Skip to main content
NCBI home page
Dose-Response logoLink to Dose-Response
. 2017 Mar 28;15(1):1559325817693345. doi: 10.1177/1559325817693345

Fine Particulate Matter and Total Mortality in Cancer Prevention Study Cohort Reanalysis

James E Enstrom 1,✉
PMCID: PMC5407529  PMID: 28473741

Abstract

Background:

In 1997 the US Environmental Protection Agency (EPA) established the National Ambient Air Quality Standard (NAAQS) for fine particulate matter (PM2.5), largely because of its positive relationship to total mortality in the 1982 American Cancer Society Cancer Prevention Study (CPS II) cohort. Subsequently, EPA has used this relationship as the primary justification for many costly regulations, most recently the Clean Power Plan. An independent analysis of the CPS II data was conducted in order to test the validity of this relationship.

Methods:

The original CPS II questionnaire data, including 1982 to 1988 mortality follow-up, were analyzed using Cox proportional hazards regression. Results were obtained for 292 277 participants in 85 counties with 1979-1983 EPA Inhalable Particulate Network PM2.5 measurements, as well as for 212 370 participants in the 50 counties used in the original 1995 analysis.

Results:

The 1982 to 1988 relative risk (RR) of death from all causes and 95% confidence interval adjusted for age, sex, race, education, and smoking status was 1.023 (0.997-1.049) for a 10 µg/m3 increase in PM2.5 in 85 counties and 1.025 (0.990-1.061) in the 50 original counties. The fully adjusted RR was null in the western and eastern portions of the United States, including in areas with somewhat higher PM2.5 levels, particularly 5 Ohio Valley states and California.

Conclusion:

No significant relationship between PM2.5 and total mortality in the CPS II cohort was found when the best available PM2.5 data were used. The original 1995 analysis found a positive relationship by selective use of CPS II and PM2.5 data. This independent analysis of underlying data raises serious doubts about the CPS II epidemiologic evidence supporting the PM2.5 NAAQS. These findings provide strong justification for further independent analysis of the CPS II data.

Keywords: epidemiology, PM2.5, deaths, CPS II, reanalysis

Introduction

In 1997 the US Environmental Protection Agency (EPA) established the National Ambient Air Quality Standard (NAAQS) for fine particulate matter (PM2.5), largely because of its positive relationship to total mortality in the 1982 American Cancer Society (ACS) Cancer Prevention Study (CPS II) cohort, as published in 1995 by Pope et al.1 The EPA uses this positive relationship to claim that PM2.5 causes premature deaths. However, the validity of this finding was immediately challenged with detailed and well-reasoned criticism.2-4 The relationship still remains contested and much of the original criticism has never been properly addressed, particularly the need for truly independent analysis of the CPS II data.

The EPA claim that PM2.5 causes premature deaths is implausible because no etiologic mechanism has ever been established and because it involves the lifetime inhalation of only about 5 g of particles that are less than 2.5 µm in diameter.5 The PM2.5 mortality relationship has been further challenged because the small increased risk could be due to well-known epidemiological biases, such as, the ecological fallacy, inaccurate exposure measurements, and confounding variables like copollutants. In addition, there is extensive evidence of spatial and temporal variation in PM2.5 mortality risk (MR) that does not support 1 national standard for PM2.5.

In spite of these serious problems, EPA and the major PM2.5 investigators continue to assert that their positive findings are sufficient proof that PM2.5 causes premature deaths. Their premature death claim has been used to justify many costly EPA regulations, most recently, the Clean Power Plan.6 Indeed, 85% of the total estimated benefits of all EPA regulations have been attributed to reductions in PM2.5-related premature deaths. With the assumed benefits of PM2.5 reductions playing such a major role in EPA regulatory policy, it is essential that the relationship of PM2.5 to mortality be independently verified with transparent data and reproducible findings.

In 1998, the Health Effects Institute (HEI) in Boston was commissioned to conduct a detailed reanalysis of the original Pope 1995 findings. The July 2000 HEI Reanalysis Report (HEI 2000) included “PART I: REPLICATION AND VALIDATION” and “PART II: SENSITIVITY ANALYSES.”7 The HEI Reanalysis Team lead by Daniel Krewski successfully replicated and validated the 1995 CPS II findings, but they did not analyze the CPS II data in ways that would determine whether the original results remained robust using different sources of air pollution data. For instance, none of their models used the best available PM2.5 measurements as of 1995.

Particularly troubling is the fact that EPA and the major PM2.5 investigators have ignored multiple null findings on the relationship between PM2.5 and mortality in California. These null findings include my 2005 paper,8 2006 clarification,9 2012 American Statistical Society Joint Statistical Meeting Proceedings paper,10 and 2015 International Conference on Climate Change presentation about the Clean Power Plan and PM2.5-related cobenefits.6 There is now overwhelming evidence of a null PM2.5 mortality relationship in California dating back to 2000. The problems with the PM2.5 mortality relationship have generated substantial scientific and political concern.

During 2011 to 2013, the US House Science, Space, and Technology Committee (HSSTC) repeatedly requested that EPA provide access to the underlying CPS II data, particularly since substantial Federal funding has been used for CPS II PM2.5 mortality research and publications. On July 22, 2013, the HSSTC made a particularly detailed request to EPA that included 49 pages of letters dating back to September 22, 2011.11 When EPA failed to provide the requested data, the HSSTC issued an August 1, 2013 subpoena to EPA for the CPS II data.12 The ACS refused to comply with the HSSTC subpoena, as explained in an August 19, 2013 letter to EPA by Chief Medical Officer Otis W. Brawley.13 Then, following the subpoena, ACS has refused to work with me and 3 other highly qualified investigators regarding collaborative analysis of the CPS II data.14 Finally, HEI has refused to conduct my proposed CPS II analyses.15 However, my recent acquisition of an original version of the CPS II data has made possible this first truly independent analysis.

Methods

Computer files containing the original 1982 ACS CPS II deidentified questionnaire data and 6-year follow-up data on deaths from September 1, 1982 through August 31, 1988, along with detailed documentation, were obtained from a source with appropriate access to these data, as explained in the “Acknowledgments.” This article presents my initial analysis of the CPS II cohort and it is subject to the limitations of data and documentation that is not as complete and current as the data and documentation possessed by ACS.

The research described below is exempt from human participants or ethics approval because it involved only statistical analysis of existing deidentified data. Human participants’ approval was obtained by ACS in 1982 when each individual enrolled in CPS II. Because of the epidemiologic importance of this analysis, an effort will be made to post on my Scientific Integrity Institute website a version of the CPS II data that fully preserves the confidentiality of all of participants and that contains enough information to verify my findings.

Of the 1.2 million total CPS II participants, analysis has been done on 297 592 participants residing in 85 counties in the continental United States with 1979 to 1983 EPA Inhalable Particulate Network (IPN) PM2.5 measurements.16,17 Among these participants, there were 18 612 total deaths from September 1, 1982 through August 31, 1988; 17 329 of these deaths (93.1%) had a known date of death. Of the 297 592 participants, 292 277 had age at entry of 30 to 99 years and sex of male [1] or female [2]. Of the 292 277 participants, 269 766 had race of white [1,2,5] or black [3,4]; education level of no or some high school [1,2], high school graduate [3], some college [4,5], college graduate [6], or graduate school [7]; and smoking status of never [1], former [5-8 for males and 3 for females], or current [2-4 for males and 2 for females]. Those participants reported to be dead [D, G, K] but without an exact date of death have been assumed to be alive in this analysis. The unconfirmed deaths were randomly distributed and did not impact relative comparisons of death in a systematic way. The computer codes for the above variables are shown in brackets.

CPS II participants were entered into the master data file geographically. Since this deidentified data file does not contain home addresses, the Division number and Unit number assigned by ACS to each CPS II participant have been used to define their county of residence. For instance, ACS Division 39 represents the state of Ohio and its Unit 041 represents Jefferson County, which includes the city of Steubenville, where the IPN PM2.5 measurements were made. In other words, most of the 575 participants in Unit 041 lived in Jefferson County as of September 1, 1982. The IPN PM2.5 value of 29.6739 µg/m3, based on measurements made in Steubenville, was assigned to all CPS II participants in Unit 041. This PM2.5 value is a weighted average of 53 measurements (mean of 33.9260 µg/m3) and 31 measurements (mean of 29.4884 µg/m3) made during 1979 to 198216 and 53 measurements (mean of 27.2473 µg/m3) and 54 measurements (mean of 28.0676 µg/m3) made during 1983.17 The IPN PM2.5 data were collected only during 1979 to 1983, although some other IPN air pollution data were collected through 1984. The values for each county that includes a city with CPS II participants and IPN PM2.5 measurements are shown in Appendix Table A1.

To make the best possible comparison with Pope 1995 and HEI 2000 results, the HEI PM2.5 value of 23.1 µg/m3 for Steubenville was assigned to all participants in Unit 041. This value is the median of PM2.5 measurements made in Steubenville and is shown in HEI 2000 Appendix D “Alternative Air Pollution Data in the ACS Study.”7 Analyses were done for the 50 counties containing the original 50 cities with CPS II participants and HEI PM2.5 values used in Pope 1995 and HEI 2000. Additional analyses were done for all 85 counties containing cities with both CPS II participants and IPN PM2.5 data. Without explanation, Pope 1995 and HEI 2000 omitted from their analyses, 35 cities with CPS II participants and IPN PM2.5 data. To be clear, these analyses are based on the CPS II participants assigned to each Unit (county) that included a city with IPN PM2.5 data. The original Pope 1995 and HEI 2000 analyses were based on the CPS II participants assigned to each metropolitan area (MA) that included a city with HEI PM2.5 data, as defined in HEI 2000 Appendix F “Definition of Metropolitan Areas in the ACS Study.”7 The MA, which was equivalent to the US Census Bureau Standard Metropolitan Statistical Area (SMSA), always included the county containing the city with the HEI PM2.5 data and often included 1 or more additional counties.

The SAS 9.4 procedure PHREG was used to conduct Cox proportional hazards regression.18 Relative risks (RRs) for death from all causes and 95% confidence intervals (CI) were calculated using age–sex adjustment and full adjustment (age, sex, race, education, and smoking status, as defined above). Each of the 5 adjustment variables had a strong relationship to total mortality. Race, education, and smoking status were the 3 adjustment variables that had the greatest impact on the age–sex-adjusted RR. The Pope 1995 and HEI 2000 analyses used 4 additional adjustment variables that had a lesser impact on the age–sex-adjusted RR.

In addition, county-level ecological analyses were done by comparing IPN PM2.5 and HEI PM2.5 values to 1980 age-adjusted white total death rates (DRs) determined by the Centers for Disease Control and Prevention (CDC) WONDER19 and mortality risks (MRs) as shown in Figures 5 and 21 of HEI 2000.7 Death rates are age adjusted to the 2000 US Standard Population and are expressed as annual deaths per 100 000 persons. The SAS 9.4 procedure REGRESSION was used to conduct linear regression of PM2.5 values with DRs and MRs.

Appendix Table A1 lists the 50 original cities used in Pope 1995 and HEI 2000 and includes city, county, state, ACS Division and Unit numbers, Federal Information Processing Standards (FIPS) code, IPN average PM2.5 level, HEI median PM2.5 level, 1980 DR, and HEI MR. Appendix Table A1 also lists similar information for the 35 additional cities with CPS II participants and IPN PM2.5 data. However, HEI PM2.5 and HEI MR data are not available for these 35 cities.

Results

Table 1 shows basic demographic characteristics for the CPS II participants, as stated in Pope 1995,1 HEI 2000,7 and this current analysis. There is excellent agreement on age, sex, race, education, and smoking status. However, the IPN PM2.5 averages are generally about 20% higher than the HEI PM2.5 medians, although the differences range from +78% to −28%.

Table 1.

Summary Characteristics of CPS II Participants in (1) Pope 1995 Table 1,1 (2) HEI 2000 Table 24,7 and (3) Current Analysis Based on CPS II Participants in 50 and 85 Counties.

Characteristics Pope 1995 Table 1 HEI 2000 Table 24 Current CPS II Analysis
n = 50 HEI PM2.5 n = 50 IPN PM2.5 n = 85 IPN PM2.5
Number of metro areas 50 50
Number of counties Not stated Not stated 50 50 85
Age–sex-adjusted participants 212 370 212 370 292 277
Fully adjusted participants 295 223 298 817 195 215 195 215 269 766
Age–sex-adjusted deaths 12 518 12 518 17 231
Fully adjusted deaths 20 765 23 093 11 221 11 221 15 593
Values below are for participants in fully adjusted results
Age at enrollment, mean years 56.6 56.6 56.66 56.66 56.64
Sex (% females) 55.9 56.4 56.72 56.72 56.61
Race (% white) 94.0 94.0 94.58 94.58 95.09
Less than high school education, % 11.3 11.3 11.71 11.71 11.71
Never smoked regularly, % 41.69 41.69 41.57
Former smoker, % 33.25 33.25 33.67
Former cigarette smoker, % 29.4 30.2 30.43 30.43 30.81
Current smoker, % 25.06 25.06 24.76
Current cigarette smoker, % 21.6 21.4 21.01 21.01 20.76
Fine particles, µg/m3
 Average 18.2 18.2 17.99 21.37 21.16
 SD 5.1 4.4 4.52 5.30 5.98
 Range 9.0-33.5 9.0-33.4 9.0-33.4 10.77-29.67 10.63-42.01
Open in a new tab

Abbreviations: CPS, Cancer Prevention Study; HEI, Health Effects Institute; IPN, Inhalable Particulate Network; PM2.5, fine particulate matter.

Table 2 shows that during 1982 to 1988, there was no significant relationship between IPN PM2.5 and total mortality in the entire United States. The fully adjusted RR and 95% CI was 1.023 (0.997-1.049) for a 10 µg/m3 increase in PM2.5 in all 85 counties and 1.025 (0.990-1.061) in the 50 original counties. Indeed, the fully adjusted RR was not significant in any area of the United States, such as, the states west of the Mississippi River, the states east of the Mississippi River, the 5 Ohio Valley states (Indiana, Kentucky, Ohio, Pennsylvania, and West Virginia), and the states other than the Ohio Valley states. The age–sex-adjusted and fully adjusted RRs in the states other than the Ohio Valley states are all consistent with no relationship and most are very close to 1.00. The slightly positive age–sex-adjusted RRs for the entire United States and the Ohio Valley states became statistically consistent with no relationship after controlling for the 3 confounding variables of race, education, and smoking status.

Table 2.

Age–Sex-Adjusted and Fully Adjusted Relative Risk of Death From All Causes (RR and 95% CI) From September 1, 1982 Through August 31, 1988 Associated With Change of 10 µg/m3 Increase in PM2.5 for CPS II Participants Residing in 50 and 85 Counties in the Continental United States With 1979 to 1983 IPN PM2.5 Measurements.a

PM2.5 Years and Source Number of Counties Number of Participants Number of Deaths RR 95% CI Lower Upper Average PM2.5
Age–sex adjusted RR for the continental United States
 1979-1983 IPN 85 292 277 17 321 1.038 (1.014-1.063) 21.16
 1979-1983 IPN 50 212 370 12 518 1.046 (1.013-1.081) 21.36
 1979-1983 HEI 50 212 370 12 518 1.121 (1.078-1.166) 17.99
Fully adjusted RR for the continental United States
 1979-1983 IPN 85 269 766 15 593 1.023 (0.997-1.049) 21.15
 1979-1983 IPN 50 195 215 11 221 1.025 (0.990-1.061) 21.36
 1979-1983 HEI 50 195 215 11 221 1.082 (1.039-1.128) 17.99
Age–sex adjusted RR for Ohio Valley States (IN, KY, OH, PA, WV)
 1979-1983 IPN 17 56 979 3649 1.126 (1.011-1.255) 25.51
 1979-1983 IPN 12 45 303 2942 1.079 (0.951-1.225) 25.76
 1979-1983 HEI 12 45 303 2942 1.153 (1.027-1.296) 22.02
Fully adjusted RR for Ohio Valley states (IN, KY, OH, PA, WV)
 1979-1983 IPN 17 53 026 3293 1.096 (0.978-1.228) 25.51
 1979-1983 IPN 12 42 174 2652 1.050 (0.918-1.201) 25.75
 1979-1983 HEI 12 42 174 2652 1.111 (0.983-1.256) 22.02
Age–sex adjusted RR for states other than the Ohio Valley states
 1979-1983 IPN 68 235 298 13 672 0.999 (0.973-1.027) 20.11
 1979-1983 IPN 38 167 067 9576 0.983 (0.946-1.021) 20.18
 1979-1983 HEI 38 167 067 9576 1.045 (0.997-1.096) 16.90
Fully adjusted RR for states other than the Ohio Valley states
 1979-1983 IPN 68 216 740 12 300 0.994 (0.967-1.023) 20.09
 1979-1983 IPN 38 153 041 8569 0.975 (0.936-1.015) 20.15
 1979-1983 HEI 38 153 041 8569 1.025 (0.975-1.078) 16.89
Open in a new tab

Abbreviations: CI, confidence interval; CPS, Cancer Prevention Study; HEI, Health Effects Institute; IPN, Inhalable Particulate Network; PM2.5, particulate matter.

aAnalysis includes continental United States, 5 Ohio Valley states, and remainder of the states. Appendix Table A1 lists the 85 cities and counties with PM2.5 measurements.

However, the fully adjusted RR for the entire United States was 1.082 (1.039-1.128) when based on the HEI PM2.5 values in 50 counties. This RR agrees quite well with the fully adjusted RR of 1.067 (1.037-1.099) for 1982 to 1989, which is shown in Table 34 of the June 2009 HEI Extended Follow-up Research Report (HEI 2009).20 Thus, the positive nationwide RRs in the CPS II cohort depend upon the use of HEI PM2.5 values. The nationwide RRs are consistent with no effect when based on IPN PM2.5 values. The findings in Table 2 clearly demonstrate the large influence of PM2.5 values and geography on the RRs.

Table 3 shows that the fully adjusted RR in California was 0.992 (0.954-1.032) when based on IPN PM2.5 values in all 11 California counties. This null finding is consistent with the 15 other findings of a null relationship in California, all of which are shown in Appendix Table B1. However, when the RR is based on the 4 California counties used in Pope 1995 and HEI 2000, there is a significant inverse relationship. The fully adjusted RR is 0.879 (0.805-0.960) when based on the IPN PM2.5 values and is 0.870 (0.788-0.960) when based on the HEI PM2.5 values. This significant inverse relationship is in exact agreement with the finding of a special analysis of the CPS II cohort done for HEI by Krewski in 2010, which yielded a fully adjusted RR of 0.872 (0.805-0.944) during 1982 to 1989 in California when based on HEI PM2.5 values.21 In this instance, the California RRs are clearly dependent upon the number of counties used.

Table 3.

Age–Sex-Adjusted and Fully Adjusted Relative Risk of Death From All Causes (RR and 95% CI) From September 1, 1982 Through August 31, 1988 Associated With 10 µg/m3 Increase in PM2.5 for California CPS II Participants Living in 4 and 11 Counties With 1979 to 1983 IPN PM2.5 Measurements.a

PM2.5 Years and Source Number of Counties Number of Participants Number of Deaths RR 95% CI of RR Lower Upper Average PM2.5
Age–sex adjusted RR for California during 1982 to 1988
 1979-1983 IPN 11 66 615 3856 1.005 (0.968-1.043) 24.08
 1979-1983 IPN 4 40 527 2146 0.904 (0.831-0.983) 24.90
 1979-1983 HEI 4 40 527 2146 0.894 (0.817-0.986) 18.83
Fully adjusted (age, sex, race, education, and smoking status) RR for California during 1982 to 1988
 1979-1983 IPN 11 60 521 3512 0.992 (0.954-1.032) 24.11
 1979-1983 IPN 4 36 201 1939 0.879 (0.805-0.960) 25.01
 1979-1983 HEI 4 36 201 1939 0.870 (0.788-0.960) 18.91
Fully adjusted (44 confounders) RR for California during 1982 to 1989 as per Krewski21
 “Same” Standard Cox Model 1979-1983 HEI 4 40 408 0.872 (0.805-0.944) ∼19
 “Different” Standard Cox Model 1979-1983 HEI 4 38 925 0.893 (0.823-0.969) ∼19
Open in a new tab

Abbreviations: CI, confidence interval; CPS, Cancer Prevention Study; HEI, Health Effects Institute; IPN, Inhalable Particulate Network; PM2.5, particulate matter.

aAlso, fully adjusted RR for California participants in 4 counties from September 1, 1982 through December 31, 1989 as calculated by Krewski.21

Table 4 shows that the ecological analysis based on linear regression is quite consistent with the proportional hazard regression results in Tables 2 and 3, in spite of the fact that the regression results are not fully adjusted. Using 1980 age-adjusted white total DRs versus HEI PM2.5 values in 50 counties, linear regression yielded a regression coefficient of 6.96 (standard error [SE] = 1.85) that was statistically significant at the 95% confidence level. Pope 1995 reported a significant regression coefficient for 50 cities of 8.0 (SE = 1.4). However, this positive coefficient is misleading because both DRs and PM2.5 levels are higher in the East than in the West. Regional regression analyses did not generally yield significant regression coefficients. Specifically, there were no significant regression coefficients for California, the 5 Ohio Valley states, or all states west of the Mississippi River. These findings reinforce the CPS II cohort evidence of statistically insignificant PM2.5 MR throughout the United States.

Table 4.

Linear Regression Results for 1979 to 1983 IPN PM2.5 and 1979 to 1983 HEI PM2.5 Versus 1980 Age-Adjusted White Total Death Rate (DR) for 85 Counties With IPN PM2.5 Data and for 50 HEI 2000 Counties With IPN PM2.5 and HEI PM2.5 data.

DR or MR, PM2.5 Years and Source Number of Counties DR or MR Intercept DR or MR Slope Lower 95% CI of DR or MR Slope Upper P Value
Entire continental United States
 DR and 1979-1983 IPN 85 892.68 6.8331 3.8483 9.8180 0.0000
 DR and 1979-1983 HEI 50 910.92 6.9557 3.2452 10.6662 0.0004
 MR and 1979-1983 IPN 50 0.6821 0.0102 0.0044 0.0160 0.0009
 MR and 1979-1983 HEI 50 0.6754 0.0121 0.0068 0.0173 0.0000
Ohio Valley states (IN, KY, OH, PA, and WV)
 DR and 1979-1983 IPN 17 941.77 6.0705 −0.0730 12.2139 0.0524
 DR and 1979-1983 HEI 12 1067.29 1.3235 −7.3460 9.9930 0.7408
 MR and 1979-1983 IPN 12 0.8153 0.0077 −0.0054 0.0208 0.2202
 MR and 1979-1983 HEI 12 0.9628 0.0020 −0.0080 0.0121 0.6608
States other than the Ohio Valley states
 DR and 1979-1983 IPN 68 921.45 4.8639 0.9093 8.8186 0.0167
 DR and 1979-1983 HEI 38 934.66 4.8940 −0.4337 10.2218 0.0706
 MR and 1979-1983 IPN 38 0.8111 0.0020 −0.0054 0.0094 0.5891
 MR and 1979-1983 HEI 38 0.7334 0.0072 0.0000 0.0144 0.0491
States west of the Mississippi river
 DR and 1979-1983 IPN 36 920.10 4.0155 −0.9396 8.9706 0.1088
 DR and 1979-1983 HEI 22 930.11 4.1726 −5.2015 13.5468 0.3642
 MR and 1979-1983 IPN 22 0.8663 −0.0025 −0.0162 0.0112 0.7067
 MR and 1979-1983 HEI 22 0.6413 0.0134 −0.0018 0.0285 0.0807
California
 DR and 1979-1983 IPN 11 921.71 3.6516 −1.8230 9.1262 0.1656
 DR and 1979-1983 HEI 4 992.50 1.9664 −46.6929 50.6256 0.8780
 MR and 1979-1983 IPN 4 0.9529 −0.0074 −0.0600 0.0453 0.6072
 MR and 1979-1983 HEI 4 0.8336 −0.0021 −0.0618 0.0576 0.8935
Open in a new tab

Abbreviations: CI, confidence interval; HEI, Health Effects Institute; IPN, Inhalable Particulate Network; MR, mortality risk; PM2.5, particulate matter.

aLinear regression results are also shown for 1979 to 1983 IPN PM2.5 and 1979 to 1983 HEI PM2.5 versus MR for the 50 “cities” (metropolitan areas) in figures 5 and 21 in HEI 2000.

Conclusion

This independent analysis of the CPS II cohort found that there was no significant relationship between PM2.5 and death from all causes during 1982 to 1988, when the best available PM2.5 measurements were used for the 50 original counties and for all 85 counties with PM2.5 data and CPS II participants. However, a positive relationship was found when the HEI PM2.5 measurements were used for the 50 original counties, consistent with the findings in Pope 1995 and HEI 2000. This null and positive evidence demonstrates that the PM2.5 mortality relationship is not robust and is quite sensitive to the PM2.5 data and CPS II participants used in the analysis.

Furthermore, the following statement on page 80 of HEI 2000 raises serious doubts about the quality of the air pollution data used in Pope 1995 and HEI 2000: “AUDIT OF AIR QUALITY DATA. The ACS study was not originally designed as an air pollution study. The air quality monitoring data used for the ACS analyses came from various sources, some of which are now technologically difficult to access. Documentation of the statistical reduction procedures has been lost. Summary statistics for different groups of standard metropolitan statistical areas had been derived by different investigators. These data sources do not indicate whether the tabulated values refer to all or a subset of monitors in a region or whether they represent means or medians.”7

The Pope 1995 and HEI 2000 analyses were based on 50 median PM2.5 values shown in Appendix A of the 1988 Brookhaven National Laboratory Report 52122 by Lipfert et al.22 These analyses did not use or cite the high quality and widely known EPA IPN PM2.5 data in spite of the fact that these data have been available in 2 detailed EPA reports since 1986.16,17 Lipfert informed HEI about the IPN data in 1998: “During the early stages of the Reanalysis Project, I notified HEI and the reanalysis contractors of the availability of an updated version of the IPN data from EPA, which they apparently obtained. This version includes more locations and a slightly longer period of time. It does not appear that the newer IPN data are listed in Appendix G, and it is thus not possible to confirm if SMSA assignments were made properly.”23

Thus, the HEI Reanalysis Team failed to properly “evaluate the sensitivity of the original findings to the indicators of exposure to fine particle air pollution used by the Original Investigators” and failed to select “all participants who lived within each MA for which data on sulfate or fine particle pollution were available.”7 Furthermore, HEI 2009 did not use these data even though the investigators were aware of my 2005 null PM2.5 mortality findings in California,8 which were based on the IPN data for 11 California counties, instead of the 4 California counties used in Pope 1995 and HEI 2000. Indeed, HEI 2009 did not cite my 2005 findings, in spite of my personal discussion of these findings with Pope, Jerrett, and Burnett on July 11, 2008.24 Finally, HEI 2009 did not acknowledge or address my 2006 concerns about the geographic variation in PM2.5 MR clearly shown in HEI 2000 Figure 21,7 which is included here as Appendix Figure C1. HEI 2009 entirely avoided the issue of geographic variation in PM2.5 MR and omitted the equivalent to HEI 2000 Figure 21.

Since 2002, HEI has repeatedly refused to provide the city-specific PM2.5-related MR for the 50 cities included in HEI 2000 Figure 21.15 I estimated these MRs in 2010 based on visual measurements of HEI 2000 Figure 5, and my estimates are shown in Appendix Table A1.25 Figure 21 and its MRs represented early evidence that there was no PM2.5-related MR in California. Appendix Table B1 shows the now overwhelming 2000 to 2016 evidence from 6 different cohorts that there is no relationship between PM2.5 and total mortality in California. Indeed, the weighted average RR of the latest results from the 6 California cohorts is RR = 0.999 (0.988-1.010).26

The authors of the CPS II PM2.5 mortality publications, which began with Pope 1995, have faced original criticism,2-4 my criticism,6-10,14,15 and the criticism of the HSSTC and its subpoena.11-13 Now, my null findings represent a direct challenge to the positive findings of Pope 1995. All of this criticism is relevant to the EPA claim that PM2.5 has a causal relationship to total mortality. The authors of Pope 1995, HEI 2000, and HEI 2009 need to promptly address my findings, as well as the earlier criticism. Then, they need to cooperate with critics on transparent air pollution epidemiology analyses of the CPS II cohort data.

Also, major scientific journals like the New England Journal of Medicine (NEJM) and Science, which have consistently written about the positive relationship between PM2.5 and total mortality, need to publish evidence of no relationship when strong null evidence is submitted to them. In 2015, Science immediately rejected without peer reviewing 3 versions of strong evidence that PM2.5 does not cause premature deaths.5 In 2016, Science immediately rejected without peer reviewing this article. Indeed, this article was rejected by NEJM, Science, and 5 other major journals, as described in a detailed compilation of relevant correspondence.27 Most troubling is the rejection by the American Journal of Respiratory and Clinical Care Medicine, which has published Pope 1995 and several other PM2.5 mortality articles based on the CPS II cohort data.

In summary, the null CPS II PM2.5 mortality findings in this article directly challenge the original positive Pope 1995 findings, and they raise serious doubts about the CPS II epidemiologic evidence supporting the PM2.5 NAAQS. These findings demonstrate the importance of independent and transparent analysis of underlying data. Finally, these findings provide strong justification for further independent analysis of CPS II cohort data.

Supplementary Material

Supplementary material

Acknowledgments

The author thanks the American Cancer Society for helping initiate my epidemiologic career (http://www.scientificintegrityinstitute.org/Detels082773.pdf), for providing me with essential research support for many years (http://www.scientificintegrityinstitute.org/MormonLAT120689.pdf), for granting me unique access to California CPS I cohort data (http://www.scientificintegrityinstitute.org/CACPSI090391.pdf), for selecting me as a Researcher who enrolled CPS II participants and worked with CPS II epidemiologists (http://www.scientificintegrityinstitute.org/Enstrom090213.pdf), and for making it possible for me to obtain unique access to the CPS II cohort data and detailed documentation. In addition, the author sincerely thanks Professors Melvin Schwartz, Lester Breslow, and Nikolai Vavilov, as well as Mr. Lehman Feldenstein, for the training and inspiration that made it possible for me to conduct and publish this research (http://www.scientificintegrityinstitute.org/AFAJEEAS051715.pdf).

Appendix A

Table A1.

List of the 85 Counties Containing the 50 Cities Used in Pope 1995, HEI 2000, and This Analysis, as well as the 35 Additional Cities Used Only in This Analysis.a

State ACS Div-Unit FIPS Code IPN/HEI County Containing IPN/HEI City IPN/HEI City With PM2.5 Measurements 1979-1983 IPN PM2.5, μg/m3, (Weighted Average) 1979-1983 HEI PM2.5, μg/m3 (Median) 1980 Age-Adj White Death Rate (DR) HEI Figure 5 Mortality Risk (MR)
AL 01037 01073 Jefferson Birmingham 25.6016 24.5 1025.3 0.760
AL 01049 01097 Mobile Mobile 22.0296 20.9 1067.2 0.950
AZ 03700 04013 Maricopa Phoenix 15.7790 15.2 953.0 0.855
AR 04071 05119 Pulaski Little Rock 20.5773 17.8 1059.4 0.870
CA 06001 06001 Alameda Livermore 14.3882 1016.6
CA 06002 06007 Butte Chico 15.4525 962.5
CA 06003 06013 Contra Costa Richmond 13.9197 937.1
CA 06004 06019 Fresno Fresno 18.3731 10.3 1001.4 0.680
CA 06008 06029 Kern Bakersfield 30.8628 1119.3
CA 06051 06037 Los Angeles Los Angeles 28.2239 21.8 1035.1 0.760
CA 06019 06065 Riverside Rubidoux 42.0117 1013.9
CA 06020 06073 San Diego San Diego 18.9189 943.7
CA 06021 06075 San Francisco San Francisco 16.3522 12.2 1123.1 0.890
CA 06025 06083 Santa Barbara Lompoc 10.6277 892.8
CA 06026 06085 Santa Clara San Jose 17.7884 12.4 921.9 0.885
CO 07004 08031 Denver Denver 10.7675 16.1 967.3 0.925
CO 07047 08069 Larimer Fort Collins 11.1226 810.5
CO 07008 08101 Pueblo Pueblo 10.9155 1024.1
CT 08001 09003 Hartford Hartford 18.3949 14.8 952.0 0.845
CT 08004 09005 Litchfield Litchfield 11.6502 941.5
DE 09002 10001 Kent Dover 19.5280 959.4
DE 09004 10003 New Castle Wilmington 20.3743 1053.7
DC 10001 11001 Dist Columbia Washington 25.9289 22.5 993.2 0.850
FL 11044 12057 Hillsborough Tampa 13.7337 11.4 1021.8 0.845
GA 12027 13051 Chatham Savannah 17.8127 1029.6
GA 12062 13121 Fulton Atlanta 22.5688 20.3 1063.5 0.840
ID 13001 16001 ADA Boise 18.0052 12.1 892.6 0.600
IL 14089 17031 Cook Chicago 25.1019 21.0 1076.3 0.945
IL 14098 17197 Will Braidwood 17.1851 1054.0
IN 15045 18089 Lake Gary 27.4759 25.2 1129.8 0.995
IN 15049 18097 Marion Indianapolis 23.0925 21.1 1041.2 0.970
KS 17287 20173 Sedgwick Wichita 15.0222 13.6 953.4 0.890
KS 17289 20177 Shawnee Topeka 11.7518 10.3 933.7 0.830
KY 18010 21019 Boyd Ashland 37.7700 1184.6
KY 18055 21111 Jefferson Louisville 24.2134 1095.7
MD 21106 24510 Baltimore City Baltimore 21.6922 1237.8
MD 21101 24031 Montgomery Rockville 20.2009 881.9
MA 22105 25013 Hampden Springfield 17.5682 1025.3
MA 22136 25027 Worcester Worcester 16.2641 1014.6
MN 25001 27053 Hennepin Minneapolis 15.5172 13.7 905.3 0.815
MN 25150 27123 Ramsey St Paul 15.5823 935.7
MS 26086 28049 Hinds Jackson 18.1339 15.7 1087.4 0.930
MO 27001 29095 Jackson Kansas City 17.8488 1090.3
MT 28009 30063 Missoula Missoula 17.6212 938.0
MT 28011 30093 Silver Bow Butte 16.0405 1299.5
NE 30028 31055 Douglas Omaha 15.2760 13.1 991.0 0.880
NV 31101 32031 Washoe Reno 13.1184 11.8 1049.5 0.670
NJ 33004 34007 Camden Camden 20.9523 1146.9
NJ 33007 34013 Essex Livingston 16.4775 1072.7
NJ 33009 34017 Hudson Jersey City 19.9121 17.3 1172.6 0.810
NM 34201 35001 Bernalillo Albuquerque 12.8865 9.0 1014.7 0.710
NY 36014 36029 Erie Buffalo 25.1623 23.5 1085.6 0.960
NY 35001 36061 New York New York City 23.9064 1090.4
NC 37033 37063 Durham Durham 19.4092 16.8 1039.2 1.000
NC 37064 37119 Mecklenburg Charlotte 24.1214 22.6 932.8 0.835
OH 39009 39017 Butler Middletown 25.1789 1108.3
OH 39018 39035 Cuyahoga Cleveland 28.4120 24.6 1089.1 0.980
OH 39031 39061 Hamilton Cincinnati 24.9979 23.1 1095.2 0.980
OH 39041 39081 Jefferson Steubenville 29.6739 23.1 1058.6 1.145
OH 39050 39099 Mahoning Youngstown 22.9404 20.2 1058.4 1.060
OH 39057 39113 Montgomery Dayton 20.8120 18.8 1039.5 0.980
OH 39077 39153 Summit Akron 25.9864 24.6 1064.0 1.060
OK 40055 40109 Oklahoma Oklahoma City 14.9767 15.9 1050.4 0.985
OR 41019 41039 Lane Eugene 17.1653 885.5
OR 41026 41051 Multnomah Portland 16.3537 14.7 1060.8 0.830
PA 42101 42003 Allegheny Pittsburgh 29.1043 17.9 1115.6 1.005
PA 42443 42095 Northampton Bethlehem 19.5265 998.6
PA 43002 42101 Philadelphia Philadelphia 24.0704 21.4 1211.0 0.910
RI 45001 44007 Providence Providence 14.2341 12.9 1006.1 0.890
SC 46016 45019 Charleston Charleston 16.1635 1023.5
TN 51019 47037 Davidson Nashville 21.8944 20.5 981.9 0.845
TN 51088 47065 Hamilton Chattanooga 18.2433 16.6 1087.9 0.840
TX 52811 48113 Dallas Dallas 18.7594 16.5 1024.9 0.850
TX 52859 48141 El Paso El Paso 16.9021 15.7 903.5 0.910
TX 52882 48201 Harris Houston 18.0421 13.4 1025.7 0.700
UT 53024 49035 Salt Lake Salt Lake City 16.6590 15.4 954.3 1.025
VA 55024 51059 Fairfax Fairfax 19.5425 925.7
VA 55002 51710 Norfolk City Norfolk 19.5500 16.9 1139.3 0.910
WA 56017 53033 King Seattle 14.9121 11.9 943.6 0.780
WA 56032 53063 Spokane Spokane 13.5200 9.4 959.2 0.810
WV 58130 54029 Hancock Weirton 25.9181 1094.8
WV 58207 54039 Kanawha Charleston 21.9511 20.1 1149.5 1.005
WV 58117 54069 Ohio Wheeling 23.9840 33.4 1117.5 1.020
WI 59005 55009 Brown Green Bay 20.5462 931.0
WI 59052 55105 Rock Beloit 19.8584 1019.4
Open in a new tab

aEach location includes State, ACS Division Unit number, Federal Information Processing Standards (FIPS) code, IPN/HEI county, IPN/HEI city with PM2.5 measurements, 1979-1983 IPN average PM2.5 level, 1979-1983 HEI median PM2.5 level, 1980 age-adjusted white county total death rate (annual deaths per 100 000), and HEI 2000 figure 5 mortality risk for HEI city (metropolitan area). List also includes 35 additional counties containing cities with IPN PM2.5 data used in this analysis. These 35 counties do not have HEI PM2.5 data.

Appendix B

Table B1.

Epidemiologic Cohort Studies of PM2.5 and Total Mortality in California, 2000 to 2016: Relative Risk of Death From All Causes (RR and 95% CI) Associated With Increase of 10 µg/m3 in PM2.5 (http://scientificintegrityinstitute.org/NoPMDeaths081516.pdf).

Krewski 2000 and 2010a,b CA CPS II Cohort N = 40 408 RR = 0.872 (0.805-0.944) 1982-1989
(N = [18 000 M + 22 408 F]; 4 MSAs; 1979-1983 PM2.5; 44 covariates)
McDonnell 2000c CA AHSMOG Cohort N ∼ 3800 RR ∼ 1.00 (0.95-1.05) 1977-1992
(N∼[1347 M + 2422 F]; SC&SD&SF AB; M RR = 1.09 (0.98-1.21) & F RR∼0.98 (0.92-1.03))
Jerrett 2005d CPS II Cohort in LA Basin N = 22 905 RR = 1.11 (0.99-1.25) 1982-2000
(N = 22 905 M and F; 267 zip code areas; 1999-2000 PM2.5; 44 cov + max confounders)
Enstrom 2005e CA CPS I Cohort N = 35 783 RR = 1.039 (1.010-1.069) 1973-1982
(N = [15 573 M + 20 210 F]; 11 counties; 1979-1983 PM2.5) RR = 0.997 (0.978-1.016) 1983-2002
Enstrom 2006f CA CPS I Cohort N = 35 783 RR = 1.061 (1.017-1.106) 1973-1982
(N = [15 573 M + 20 210 F]; 11 counties; 1979-1983 and 1999-2001 PM2.5) RR = 0.995 (0.968-1.024) 1983-2002
Zeger 2008g MCAPS Cohort “West” N = 3 100 000 RR = 0.989 (0.970-1.008) 2000-2005
(N = [1.5 M M + 1.6 M F]; Medicare enrollees in CA + OR + WA (CA = 73%); 2000-2005 PM2.5)
Jerrett 2010h CA CPS II Cohort N = 77 767 RR ∼ 0.994 (0.965-1.025) 1982-2000
(N = [34 367 M + 43 400 F]; 54 counties; 2000 PM2.5; KRG ZIP; 20 ind cov + 7 eco var; slide 12)
Krewski 2010b (2009) CA CPS II Cohort
(4 MSAs; 1979-1983 PM2.5; 44 cov) N = 40 408 RR = 0.960 (0.920-1.002) 1982-2000
(7 MSAs; 1999-2000 PM2.5; 44 cov) N = 50 930 RR = 0.968 (0.916-1.022) 1982-2000
Jerrett 2011i CA CPS II Cohort N = 73 609 RR = 0.994 (0.965-1.024) 1982-2000
(N = [32 509 M + 41 100 F]; 54 counties; 2000 PM2.5; KRG ZIP Model; 20 ind cov + 7 eco var; Table 28)
Jerrett 2011i CA CPS II Cohort N = 73 609 RR = 1.002 (0.992-1.012) 1982-2000
(N = [32 509 M + 41 100 F]; 54 counties; 2000 PM2.5; Nine Model Ave; 20 ic + 7 ev; Figure 22 and Tables 27-32)
Lipsett 2011j CA Teachers Cohort N = 73 489 RR = 1.01 (0.95-1.09) 2000-2005
(N = [73 489 F]; 2000-2005 PM2.5)
Ostro 2011k CA Teachers Cohort N = 43 220 RR = 1.06 (0.96-1.16) 2002-2007
(N = [43 220 F]; 2002-2007 PM2.5)
Jerrett 2013l CA CPS II Cohort N = 73 711 RR = 1.060 (1.003-1.120) 1982-2000
(N = [∼32 550 M + ∼41 161 F]; 54 counties; 2000 PM2.5; LUR Conurb Model; 42 ind cov + 7 eco var + 5 metro; Table 6)
Jerrett 2013l CA CPS II Cohort N = 73 711 RR = 1.028 (0.957-1.104) 1982-2000
(Same parameters and model as above, except including co-pollutants NO2 and Ozone; Table 5)
Ostro 2015m CA Teachers Cohort N = 101 884 RR = 1.01 (0.98-1.05) 2001-2007
(N = [101 881 F]; 2002-2007 PM2.5) (all natural causes of death)
Thurston 2016n CA NIH-AARP Cohort N = 160 209 RR = 1.02 (0.99-1.04) 2000-2009
(N = [∼95 965 M + ∼64 245 F]; full baseline model: PM2.5 by zip code; Table 3) (all natural causes of death)
Enstrom 2016 unpublished CA NIH-AARP Cohort N = 160 368 RR = 1.001 (0.949-1.055) 2000-2009
(N = [∼96 059 M + ∼64 309 F]; full baseline model: 2000 PM2.5 by county)
Open in a new tab

aKrewski D. “Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality: HEI Special Report. July 2000”. 2000. Figure 5 and Figure 21 of Part II: Sensitivity Analyses http://www.scientificintegrityinstitute.org/HEIFigure5093010.pdf.

bKrewski D. August 31, 2010 letter from Krewski to Health Effects Institute and CARB with California-specific PM2.5 mortality results from Table 34 in Krewski 2009. 2010. http://www.arb.ca.gov/research/health/pm-mort/HEI_Correspondence.pdf

cMcDonnell WF, Nishino-Ishikawa N, Petersen FF, Chen LH, Abbey DE. Relationships of mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers. J Expo Anal Environ Epidemiol. 2000;10(5):427-436. http://www.scientificintegrityinstitute.org/JEAEE090100.pdf

dJerrett M, Burnett RT, Ma R, et al. Spatial Analysis of Air Pollution and Mortality in Los Angeles. Epidemiology. 2005;16(6):727-736. http://www.scientificintegrityinstitute.org/Jerrett110105.pdf

eEnstrom JE. Fine particulate air pollution and total mortality among elderly Californians, 1973-2002. Inhal Toxicol. 2005;17(14):803-816. http://www.arb.ca.gov/planning/gmerp/dec1plan/gmerp_comments/enstrom.pdf, and http://www.scientificintegrityinstitute.org/IT121505.pdf

fEnstrom JE. Response to “A Critique of ‘Fine Particulate Air Pollution and Total Mortality Among Elderly Californians, 1973-2002” by Bert Brunekreef, PhD, and Gerard Hoek, PhD’. Inhal Toxicol. 2006:18:509-514. http://www.scientificintegrityinstitute.org/IT060106.pdf, and http://www.scientificintegrityinstitute.org/ITBH060106.pdf

gZeger SL, Dominici F, McDermott A, Samet JM. Mortality in the Medicare Population and Chronic Exposure to Fine Particulate Air Pollution in Urban Centers (2000-2005). Environ Health Perspect. 2008;116:1614-1619. http://ehp03.niehs.nih.gov/article/info: doi/10.1289/ehp.11449

hJerrett M. February 26, 2010 CARB Symposium Presentation by Principal Investigator, Michael Jerrett, UC Berkeley/CARB Proposal No. 2624-254 “Spatiotemporal Analysis of Air Pollution and Mortality in California Based on the American Cancer Society Cohort”. 2010. http://www.scientificintegrityinstitute.org/CARBJerrett022610.pdf

iJerrett M. October 28, 2011 Revised Final Report for Contract No. 06-332 to CARB Research Screening Committee, Principal Investigator Michael Jerrett, “Spatiotemporal Analysis of Air Pollution and Mortality in California Based on the American Cancer Society Cohort” Co-Investigators: Burnett RT, Pope CA III, Krewski D, Thurston G, Christakos G, Hughes E, Ross Z, Shi Y, Thun M. 2011. http://www.arb.ca.gov/research/rsc/10-28-11/item1dfr06-332.pdf, and http://www.scientificintegrityinstitute.org/Jerrett012510.pdf, and http://www.scientificintegrityinstitute.org/JerrettCriticism102811.pdf

jLipsett MJ, Ostro BD, Reynolds P, et al. Long-term Exposure to Air Pollution and Cardiorespiratory Disease in the California Teachers Study Cohort. Am J Respir Crit Care Med. 2011;184(7);828-835. http://ajrccm.atsjournals.org/content/184/7/828.full.pdf

kOstro B, Lipsett M, Reynolds P, et al. Long-Term Exposure to Constituents of Fine Particulate Air Pollution and Mortality: Results from the California Teachers Study. Environ Health Perspect. 2010;118(3):363-369. http://ehp03.niehs.nih.gov/article/info: doi/10.1289/ehp.0901181

lJerrett M, Burnett RT, Beckerman BS, et al. Spatial analysis of air pollution and mortality in California. Am J Respir Crit Care Med. 2013;188(5):593-599. doi:10.1164/rccm.201303-0609OC. PMID:23805824.

mOstro B, Hu J, Goldberg D, et al. Associations of Mortality with Long-Term Exposures to Fine and Ultrafine Particles, Species and Sources: Results from the California Teachers Study Cohort. Environ Health Perspect. 2015;123(6):549-556. http://ehp.niehs.nih.gov/1408565/, or http://dx.doi.org/10.1289/ehp.1408565

nThurston GD, Ahn J, Cromar KR, et al. Ambient Particulate Matter Air Pollution Exposure and Mortality in the NIH-AARP Diet and Health Cohort. Environ Health Perspect. 2016;124(4):484-490. http://ehp.niehs.nih.gov/1509676/

US EPA. Regulatory Impact Analysis related to the Proposed Revisions to the National Ambient Air Quality Standards for Particulate Matter EPA-452/R-12-003. 2012. http://www.epa.gov/ttn/ecas/regdata/RIAs/PMRIACombinedFile_Bookmarked.pdf

Appendix C

Figure C1.

Figure C1.

Open in a new tab

1982 to 1989 PM2.5 mortality risk (MR) in 50 cities (metropolitan areas) shown in Figure 21 on page 197 of HEI 20007,9 and listed in Appendix Table B1. Figure 21. Spatial overlay of fine particle levels and relative risk of mortality. Interval classifications for fine particles (in g/m3): low 8.99 to 17.03: medium 17.03 to 25.07; high 25.07 to 33. Interval classifications for relative risks of mortality: low 0.052 to 0.711; medium 0.711 to 0.919; high 0.919 to 1.128.

Footnotes

Declaration of Conflicting Interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The American Cancer Society provided the funding for the establishment of the CSP II cohort in 1982, the mortality follow-up from 1982 through 1988, and the preparation of the computerized files and documentation used for this research.

Supplemental Material: The online supplemental material is available at http://journals.sagepub.com/doi/suppl/10.1177/1559325817693345.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary material
Supplement_Material.pdf (6.3MB, pdf)

Articles from Dose-Response are provided here courtesy of SAGE Publications

RESOURCES