I analysed my wound healing closure data with ANOVA but I had more than 2 groups. In your case you should use T test. I had already use T test in other wound healing experiments when having 2 groups. 

Thank you Tiago. I am already using T test for overall reduction in surface area from baseline to final measurement. Wouldn't multiple T-tests at each point in time be inappropriate?

Would a linear regression or linear mixed model be more suitable?

Also did you have any thoughts regarding time taken to achieve resolution of infection? 

Dear Meg, Unfortunately I did not perform infection in my wound healing experiments so I cant tell you. Mine was about treatment with pharmacological blockers to see improvement in wound closure only. 

Well as I said I had more than 2 groups( 1 controle plus 3 or 4 treatments ) plus several time points. And in my experience I had used anova. 

What I had was percentages of the wound closure in each day so anyway, I am not sure how use linear regression with this data.

Thank you very much Tiago.

Hi  You may not have been exposed to this yet in your undergrad stat courses, but the ideal  way would probably be a hierarchical model where you treat responses at the two time points as nested within the persons themselves. everyone has their own intercept. This is especially good if you have missing data at either of the time points.  But people have been addressing problems like this  for eons using repeated measures ANOVA

Good luck!

BF

Thank you for your advice William. Would this model work for more than 2 time points as my data will be collected at baseline then weekly thereafter (for 4 weeks). Also would this model work for between-subjects analysis to find significance between the two groups? 

Before you use t-tests check your data distribution. It is quite likely that it is not normally distributed - in which case you would need to use non-parametric methods or correct the distribution (advanced stats and probably not for an undergrad class).

the level of significance for any multiple testing needs to be corrected because the more you test the more you are likely to find a 'significant' difference.

We have found the best fit for the clinical data to be a personalized mixed-effects exponential model (see Xu et al, J Tissue Viability 2014)

Thank you both Chris and Kath, your comments have been incredibly helpful. Much appreciated :)

Measuring infection is the trickier part, and one I had to deal with in my original dissertation study design (it's since changed completely, due to other reasons).  We don't rely on cultures to diagnose infection anymore as that just gives you a number for different bacteria present in the wound.  There are two problems with using this diagnostically.  First, some bacteria that may be causing problems just won't grow in our standard culture techniques, so we can be missing some virulent pathogens.  Secondly, the host's resources are critical in determining if that bioburden will result in an infection or not.  A stronger immune system in a healthier person can keep more bacteria in check than a frail person with a weakened immune system can.  The latter may have signs of infection with fewer bacteria present, so a CFU count won't be relevant.

Instead, look at a clinical assessment of signs & symptoms approach.  Gary Sibbald and Kevin Woo have developed their NERDS & STONEES approach for recognizing superficial infections/critical colonization and deep wound infections.  Keith Cutting has published some papers using clinical signs as suggestive of infection.  He has also distinguished which signs are relevant for which type of wound.

 Thank you Renee, I will definitely look into this in more depth. I was planning to have some form of bacteriological report in regards to the specific bacterial groups present in each treatment arm as I am aware that the number of pathogens is not sufficient to indicate antibacterial efficacy, but in light of what you have pointed out I will definitely include s/s of infection in my outcome measures. One thing I am concerned about however is the fact  that infection s/s may be absent/masked in my target population...would it be plausible to suggest that eradication of infection would be based on clinical observation and supporting culture results? 

Many thanks

While the classic signs of infection resulting from an immune response, such as purulent exudate, may be absent, other signs may still be present from the destructive nature of the bacteria.   NERDS/STONEES and Cutting's criteria consider this, and area appropriate for diabetic ulcers (Cutting has a a set for this wound type in particular too).  Quantitative bacteriologic measures may be helpful in showing volume, if relevant to what you're studying, but is not sufficient for determining infection, so a combination of the two may be valuable.  

On that note, are you doing swab cultures or quantitative tissue cultures?

Hi 

Kath is correct. If you use a mixed effect modell where each individuaj has shi/her own intercept, you can model the outcome that way > each unit has an intercept and a slope. You only need two points (the intercept and one other) to estimate a slope, you can have as many as you want. TYou can have missing or delayed observations as well  A great treatment of this approach is provided in Singer & Willett's book "Applied Longitudinal Data analysis."

as for "time to infect," the K-M approach, stratified by treatment group, works fine.  I doubt you have censored data, but that won't matter.

Good luck.

 Thanks Renee. I am doing swabs; I figured this would be more applicable to routine practice as, in my experience, we only ever do swabs. 

William, thank you for your advice, I will take a look at the book. I really appreciate your help.