CORRELATIONS IN POLLUTANTS AND TOXICITIES

1
CORRELATIONS IN POLLUTANTS AND TOXICITIES

• Kovanic P. and Ocelka T.
• The Institute of Public Health, Ostrava, Czech
  Republic

2
DATA

• Actions Regular monitoring of Czech and Moravian
  rivers
• Period 2002 2007
• Profiles 21 locations of rivers Becva, Berounka,
  Bílina, Dyje, Jihlava, Jizera, Labe, Lužická
  Nisa, Lužnice, Morava, Odra, Ohre, Opava, Otava,
  Sázava, Svratka, Vltava.
• Field activity Institute of Public Health,
  Ostrava (IPH)
• (The National Reference Laboratory)
• Chemical analyses Laboratories of the IPH,
  Frýdek-Místek
• Mathematical (Gnostic) analysis IPH
• Particular problem
• Are there any interactions
• between pollutants?

3
NATURAL ASSUMPTIONS ?

• Contaminations are generated, polluted and
  accumulated mostly simultaneously, hence the
  more contaminants, the higher contamination and
  opposite.
• The more pollutant A, the more polutant B.
• Positive and significant interactions
  between
• concentrations of pollutants are expected.
• IS IT TRUE?

4
(No Transcript)
5
(No Transcript)
6
COMMENTS

• Concentrations of groups of pollutants
• differ by orders of magnitude.
• Distributions differ not only by mean
• levels but also by their forms.
• Distributions are non-Gaussian (normal)
• domains are finite, densities asymmetric.
• Data variability is strong, robust analysis
• must be applied.

7
TWO APPROACHES TO INTERACTIONS

• Robust correlation coefficients
• interdependence of deviations from the mean
  value
• Robust regression models
• interdependence of variables
• The former does not imply
• the latter automatically !

8
ROBUST CORRELATIONS

• Robust estimate low sensitivity
• to bad data
• Non-robust estimates point statistics
• (sample estimates of statistical moments)
• Many robust estimates exist producing
• different results

9
(No Transcript)
10
DIVERSITY OF ESTIMATES

• In the past lack of robust methods
• Recently abundance of robust methods
• Diversity of results
• IN WHICH METHOD TO BELIEVE?

11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
INFERENCE

• Significant interactions between groups
• of pollutants have been confirmed.
• Assumption of positive interactions was
• falsified there exist negative interactions.
• Group HCH initiates negative effects.
• Interactions of groups implie interactions
  between individual congeners.
• Which congeners interact negatively
• and how much?

15
(No Transcript)
16
DEPENDENCE OF POLLUTANTS (Y) ON THE GAMMAHCH
(X)

• Title L(Y)L(X) is to be read as natural
  logarithm of the pollutant Y presented as a
  linear function of the natural logarithm of the
  pollutant X (gammaHCH)
• GRAPHS
• Straight line is the robust linear model.
• Points depict the data values (X, Y)
• NOTE Vertical scalings (of Y) differ, the
  horizontal scale (of X) remains unchanged

17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
PARAMETERS OF THE FUNCTIONlognat(Y) Intercept
Coef lognat(gammaHCH)

• Pollutant (Y) Intercept STD(Intct) Coeff.
  STD(Coef)
• OCDD -1.01 0.23 -0.376
  0.040
• TCDD 0.08 0.34
  -0.424 0.057
• PeCDD -0.94 0.28 -0.454
  0.048
• HxCDD -2.58 0.16 -0.082
  0.027
• HpCDD -1.53 0.27 -0.312
  0.046
• OCDF -1.55 0.28
  -0.344 0.047
• TCDF 2.01 0.35
  -0.446 0.058
• PeCDF 0.36 0.33 -0.319
  0.055
• HxCDF -1.58 0.27
  -0.249 0.045
• HpCDF -2.07 0.25 -0.184
  0.041

25
PARAMETERS FOR THE HCH AND PBDE

• Pollutant (Y) Intercept STD(Intct) Coeff.
  STD(Coef)
• alfaHCH 2.14 0.38 0.373
  0.064
• betaHCH -2.39 0.57 1.054
  0.096
• deltaHCH -3.47 0.52 1.057
  0.089
• HCB 9.15 0.61
  -0.666 0.102
• PBDE28 9.66 0.73
  -1.603 0.123
• PBDE47 15.30 0.69
  -2.019 0.116
• PBDE100 11.40 0.56 -1.698
  0.095
• PBDE99 13.39 0.71 -1.826
  0.119
• PBDE154 11.42 0.77 -2.004
  0.130
• PBDE153 9.92 0.71
  -1.700 0.119
• PBDE183 2.77 0.64
  -0.534 0.105

26
RELATIVE IMPACTS OF gammaHCHImpact/mean(pollut.co
ncentr.)(How many times is the mean exceeded)
Pollutant Rel.Impact Pollutant Rel.Impact
OCDD -0.88 alfaHCH 0.49
TCDD -0.89 betaHCH 2.27
PeCDD -0.88 deltaHCH 2.28
HxCDD -0.21 HCB -0.63
HpCDD -0.68 PBDE28 -2.22
OCDF -0.53 PBDE47 -3.20
TCDF -0.75 PBDE100 -2.82
PeCDF HxCDF -0.57 -0.33 PBDE99 PBDE154 -3.06 -3.33
HpCDF -0.31 PBDE153 -2.37
PBDE183 -0.20
27
POLLUTANTS TOXICITY

• Four methods to measure toxicity
• Daphnia Magna
• Vibrio Fischeri
• Desmodemus subspicatus
• Saprobita

28
(No Transcript)
29
NATURAL ASSUMPTIONS

• A) Methods measuring the same give the same
  results or
• B) Results of measuring the same are at
• least similar (correlated)
• C) The more pollutants concentration, the more
  toxic effects

30
SIGNIFICANT CORRELATIONS WITH TOXICITIES
Correlation Cor. Coef. Prob0
(Vibrio F., Desm. Subsp.) 0.524 0.022
(sumPAH, Desm. Subsp.) 0.643 0.010
(sumDDT, Daphnia Magna) 0.501 0.035
Other correlations are not significant. Natural
assumptions A) through C) are not supported by
the data. Let us try the MD-models !
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
WORTHWHILE

• MD-models confirm the existence of contrary
  toxic effects.
• The group PCB affects the toxicity contrary to
  other groups of pollutants in 3 of 4 MD-models in
  spite of the positiveness of all correlations
• (pollutant, toxicity).

36
SUMMARY

• Statistically significant (mostly positive)
  correlations in organic pollutants exist.
• Negative correlations exist as well.
• The most negatively active is gammaHCH.
• Its strongest negative effects are manifested by
  the congeners of PBDE.
• Contrary toxicity impacts of pollutants exist.
• HYPOTHESES MUST BE TESTED !

37
OPEN PROBLEMS

• Are these effects caused by some real chemical or
  physical reactions of the substances or only by
  different rates of their production and
  pollution?
• Are they worth of further investigation?
• EXPERIENCE
• DATA TREATMENT MUST BE ROBUST
• AND HYPOTHESES MUST BE TESTED !

38
FUNDING

• European Commission Sixth Framework Program,
  Priority 6 (Global change and ecosystems),
  project 2-FUN (contract036976)