New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

allometric scaling in pksim #120

Closed
elmokadem opened this Issue Jan 9, 2018 · 10 comments

Comments

Projects
None yet
4 participants
@elmokadem

elmokadem commented Jan 9, 2018

Hi,

I was curious about how the scaling of a typical individual physiologic parameters is done in pksim. I traced back the literature to an excellent paper by Willmann et al in 2007 that states that allometric scaling is being adopted. For example, scaling of organ weights (OW) follows:

OW_target = (Height_target/Height_mean)^0.75 * OW_mean

From what I understood, the Height_mean and OW_mean belong to a mean adult individual. So for example, the liver weight for a 5 year old with a height of ~109 cm would be:
OW_target = (109/176)^0.75 * 2.35
with 176 and 2.35 being the adult mean height and liver weight, respectively.

The thing is that when I apply this equation, the organ weights I get are way off the values mentioned in the measured ICRP database. Also, the results I get from the equation are very different than the results I get from the generated individuals in pksim. Furthermore, the generated results from pksim agree well with the ICRP data for some parameters but differ greatly for others.

So, I guess the short version of my question is: how is the scaling done in pksim?

Thanks and regards,
Ahmed.

@msevestre msevestre added the question label Jan 10, 2018

@StephanSchaller

This comment has been minimized.

Show comment
Hide comment
@StephanSchaller

StephanSchaller Jan 23, 2018

Member

@JanSchlender

Or Meanwhile you will find some answers in #87

Member

StephanSchaller commented Jan 23, 2018

@JanSchlender

Or Meanwhile you will find some answers in #87

@JanSchlender

This comment has been minimized.

Show comment
Hide comment
@JanSchlender

JanSchlender Jan 26, 2018

Hi Ahmed,

Well observed. You may also read Edginton et al 2006 (https://www.ncbi.nlm.nih.gov/pubmed/16984214). The ICRP population database is well explained in this publication.
Best regards,

Jan

JanSchlender commented Jan 26, 2018

Hi Ahmed,

Well observed. You may also read Edginton et al 2006 (https://www.ncbi.nlm.nih.gov/pubmed/16984214). The ICRP population database is well explained in this publication.
Best regards,

Jan

@msevestre

This comment has been minimized.

Show comment
Hide comment
@msevestre

msevestre Jan 26, 2018

Member

@elmokadem

You can also check the wiki page here describing how the Create Individual Algorithm is implemented in details. This should complement the above mentioned publication nicely

Cheers,

Member

msevestre commented Jan 26, 2018

@elmokadem

You can also check the wiki page here describing how the Create Individual Algorithm is implemented in details. This should complement the above mentioned publication nicely

Cheers,

@elmokadem

This comment has been minimized.

Show comment
Hide comment
@elmokadem

elmokadem Jan 26, 2018

Thanks Stephan, Jan and Michael for your replies. The Edginton paper was very helpful in explaining some of the discrepancies between organ volumes/weights reported in ICRP and those calculated by PKSim. Nevertheless, it still doesn't explain why a scaling method based on weight or BSA wasn't adopted instead of a clearly erroneous height scaling method.
The wiki page was also very useful as it confirmed that the scaling with height is indeed the method used to calculate the mean organ volumes and this is where I am having a problem.
Take liver weight as an example. If we use the formula mentioned in my initial comment (and in the wiki page) to calculate the liver weights for a range of individuals ranging in heights from 51 to 176 cm, the equation will be:
Li_wt = 1.8*(HT/176)^0.75
This will result in liver weights that are so different from the ICRP reported weights as shown in the figure below. This means that when we use these scaled liver weights as means to sample random weights that are normally distributed, they will spread around the line shown in the plot and will not reflect a distribution close to the ICRP datapoints. On the other hand, if we simply use linear scaling with weight, the results now agree very well with the ICRP data.

li_scaling

elmokadem commented Jan 26, 2018

Thanks Stephan, Jan and Michael for your replies. The Edginton paper was very helpful in explaining some of the discrepancies between organ volumes/weights reported in ICRP and those calculated by PKSim. Nevertheless, it still doesn't explain why a scaling method based on weight or BSA wasn't adopted instead of a clearly erroneous height scaling method.
The wiki page was also very useful as it confirmed that the scaling with height is indeed the method used to calculate the mean organ volumes and this is where I am having a problem.
Take liver weight as an example. If we use the formula mentioned in my initial comment (and in the wiki page) to calculate the liver weights for a range of individuals ranging in heights from 51 to 176 cm, the equation will be:
Li_wt = 1.8*(HT/176)^0.75
This will result in liver weights that are so different from the ICRP reported weights as shown in the figure below. This means that when we use these scaled liver weights as means to sample random weights that are normally distributed, they will spread around the line shown in the plot and will not reflect a distribution close to the ICRP datapoints. On the other hand, if we simply use linear scaling with weight, the results now agree very well with the ICRP data.

li_scaling

@msevestre

This comment has been minimized.

Show comment
Hide comment
@msevestre

msevestre Jan 26, 2018

Member

@elmokadem

The formula Li_wt = 1.8*(HT/176)^0.75 that you used your calculation is not accurate.

PK-Sim is not using 1.8 and 176 all the time (Mean Liver weight and mean height for adult 30 years old ICRP).

Instead, PK-Sim used the mean liver weight and mean height for a randomly generated input age.

In the case of a population, here is what's happening:

  • A new age between [Age min and Age max] is generated
  • For this VERY age, mean_liver and mean_height is calculated based on ICRP data for that age
  • Only then is a scaling performed for the randomly generated target weight (calculated for the height distribution for the previously generated)

That means that the scaling with height will be pretty accurate and very close to the ICRP data by construction

Hope this clarifies things,

Member

msevestre commented Jan 26, 2018

@elmokadem

The formula Li_wt = 1.8*(HT/176)^0.75 that you used your calculation is not accurate.

PK-Sim is not using 1.8 and 176 all the time (Mean Liver weight and mean height for adult 30 years old ICRP).

Instead, PK-Sim used the mean liver weight and mean height for a randomly generated input age.

In the case of a population, here is what's happening:

  • A new age between [Age min and Age max] is generated
  • For this VERY age, mean_liver and mean_height is calculated based on ICRP data for that age
  • Only then is a scaling performed for the randomly generated target weight (calculated for the height distribution for the previously generated)

That means that the scaling with height will be pretty accurate and very close to the ICRP data by construction

Hope this clarifies things,

@elmokadem

This comment has been minimized.

Show comment
Hide comment
@elmokadem

elmokadem Jan 29, 2018

Thanks a lot Michael for the clarification. I guess my last question here is: because ICRP lacks the organ specific data (volumes and flows) for every possible age, what kind of extrapolation do you do for covering these gaps? In the Edgington paper mentioned in Jan's comment, it is mentioned that cubic splines were used for some parameters as blood volume and heights, so is the same extrapolation used for organ volumes?

elmokadem commented Jan 29, 2018

Thanks a lot Michael for the clarification. I guess my last question here is: because ICRP lacks the organ specific data (volumes and flows) for every possible age, what kind of extrapolation do you do for covering these gaps? In the Edgington paper mentioned in Jan's comment, it is mentioned that cubic splines were used for some parameters as blood volume and heights, so is the same extrapolation used for organ volumes?

@msevestre msevestre added the answer label Jan 29, 2018

@msevestre

This comment has been minimized.

Show comment
Hide comment
@msevestre

msevestre Jan 29, 2018

Member

@elmokadem PK-Sim does not use cubic splines interpolation anymore.
At the moment, a simple linear interpolation is used between available ages and for all parameters

Member

msevestre commented Jan 29, 2018

@elmokadem PK-Sim does not use cubic splines interpolation anymore.
At the moment, a simple linear interpolation is used between available ages and for all parameters

@elmokadem

This comment has been minimized.

Show comment
Hide comment
@elmokadem

elmokadem Jan 29, 2018

Thanks a lot Michael.. That was really helpful.

elmokadem commented Jan 29, 2018

Thanks a lot Michael.. That was really helpful.

@msevestre

This comment has been minimized.

Show comment
Hide comment
@msevestre

msevestre Jan 29, 2018

Member

@elmokadem Awesome. Glad I could help
FYI: We will release in the next few weeks more implementation details in the wiki. (Population Algorithm, Aging Individual etc..). Stay tuned.

Member

msevestre commented Jan 29, 2018

@elmokadem Awesome. Glad I could help
FYI: We will release in the next few weeks more implementation details in the wiki. (Population Algorithm, Aging Individual etc..). Stay tuned.

@elmokadem

This comment has been minimized.

Show comment
Hide comment
@elmokadem

elmokadem Jan 29, 2018

Looking forward to it :)

elmokadem commented Jan 29, 2018

Looking forward to it :)

@msevestre msevestre closed this Feb 1, 2018

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment