270 likes | 291 Views
Low probability and great potential damage – statistical comments on biomonitoring Carsten Lüring Kiel - 2010. Why use Biomonitors?. Fast and online detection of toxic substances Detection of known as well as unknown or unexpected toxic substances
E N D
Low probability and great potential damage – statistical comments onbiomonitoringCarsten LüringKiel - 2010
Why use Biomonitors? • Fast and online detection of toxic substances • Detection of known as well as unknown or unexpected toxic substances • Massive reduction of efforts for chemical analysis • Detection of rare events • Prevention of high damage
Statistical proof of God’s existenceBlaise Pascale (1623 - 1662)
The statistics of biomonitoring (2010) -> consider the use of biomonitoring
Organisms and Biomonitors • daphnids • algae • luminescent bacteria • mussels • fish
Performance of a water monitoring instrument: • Question: • What is the probability of a toxic event if an alarm is indicated? • 1. ca. 99,9% • 2. ca. 98% • 3. ca. 5% • 4. ca. 2% • 5. I don’t know • 6. Nobody knows
Performance of a medical test: You are travelling to a tropical country. After returning, you notice that some travellers - let´s say 1 of 1000 - are infected by a painful sickness. To check whether you are infected or not, a diagnostic test with the following performance is conducted: Test result: infected How high is the probability of infection?
Performance of a medical test: • Your personal test result: infected • How high is the probability of real infection? • 1. ca. 99% • 2. ca. 98% • 3. ca. 5% • 4. ca. 2% • 5. I don’t know • 6. Nobody knows
Result of the medical test: • Your personal test result: infected • How high is the probability of real infection? • 1. ca. 99% • 2. ca. 98% • 3. ca. 5% • 4. ca. 2% • 5. I don’t know • 6. Nobody knows
Result of the medical test: Example: Number of travellers: Number of not infected travellers Number of infected travellers Number of not infected travellers with a false positive test Number of infected travellers with a positive test Number of infected travellers with a false negative test 100,000 99,900 100 2,000 99 1 Probability: 99 of 2,099 -> ca. 5%
Consequence: Redo the test • Your personal second test result: infected • How high is the probability of real infection? • 1. ca. 99% • 2. ca. 98% • 3. ca. 70% • 4. ca. 35% • 4. ca. 15% • 5. I don’t know
Consequence: Redo the test Example: Number of travellers with a positive test: Number of not infected travellers with a positive test: Number of infected travellers with a positive test: Number of not infected travellers with a second false positive test Number of infected travellers with a second positive test Number of infected travellers with a false negative test 2,099 2,000 99 40 98 1 Probability: 98 of 138 -> ca. 70%
Performance of a test: • To answer the questions of the probability of being infected, 3 pieces of information are required: • Error of the test showing an infection although there is none • Error of the test showing no infection although there is an infection • A priori probability of the population
Performance of a water monitoring instrument: • Question: • What is the probability of a toxic event if an alarm is indicated? • 1. ca. 99.9% • 2. ca. 98% • 3. ca. 5% • 4. ca. 2% • 5. I don’t know • 6. Nobody knows
Performance of 10 biomonitors in a drinking water supply system:
Consequences for biomonitoring: • Double check in case of an alarm • Reduce the number of alarms although there is no toxin by enhancing the instrument
Realization in bbe biomonitoring instruments: • Reduce the number of alarms although there is no alarm by enhancing the instrument by: • using special alarm detection algorithms • using more than on parameter to detect an alarm • - DaphTox / FTOX: 2-3 of the behavioural parameters need to be suspicious • - ToxProtect: Activity and “swimming in the top row” • supervising the hardware for illegal states • - Temperature • - Flowrate • - Function of heating / cooling devices • - Status of the organisms
Double check in case of an alarm: Using 2 instruments of the same type BioMonitor A1 BioMonitor A2 -> expensive
Not usable for a double checking BioMonitor A BioMonitor B Double check in case of an alarm: Using 2 instruments with different organisms Detecting a wider range of toxins BioMonitor A BioMonitor B
Realization in bbe biomonitoring instruments: • Double check in case of an alarm: • DaphTox / FTOX: Two chamber systems with parallel alarm evaluation • ToxProtect: Alarm evaluation by switching off the light • Algae Toximeter: Redo the measurement 2 or 3 times
The bbe Daphnia Toximeter verification mechanism - current subjects of research - • Using the phototaxis of the daphnia for a second test • Additional alternating blue light on the left and right side in case of alarm • Horizontal distribution of daphnia by adding 5µg/l Pyrethrum
Take home message: • The lower the probability of a toxic event - • the more important is the use of a biomonitor with an automatic double check of the alarm.
Special thanks to: Dr. Hans-Hermann Dubben: Der Hund, der Eier legt