1. “Split-Mouth” Clinical Trials Methodological Challenges in the Design and Analysis Introduction
Bruce L. Pihlstrom, DDS, MS
Bethesda, Maryland USA
Professor Emeritus, University of Minnesota
Adjunct Professor, Oregon Health Sciences University
Independent Oral Health Research Consultant

2. Presentation Outline History and Development
Advantages / Disadvantages
Limitations
Recommendations

3. Parallel Arm Trial Design Effect of NaFl rinse on caries
Such experimental designs are particularly well suited to farmers. Treatments can be laid out in strips, with length of the plots determined by the length of the field and the width by the equipment you use.
Parallel Arm Trial Design Effect of NaFl rinse on caries
0.05% NaFl Daily Rinse Vs
Placebo Daily Rinse
Randomized Intervention
Example
0.05% NaFl Daily Rinse
Randomize Subject
Randomize Subject Y X
Placebo Daily Rinse

4. History and Development
Split-Mouth Design
Derived from Split-Plot Agricultural Research ??
Evaluates multiple agricultural methods by dividing fields (sampling units) into main plots & sub-plots

5. Split Plot Trial Design Effect of two soil treatments on crop yield
Such experimental designs are particularly well suited to farmers. Treatments can be laid out in strips, with length of the plots determined by the length of the field and the width by the equipment you use.
Split Plot Trial Design Effect of two soil treatments on crop yield
Main plot factor:
Randomize Fertilizer
Fertilizer 1 = 20 lb Nitrogen / acre
Fertilizer 2 = 40 lb Nitrogen / acre
Plot
Randomize Main Treatment
(Fertilizer per Acre) X Y Y X Y X
Plot
Sub-plot factor:
Randomize Tillage
to subplots
X = Conservative Tillage
Y = Intensive Tillage

6. Analogous Split Mouth Trial Design Effect of two treatments on caries
Such experimental designs are particularly well suited to farmers. Treatments can be laid out in strips, with length of the plots determined by the length of the field and the width by the equipment you use.
Analogous Split Mouth Trial Design Effect of two treatments on caries
Main subject factor:
Randomize fluoride use
NaFl: 0.05% daily rinse
NaFl: 0.24% daily toothpaste
Subject X Y
Randomize Main Treatment
(Fluoride Rinse)
Subject Y X Y X
Sub-subject factor:
Randomize sealants
to dentition segments
X = Sealants Placed
Y = No Sealants Placed

7. Common Split Mouth Trial Design Effect of one treatment on caries
Such experimental designs are particularly well suited to farmers. Treatments can be laid out in strips, with length of the plots determined by the length of the field and the width by the equipment you use.
Common Split Mouth Trial Design Effect of one treatment on caries
Randomized Intervention
X = Sealants Placed
Y = No Sealants Placed
Randomize intervention to dentition segments (½ mouths) in each subject
X Y
Randomize intervention to dentition segments (½ mouths) in each subject
Y X

8. History of Split Mouth Trials
Introduced in 1968 by Ramfjord, Nissle, Shick, et al.1
Randomly assigned subgingival curettage or surgical pocket elimination to ½ mouths divided by mid-sagittal plane between central incisors
Later, definition enlarged by others to include randomly assigned treatments to quadrants / sextants for other dental diseases
11 different variants found in periodontal literature by Hujoel and Loesche in 19902
J Periodontol 1968;39:167
J Clin Periodontol 1990;17:722

9. Potential Advantage of Split Mouth Trials*
Achieves increased efficiency by:
Reduced inter-subject variability (error variance) and therefore:
Increase power
Increase precision of estimated treatment effect
*Hujoel P. and DeRouen, J Clin Periodontol 1992;19:

10. Example of a Small Randomized Clinical Trial Surgical vs
Example of a Small Randomized Clinical Trial Surgical vs. Non-surgical Periodontal Therapy J Periodontol May;52(5):

13. Hujoel & Loesche* Efficiency of Split Mouth Designs
1. Baseline data from 69 patients with periodontal disease investigated for disease similarity and suitability for 11 different split mouth designs based on disease:
Amount (mean PD & CAL)
Distribution (variance of PD & CAL)
Severity (presence of PD 4-6mm; > 6 mm)
Subset of 38 patients received scaling and root planing in all within-subject experimental units
Re-evaluated at 6 weeks to determine relative efficiency of 11 different split mouth designs
* J Periodontol 1990; 17:

14. Hujoel & Loesche* Homogeneity Within Experimental Units
Split Mouth Design
Within-Patient Experimental Units
Disease Amount
Pocket Depth Mean
Differences
(p-values)
Clinical Att. Loss
Mean Differences
Ipsi-lateral Quadrants 2
0.08
0.89
Ipsi-lateral Sextants
0.03
0.25
Contra-lateral Quadrants
0.80
0.33
Contra-lateral Sextants
0.76
0.45
Quadrants 4
0.0001
0.23
Posterior Sextants
0.1
Sextants 6
* J Periodontol 1990; 17:

15. Hujoel & Loesche* Variability within Experimental Units
Split Mouth Design
Within-Patient Experimental Units
Probing Depth Variance
(p-values)
Clinical Attachment Level Variance
Ipsi-lateral Quadrants 2
0.11
0.19
Ipsi-lateral Sextants
0.03
0.006
Contra-lateral Quadrants
0.83
0.01
Contra-lateral Sextants
0.98
0.26
Quadrants 4
0.02
0.001
Posterior Sextants
0.04
0.0007
Sextants 6
0.0001
0.002
* J Periodontol 1990; 17:

16. Hujoel & Loesche* Distribution & Asymmetry Within Experimental Units
Split Mouth Design
Patients with asymmetric probing depths 4-6 mm (%)
Patients with asymmetric probing depths > 6mm (%)
Ipsi-lateral Quadrants
0 %
17%
Ipsi-lateral Sextants 19
Contra-lateral Quadrants 15
Contra-lateral Sextants 16
Quadrants 55
Posterior Sextants 9 68
Sextants 14 83
* J Periodontol 1990; 17:

17. Hujoel & Loesche* Relative Efficiencies of Split Mouth Designs
Efficiency Factor Gain
4-6 mm probing depths
Efficiency Factor Gain
> 6 mm probing depths
Ipsi-lateral Quadrants (2)
3.2
0.8
Ipsi-lateral Sextants (2)
1.9
Contra-lateral Quadrants (2)
4.8
1.7
Contra-lateral Sextants (2)
1.0
Quadrants (4)
2.0
Posterior Sextants (4)
0.9
0.7
Sextants (6)
Additional Patients needed for whole mouth design = Split-mouth design x Efficiency Factor Gain
* Ratios of full mouth vs. split mouth error variance J Periodontol 1990; 17:

18. Potential Disadvantages of Split Mouth Trials
“Carry-cross” from one randomized segment to another may introduce unknown bias 1
“Carry-cross” effect precludes use of this design for equivalence trials because of potential contamination across randomized segments 2,3
Screening and recruitment may need to be increased because of need for symmetry in randomized segments; e.g. especially for trials with very restrictive inclusion criteria1
1. Hujoel PP. Community Dent Oral Epidemiol 1998;26:85-6
2. Imrey P, Chilton N. J Periodontol 1992 (Suppl);
3. Fleiss JL. J Periodont Res 1992;27:

19. Potential Disadvantages of Split Mouth Trials
More complex statistical analysis is needed to consider paired design and correlations within subjects across randomized segments1
Depending on magnitude of within-patient correlation, may not gain efficiency; e.g. If within-patient correlation of treatment response is low, little efficiency is gained1
1. Hujoel PP. Community Dent Oral Epidemiol 1998;26:85-6

20. Split Mouth Trials - Recommendations
Always consult with biostatistician who has experience in clinical trials during design phase
Facilitate statistical analysis by not using complex designs with multiple levels of sub-patient interventions and outcomes
Facilitate screening and recruitment by simplifying inclusion / exclusion criteria
NOTE: Always a good idea for generalizability and ease of recruitment…..
BUT simplification of inclusion criteria especially important for split mouth trials where symmetry of disease in randomized segments is essential to gain (or maintain) efficiency

21. Split Mouth Trials - Recommendations
Avoid using split mouth trials for equivalency trials
Always consider possible “carry-cross” effect of intervention
Never use split mouth design if carry-cross effect is likely (possible?)
Make sure that efficiency is really being improved by screening patients for comparable amount, severity, distribution of disease within patient mouths