Bioequivalence and Bioavailability Forum

Lin-up/log-down trapezoidal avoids most trouble [NCA / SHAM]

posted by Helmut  – Vienna, Austria, 2013-05-20 16:22 (3990 d 22:40 ago) – Posting: # 10604
Views: 17,837

Hi Gracehung,

please follow the Forum’s Policy in the future. THX.

❝ As in BE study, One sample’s actual sampling time was miss recorded. However as according to protocol, this sample was analyzed by our analytical laboratory.

Perform an audit. Why was an existing sample recorded as missing?

❝ 1. it was not specified in protocol…

Don’t worry, it’s too late.

❝ … how would you handling this missed time point at statistical analysis?

In the future describe how you would handle missing samples in an SOP or – much better IMHO – in the protocol. In the latter case any procedure is more transparent to an assessor (i.e., approved by the IEC and the local authority).

❝ 2. Will this data be included in statistical analysis? If yes, what is the handling procedure?

Depends on the intellectual horsepower you are willing/able to invest. Missings are generally not problematic if unambiguously located in the absorption¹ or distribution/elimination² phase. If the missing might be C_max³ data imputation is problematic. Regulators don’t like PK modeling, so I have tried regressive splines with not very convincing results (references there).

❝ 3. If the data is excluded, what kind of method would you use in calculate your AUC?

❝ a. Would you adapt interpolation method?

❝ b. If yes, would you adapt logarithmic interpolation or linear interpolation?

I always suggest to use the lin-up/log-down trapezoidal method. I think that Pharsight will make it even the default in the next version of Phoenix/WinNonlin. No need to impute data. Fine unless the missing might be C_max… See also this presentation about different integration algos (especially slides 16–17 on what happens with missing values).

❝ 4. Is there any guideline or reference document we could follow?

Not that I recall.

---

Let’s denote denote the missing as C_i. Then if

1. C_i+1 > C_i-1 ~ absorption
   
2. C_i+1 < C_i-1 ~ distribution/elimination
   
3. C_i+1 = C_i-1 ~ C_max or pure chance? Any algo will come up with the linear interpolation C_i = C_i-1 + |(t_i – t_i-1)/(t_i+1 – t_i-1)|(C_i+1 – C_i-1) – which likely underestimates the true value.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна! 
Helmut Schütz


The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

• handling of missing data gracehung 2013-05-20 07:35 
  • Lin-up/log-down trapezoidal avoids most troubleHelmut 2013-05-20 14:22
    • Lin-up/log-down trapezoidal avoids most trouble yjlee168 2013-05-22 08:25
      • Multiple peaks: fallback to linear trapezoidal Helmut 2013-05-22 20:25
        • Multiple peaks: fallback to linear trapezoidal yjlee168 2013-05-22 21:57
          • No interpolation Helmut 2013-05-22 23:25
            • No interpolation yjlee168 2013-05-23 11:43
              • Spaghetti & other pasta Helmut 2013-05-23 15:15
                • Spaghetti & other pasta yjlee168 2013-05-23 20:47
                  • Spaghetti & other pasta Helmut 2013-05-23 21:29
                    • Spaghetti & other pasta yjlee168 2013-05-23 22:46
                      • Spaghetti & other pasta Helmut 2013-05-24 01:24
                        • Spaghetti & other pasta yjlee168 2013-05-24 09:17
            • No interpolation Ken Peh 2013-05-30 19:55
              • Different algos! Helmut 2013-05-30 22:38
                • Different algos! Ken Peh 2013-06-03 17:52
                  • Calculate what? Helmut 2013-06-03 18:11
      • Lin-up/log-down trapezoidal example Helmut 2013-05-25 15:09
        • Lin-up/log-down trapezoidal example yjlee168 2013-05-25 19:45
  • handling of missing data Ohlbe 2013-05-20 21:45
    • Sorry Helmut 2013-05-21 13:53
      • Sorry gracehung 2013-05-23 01:06
        • Uncertain time point Helmut 2013-05-23 01:35