Challenges and Opportunities in Estimating the Direct Effects of

Climate on Health

Joel Schwartz

Diane Gold Antonella Zanobetti

Harvard School of Public Health

Department of Environmental Health

What do we know about the Direct Effects of Temperature?

• More people die when it gets hot • More people die when it gets cold • More surprisingly, more people seem to

die when it is only moderately hot and when it is only moderately cold

• Could these effects at less extreme temperature by confounded?

Miami with and without control for Season

Cold effects sensitive to degrees of freedom

But should we control for season?

Schwartz J et al.

Detroit, Ml Honolulu, HI

40 40

u 30 ,........,

~ ~ ~ ~ J ()

0

~ ~ • 0

.......... 20 .......... 20

Q) J

Q)

'- '-~ 10 ~ ..... ..... ro ro

0 '- 0 '-Q) Q)

Cl. -10

Cl.

E E Q)

-20 Q)

-20 I- I-

-30 1992 1995 1998 2001 1992 1995 1998 2001

Q) Date Q) Date ..... ..... ro ro '- 30

'- 30 >. >. ~

..... ro 20

ro 20 t t 0 0

E 10 E 10

c c ro ro Q) 0

Q) 0 E E E -10

E -10 0 0 '- '-- -~ -20 ~ -20 -0 -0

~ 1992 1995 1998 2001 ~ 1992 1995 1998 2001 0 0

Year Year

Von Klot and Schwartz, Environ Health 2012

A Standard Approach in Epidemiology

• Is Stratified Analyses • What if we look at effects of temperature

within month? • Also addresses the question of whether

the effect of a given temperature is independent of when it occurs (60 ° in December Vs in May)

Temperature and Mortality in New York

0 Jan May <!) Sep 0 ..;-

Oct " Feb " Jun e

Mar Jul xx Nov

"' x Apr Aug IB Dec £ "' 0 Q) N 0 c: x ,,.,, "' Q)

:::; E 0

..'= 0

Q)

"" (JJ)

~ Q)

~ of"" Ci c Q) 0

" ~ Q; 0..

0 'i

-20 -10 0 10 20 30 40

Temperature

"' L:: -ro "' 0

:ii " :> c p

'D

~ "' ;,, "' Ci c " e 'D

()_

0

"

u N

0 -

n "i -

0 'f -

0 Jan

" Feb

+ 11.!Ar

" Aµ1

-20

May

'V Jun

Jul

Auy

'" Ill " <>O

'll

-10

Temperature and mortality in New York

" Sep

Oct

' lX Nov

" Dec

10 20 30 40

Temperature (S day Mean) Schwartz J et al.

-

0

0 Cll 0 CL 2

0 CL 3

0 CL4

0 CL S

0 CL 6

0 CL 7

• CL 8

We clustered 200 cities based on Temperature and its standard deviation

POOLED · Lag 0

~ Clusters

CL1

CL2

.., CL3

CL4

CLS

CL6 N

CL7

CL8 er er

"-:

~ 0

POOLED· MA 15 m ci

Clusters-20 -10 0 10 20

CL1

CL2

CL3

CL4

T~ature(C)

CL5

CL6

CL7

CL8

O'I 0

-20 -10 0 10 20 30 40

Tefllle{at ... e (C)

What about Exposure

• What is the connection between airport temperature and temperature where you live?

• What is the connection to temperature you are exposed to?

• We used satellite remote sensing of radiance on a 1 km grid and calibrated it with air temperature readings

predicted air temperature (C)

0 15 30 60 Ki 1ome1e,..

Figure 5

•

Massachusetts A 1 highways

Annual mean minimum

---CJ

0.50 . 1.89 1.89 . 2.61 2.61 . 3.27

3.27- 3.77 3.n - 4.17

4.17 - 4.54

4.54 . 5.05

5.05 - 5.81 Massachusetts state border

[] Selected locations

Kloog I, Chudnovsky A, Koutrakis P ,Schwartz et al. Sci Total Environ 2012 ;432:85-92

Scatterplot and regression results relating indoor to outdoor (a) temperature, (b) apparent temperature, (c) relative humidity, and (d) absolute humidity from May 2011 – April 2012, Greater Boston, Massachusetts. Red line, piecewise linear regression for (a) and (b) and linear regression for (c) and (d); black solid line, reference line placed at knot value; black dashed line, y=x (45 degree) reference line.

Nguyen J, Schwartz J, Dockery D

How can we examine Acclimatization?

• If weak acclimatization to cold is true – We expect the same effect of an unusually

cold day (say at the 1st percentile of temperature) in all cities, since it is equally unusual

– If absolute temperature matters, the effect at the 1st percentile should vary with temperature

– May also vary with other characteristics – Same for unusually hot days (99th percentile)

Extreme cold effect Extreme heat effect

New York City , NY Chicago, IL Boston, MA Houston, TX Detroit, MI Pittsburgh, PA Philadelphia, PA Atlanta, GA Clev eland, OH Seattle, WA Baltimore, MD Dallas, TX St Louis, MO Minneapolis, MN Sacramento, CA Columbus, OH Milwaukee, WI Cincinnati, OH Birmingham, AL Oklahoma City , OK San Francisco, CA Washington DC Kansas City , KS New Hav en, CT Portland, OR Nashv ille, TN New Orleans, LA Tulsa, OK Denv er, CO Charlotte, NC Salt Lake City , UT Jersey City , NJ Youngstown, OH Spokane, WA Albuquerque, NM Canton, OH Austin, TX Greensborough, NC Colorado Springs, CO Prov o, UT Boulder, CO Terra Haute, IN

Summary

2-day cummulative percent change in

-20 -15 -10 -5 0 5 10 15 20 25 30

New York City , NY Los Angeles, CA Chicago, IL Houston, TX Mami, FL Pittsburgh, PA St Louis, MO Boston, MA San Diego, CA Atlanta, GA Detroit, MI Baltimore, MD Philadelphia, PA Broward, FL Phoenix, AZ Clev eland, OH Dallas, TX Tampa, FL Kansas City , KS Honolulu, HI Minneapolis, MN New Hav en, CT Milwaukee, WI Cincinnati, OH Columbus, OH Washington DC Birmingham, AL Orlando, FL Oklahoma City , OK New Orleans, LA Nashv ille, TN Austin, TX Tulsa, OK Salt Lake City , UT Albuquerque, NM Jersey City , NJ Charlotte, NC Greensborough, NC Youngstown, OH Canton, OH Prov o, UT Terra Haute, IN

Summary

2-day cummulative percent chang

-20 -15 -10 -5 0 5 10 15 20 25 30

I2 =3% I2 =86%

Medina-Ramon M, Schwartz J et al.

So it looks like

• The effects of the coldest 1% of days are the same everywhere in the US

• So warming the winteris unlikely to reduce deaths

• The effects of the hottest 1% of days differs substantially across US cities

• So even if we use the dose-response fromcity A which already has the predicted temperature for city B,we will expect changes in mortality

Effect modification by medical condition and subject and area level characteristics of the effect of extreme hot and extreme cold

temperature (1th and 99th percentiles of temperature) on total mortality: meta-analysis of 123 and 111 U.S. cities respectively

during the period 1985–2006, among Medicare enrollees. 1.15

1.10

1.05

1.00

0.95

0.90

Zanobetti A, O’ Neill M, Gronlund C, Schwartz J, ATS conference, 2012

Col

d 1%

Hot

99%

Col

d 1%

Hot

99%

Col

d 1%

Hot

99%

Col

d 1%

Hot

99%

Col

d 1%

Hot

99%

Col

d 1%

Hot

99%

Alzheimer's Dementia Afib Non-white race

High % poverty

High % no High School

Cohort Study of Medicare Participants in 135 US Cities

• Follow a population over time and look atlife expectancy

• What is the Exposure? – Problem – Life expectancy in Finland and Greece is

about the same – Possible Answer – Variability of summer temperature predicts

mortality differences

Hazard Ratio (HR) and 95% Confidence Intervals (CI) for a 1 °C increase in yearly summer temperature standard deviation across the 135 U.S. cities in four cohorts

Summer (June-August) HR 95% CI

Adjusting for ozone

COPD 1.037 1.019 1.055

Diabetes 1.040 1.022 1.059

MI 1.038 1.021 1.055

CHF 1.028 1.013 1.042

Zanobetti A, O’ Neill M, Gronlund C, Schwartz J, Proc Natl Acad Sci USA, 2012, 24;109(17):6608-13

What are potential Mechanisms

• Particularly at less extreme temperatures?

Percent increase in blood pressure for 10 degree increase in temperature and apparent

1.0 0.0

-1.0 -2.0 -3.0 -4.0 -5.0 -6.0

temperature Te

mp

4 da

ys a

ve

adj P

M a

nd o

zone

Tem

p 4

days

ave

adj P

M a

nd o

zone

Tem

p 7

days

ave

adj B

C a

nd o

zone

App

Tem

p 7

days

ave

adj B

C a

nd o

zone

Tem

p 7

days

ave

adj B

C a

nd o

zone

App

Tem

p 7

days

ave

adj B

C a

nd o

zone

Diastolic Blood

pressure

Systolic Blood pressure

Diastolic Blood pressure Systolic Blood pressure

70 Diabetic subjects (322 obs) VA NAS, 928 men (2343 obs)

Hoffmann B, Gold DR, et al. Environ Health Perspect, Halonen JI, Zanobetti A, Sparrow D, Vokonas PS, Schwartz J 2012;120(2):241-6 Occup Environ Med. 2011;68(4):296-301

Changes in mm (95% CI) in brachial artery diameter (BAD) before and after occlusion, for an IQR increase in the prior 24 hours, 2-day and 5-day average temperature (TEMP) and

absolute humidity (AH) in diabetic subjects. 0.3

0.2

0.2

0.1

0.1

0.0

-0.1

TEM

P -

24 h

r

TEM

P -

2 d

ave

TEM

P -

5 d

ave

AH

- 24

hr

AH

- 2

d av

e

AH

- 5

d av

e

TEM

P -

24 h

r

TEM

P -

2 d

ave

TEM

P -

5 d

ave

AH

- 24

hr

AH

- 2

d av

e

AH

- 5

d av

e

Diameter before FMD Diameter after FMD

Zanobetti, Gold DR, et al. , Am J Respir Critic Care Med Vol 185. pp. A3203

Risk factors for cardiovascular disease

30.0

20.0

10.0

0.0

-10.0

-20.0

-30.0

App

tem

p 4

d av

e

App

tem

p 4

d av

e

Tem

p 28

d a

ve

Tem

p 28

d a

ve

Tem

p 28

d a

ve

Tem

p 7

d av

e

App

Tem

p 7

d av

e

Dew

poi

nt 7

d a

ve

Tem

p 14

d a

ve

App

Tem

p 14

d a

ve

BNP CRP CRP VCAM ICAM LDL HDL

100 heart failure sbjects (285 obs)

VA NAS,673 men (1254 obs) VA NAS,478 men (862 obs)

Wilker E, et al. Environ Health Perspec 2012, 120(8):1083-7 Halonen J, et al. Environ Res, 2011;111(2):281-7. Halonen J, et al. Environ Health; 2010: 9:42.

Impact of Meteorological Variables on Atrial Fibrillation

Matched Parameters included in model IQR Odds Ratio (95% CI) P

Adjusted for: Outdoor temperature --- 6.3°C 1.13 (0.89, 1.43) 0.32

Indoor Temperature 6.3°C 1.01 (0.75, 1.37) 0.93 Absolute Humidity 6.3°C 0.96 (0.72, 1.28) 0.79

Absolute Humidity --- 3.5 g/m3 1.33 (1.02, 1.74) 0.037 Outdoor Temperature 3.5 g/m3 1.37 (0.98, 1.90) 0.061 Indoor Temperature 3.5 g/m3 1.28 (0.96, 1.70) 0.093

IQR, interquartile range; CI, confidence interval. * All models were adjusted for the 24-hour moving average barometric pressure, 2-hour moving average particulate matter less than 2.5 μm in aerodynamic diameter (PM2.5) concentration, and matched on month, day of the week, and hour.

Nguyen J, Dockery D, et al

Effect of temperature on Heart Rate Variability and

Ren C, et al. Am J Epidemiol 2011;173(9):1013-21

Effects of temperature variability on QTc (ms)

14.0 12.0 10.0

8.0 6.0 4.0 2.0 0.0

-2.0

All

DM

Non

-DM

CH

D (P

-val

ue <

0.0

5)

Non

-CH

D

Obe

se (P

-val

ue <

0.0

5)

Non

-obe

se

Standard deviation of temperature, 4 d average

VA NAS, QTc (ms) in 547 subjects (891 obs)

Mehta A, Schwartz J, Zanobetti A, et al.

Effect of Meteorology on Lung Function

FVC A 1.0

0.5

0.0

-0.5

-1.0

-1.5

0.20

0.10

0.00

-0.10

-0.20

-0.30

-0.40

-0.50

FEV1

Cha

nge

in lu

ng fu

nctio

n (%

) C

hang

e in

lung

func

tion

(%)

Cha

nge

in lu

ng fu

nctio

n (%

) C

hang

e in

lung

func

tion

(%)

0 14h 2 3 4 5 6 Moving 4h 0 1 2 3 4 5 6 Moving average average

Lag in days Lag in days 0-2 days 0-6 days

Temperature Relative humidity

B 1.0

0.5

0.0

-0.5

-1.0

-1.5

0.20

0.10

0.00

-0.10

-0.20

-0.30

-0.40

-0.50 4h 0 1 2 3 4 5 6 Moving 4h 0 1 2 3 4 5 6 Moving

Lag in days average 0-2 days Lag in days

average 0-6 days

Temperature Relative humidity

Lepeule J, Schwartz J

~

~ Q) CL E 2 c Q) U) (IJ

~ u Q) -0 ()

~ (IJ

.2 0 .,,, ~ c (IJ Q)

E c 0 ·~

>,

-§ E ~ z 0

0

co 0

Effect of temperature on DNA methylation

Main model PM2.5 included 03 and particle number included

______ J _ _J __ __t_ ______________________________ _J __ J__J _____ _ t t t

~ '----~~~~~~~~~~~~~~~~~~~~~~~~~---' ~

F3 (1-week ma') CAM-1 (4-weeks ma') TLR-2 (4 weeks ma')

•ma moving average

w CL E 2 c Q) U) (IJ

~ ~ D

u ~ (IJ

2

l c 0 ·~

>,

~ E ~ 7 0

Q) 0) c (IJ .c u

0

Effect of temperature on DNA methylation

Main model PM2.5 included 03 and p:irti cle number included

__________ t ________ t _______ t---------------------------------------------

t t t IL-6 (2-weeks ma') iNOS (4-weeks ma')

•ma: moving avera;;ie

Effect of temperature on DNA Methylation

Bind MA, Schwartz J, et al

Decrease in Temperature and MI Incidence and ischemic stroke

3.0

2.5

2.0

1.5

1.0

0.5

0.0

cold

mon

ths

extre

me

5% c

old

war

m m

onth

s

extre

me

5% w

arm

All

year

sprin

g

sum fall

win

ter

Temp same day App temp 4 days average

Worcester Heart Attack Study, 4765 MI 1705 ischemic stroke

Madrigano J, Schwartz J, et al.Epidemiol, in press Mostofsky E, Mittleman M, et al. International Stroke conference 2013

Clustering days according to weather parameters 1

Mean (SD) 2

Mean (SD) 3

Mean (SD) 4

Mean (SD) 5

Mean (SD) # of days 1523 1308 1145 946 834 max T today (C) SD Hourly Temp (C) Water Vapor Pressure (mbar) Sea-Level Pressure (mbar) Ozone (ppb) Wind Speed (m/s) PM2.5 TEOM (ug/m3) Sulfur Dioxide (ppb)

6(5) 2(1) 5(2)

1022(6) 18 (9) 5(1)

10 (5) 7(4)

19 (4) 2(1)

14 (4) 1016(6)

22 (9) 4(1)

10 (5) 3(2)

8(6) 2(1) 6(3)

1007(7) 24 (9) 7(2) 7(4) 4(3)

29 (3) 3(1)

20 (4) 1013(5)

35 (12) 5(1)

16 (9) 3(2)

20 (6) 4(1) 9(3)

1017(7) 27 (11) 5(1)

10 (5) 4(3)

0 10

Color Ke and Hist

Cou

nt

Monthly

0 10 20 30 Value

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

1

Monthly Distribution 2

3

4

5

Austin E, Koutrakis P, Zanobetti A, Gold D, Schwartz J, et al

Citations • Zanobetti A, et al. Summer temperature variability and long-term survival among elderly people with chronic

disease. Proc Natl Acad Sci USA, 2012, 24;109(17):6608-13. • Hoffmann B, et al. Opposing Effects of Particle Pollution, Ozone and Ambient Temperature on Arterial Blood

Pressure. Environ Health Perspect, 2012;120(2):241-6. • Halonen J, et al. Outdoor temperature is associated with serum HDL and LDL. Environ Res, 2011;111(2):281-7. • Halonen J, et al. Relationship between outdoor temperature and blood pressure. Occup Environ Med.

2011;68(4):296-301 • Halonen J, et al. Associations between outdoor temperature and markers of inflammation: a cohort study.

Environ Health; 2010: 9:42. • Ren C, et al. Ambient Temperature, Air Pollution, and Heart Rate Variability in an Aging Population. Am J

Epidemiol 2011;173(9):1013-21. • Wilker E, et al. Ambient temperature and biomarkers of heart failure: a repeated measures analysis. Environ

Health Perspec 2012 Aug;120(8):1083-7. • Zanobetti A, et al. Effects of air pollution and temperature on vascular function in people with diabetes. Abstracts

ISEE. 2011, Barcelona, Spain. Environ Health Perspect • Zanobetti A et al. Effects Of Temperature And Water Vapor Pressure On Vascular Function In People With

Diabetes, ATS conference, Am J Respir Critic Care Med Vol 185. pp. A3203 • Mostofsky E et al. Short-Term Changes in Ambient Temperature and Risk of Ischemic Stroke. Abstract submitted

to the International Stroke conference 2013 • Zanobetti A, et al. Assessing Effect Modification By Zipcode Area Level Characteristic, Personal Characteristics

And Specific Cause Of Admission In A Multi-City Case-Only Analysis. Abstract to ATS conference, 2012 • Kloog I et al. Temporal and spatial assessments of minimum air temperature using satellite surface temperature

measurements in Massachusetts, USA. Sci Total Environ 2012 ;432:85-92. • Madrigano J et al. Temperature, Myocardial Infarction, and Mortality: Effect Modification by Individual and Area-

Level Characteristics, Epidemiol, in press

Acknowledgments • NIEHS R21 ES020695 • NIEHS R21 AG040027-01 • NIEHS R21 ES020194 • NIEHS PO1 ES-09825 • NIEHS P30 ES000002-49 • US EPA grant RD 83479801

• Our Climate Change R21s led to: – an RO1 proposal to evaluate indoor and outdoor

temperature/humidity and land use effects on cardiovascular and sleep outcomes in the longitudinal Jackson (Mississippi) Heart Study.

– An R21 proposal to identify the chronic effects associated with long­term exposure to fluctuations in temperature and absolute humidity,adaptation, and to investigate the health consequences of increasing variability of summertime temperatures with different future scenarios