 10+ 2. 80++ Prediction Equation Sys. BP = 104. 39 + .48 XI + . 13 X2 + .025 X3 R&S 033615 Table 8. ANOVA for Systolic Blood Pressure readings on 69 employees during 1965-66 (Alternate) Source Total Regression Residuals d.f. 68 3 65 SS 13983. 5 4530.8 9452.7 MS 1510.3 145. 4 F -Value 10 4^^ vjoemcients No. Variable XI Age X2 V Weight X3 TWA Mean 43. 2 - 2. 1 60.6 Original Correlation . 474 . 205 . 395 Coeff/S.E. 3. 64++ 2. 10 + 2. 05 + Prediction Equation Sys. BP = 99.47 + . 58 XI + . 13 X2 + . 069 X3 Table 9. ANOYA for Diastolic Blood Pressure readings on 69 employees during 1965-66. Source Total Regression Residuals d.f. 68 3 65 SS 5285.1 1618.3 3666.8 MS 539.4 56.4 F-Value 9.56** Coefficients No. Va-r-iable XI In Age X2 V Weight X3 In Age X(Yrs of 100 + Exp) Mean 3. 74 - 2. 1 10.79 Original Correlation . 461 ..200 . 379 Coeff/S. E. 3.50** 2. 13* 1.84 Prediction Equation BPd. = 25. 02 + 14. 66 XI + . 081 X2 + . 126 X3 R&S 033616 Table 10. ANOYA for Diastolic Blood Pressure readings on 69 employees during 1965-66 (Alternate) Source Total Regression Residuals df. SS MS F- Value 68 5285.1 3 1729. 9 576. 6 10.54** 65 3555.2 54.7 Coefficients No. Variable XI In Age X2 V Weight X3 TWA Mean 3.74 - 2. 1 60.6 Original Correlation . 461 . 200 .424 Coeff/S. E. 3. 39** 2. 07* 2. 35* Prediction Equation BPd = 26. 09 + 13. 94 XI + . 078 X2 + . 049 X3 Table 11. ANOVA for the change in Systolic Blood Pressure readings on 68 employees between their pre-exposure physical and their 1965-66 physical. Source Total Regression Residuals d.f. 67 3 64 SS 13357.8 4823. 8 8534.0 MS 1607.9 133. 3 F-Value 12. 04** Coefficients No. Variable XI A Age X2 Pre-Exp BP X3 TWA SyS Mean 14. 39 128.47 58. 62 Original Correlation . 330 - . 362 . 467 Coeff/S. E. 1.26 3. 67** 3. 19** Prediction Equation BP s = 48. 59 + .36 XI - . 47 X2 + . 12 X3 R&S 033617 Table 12. ANOVA for the change in Diastolic Blood Pressure readings on 68 employees between their pre-exposure physical and their 1965-66 physical. Source Total Regression Residuals d.f. 67 2 65 SS 6939.2 2944.3 3994.9 MS 1472. 1 61.46 F -Value 23.95** Coefficients No. Variable XI Pre-Exp BP,. X2 TWA Mean 80.60 58. 62 Original Correlation - . 546 . 311 Coeff/S. E. 6.13** 3. 80** Prediction Equation BFd = 54. 59 - .73 XI + . 082 X2 APPENDIX II ANALYSIS OF LAB PROCEDURES FOR YEARS DURING WHICH NO SIGNIFICANT CORRELATION WITH ONE OR MORE EXPOSURE VARIABLES WAS DETECTED AT THE . 05 LEVEL AFTER REMOVAL OF AFFECTS DUE TO NON-EXPOSURE VARIABLES Year_____ N Mean Std Dev______________Regression of Non-Exposure Variables Alkaline Phosphatase 1959 73 7. 2 2. 0 Aik Phos = 5.4 +. 044 (age) F -Value 3. 1 1 1,69 Blood Pressure, Diastolic PE 96 80. 3 8. 1 1958 69 81.2 9.0 1959 80 79. 8 7.7 1961 71 79. 9 8.4 196 5 69 81.0 8. 8 BPd = 59.66 + 6. 33(in age) + . 055 ( 7 wt) BPd = 56. 27 + 7. 08 (In age) + . 10 ( V wt) BPd = 49. 03 + 8. 57 (In age) + . 046 ( 7 wt) BPd = 42. 41 + 10. 36 (In age) + .13 ( V wt) See Appendix I F 2, 93 ,F * 2 66 F 2, 77 F 2, 68 2. 821 3. 36* 3. 84* 10.01** Blood Pressure, Systolic PE 96 127. 3 11.6 1958 69 129. 2 12. 9 1959 80 126.5 11.7 1961 71 130. 0 1 3. 4 1965 69 128.4 14. 3 BPs = 119. 57 + . 31 (age) + .14 (Vwt) BPg = 118.75 + .29 (age) + .10 ( V wt) BPg = 118. 08 + .23 (age) + . 11 ( SJ wt) BPg = 114. 66 + . 40 (age) + . 19 ( V wt) See Appendix I F 2, 93 F 2, 66 F2, 77 F2, 68 4.78* 2. 451 3. 29* 7.69** Bromsu lphalein 1959 16 1965 61 4. 06 3. 00 2. 0 1.8 Wt itself is not significant at the . 10 level - 16 obs. but after removing its affects no exposure variables can remain in the regression equation. See Appendix I 8L9CS0 SSU 1 Significant at . 1 0 level. APPENDIX Il^Pbntinued) Year N Hematocrit 1959 74 1961 59 1965 67 Mean Std De v 39. 3 45. 1 47.3 2. 9 2.7 2. 5 Regression of Non-Exposure Variables Not sign, correlated with non-exposure variables at . 1 level 48. 11 - . 072 (age) Not sign, correlated with non-exposure variables at . 1 level F -Value F1i58 = 3.S21 Hemoglobin 1958 35 1961 60 1965 44 14.6 14.4 14. 6 1.0 .8 1. 1 See Appendix I Hgb = 1 5. 34 - .023 (age) + . 58 (HistJaun) + . 0067 ( Vw*) See Appendix I F3, 56 = 3-54* Prothrombin 1959 73 96.0 8.0 P = 89. 32 + . 17 (age) Serum Glutamic Oxaloacetic Transaminase 1965 66 21. 1 7.7 SGOT = 27.9 - . 16 (age) F 1,72 = 2. 98 1 Fl,65= 2'92 Thymol Turbidity 1961 45 1. 53 1965 66 2. 30 1.2 1. 5 T = 1.59 + . 014 ( V wt) Not sign, correlated with non-exposure variables at . 1 level 1 F.1, 4. 4. = 3. 17 6t9eeo ssy