First, it provides a strong argument against the opinion that
subgroup analyses of randomised trials should be strongly discouraged because they can lead to false positive findings due to the multiple comparison problem.
In
our subgroup analysis of African American infants, at 12 weeks, African American mothers with peer counselors had odds of providing any breast milk 249 % greater than those for African American mothers without peer counselors (odds ratio, 3.59 [95 % confidence interval, 1.16 - 11.03]; P =.03)(Table 3).
A subgroup analysis of the women who did not take their own calcium supplements revealed that supplementing as part of the trial did reduce hip fracture risk by 30 percent.
This was the conclusion, for instance, of a 2008 article by Chris Gardner and his colleagues at Stanford, reporting on
a subgroup analysis of their famous A to Z study.
A subgroup analysis of high risk women who were unmarried and from low SES households (40 %) showed that home visits reduced the number of subsequent births (mean difference [MD] 0.5, p = 0.02), months that women received welfare (MD 29.9, p = 0.005), reports of behavioural impairment due to substance abuse (incidence 0.41 v 0.73, p = 0.005), records of arrests (incidence 0.16 v 0.90, p < 0.001), convictions (incidence 0.13 v 0.69, p < 0.001), and verified reports of child abuse and neglect involving the mother as perpetrator (incidence 0.11 v 0.53, p < 0.01).
Not exact matches
We conducted an a priori
subgroup analysis comparing risk
of bias, results and conclusions
of reviews with different industry sponsors (for example, cane and beet sugar industry sponsored reviews vs artificial sweetener industry sponsored reviews).
In the
subgroup analysis stratified by parity, there was an increased incidence
of the primary outcome for nulliparous women in the planned home birth group (weighted incidence 9.3 per 1000 births, 95 % confidence interval 6.5 to 13.1) compared with the obstetric unit group (weighted incidence 5.3, 3.9 to 7.3).
We investigated substantial heterogeneity for all four outcomes with
subgroup analyses for the following covariates: who delivered care, type
of support, timing
of support, background breastfeeding rate and number
of postnatal contacts.
Planned
subgroup analyses involved comparisons between different approaches to delivery for four key characteristics: group or individual or self - administered, length
of programme (same or different), index child involvement or adjunctive treatment.
Due to heterogeneity and to the small number
of trials in each
subgroup, we would advise caution when interpreting this result (
Analysis 2.1).
We reported the results
of subgroup analyses quoting the Chi ² statistic and P value, and the interaction test I ² value.
On the whole, there was no evidence
of a difference between the caseload and team
subgroups for any
of the outcomes included in the
subgroup analysis.
Because our
subgroup analyses (reported below) did not generally explain heterogeneity found in specific primary outcomes, we discuss additional sources
of heterogeneity below and in the discussion section
of the review.
On the whole, there was no evidence
of a difference between the caseload and team
subgroups for any
of the outcomes included in the
subgroup analysis, which included caesarean section, instrumental vaginal birth, spontaneous vaginal birth, intact perineum, preterm birth < 37 weeks and all fetal loss before and after 24 weeks plus neonatal death.
We conducted pre-specified
subgroup analysis to investigate heterogeneity in the above outcomes
of regional analgesia and preterm birth.
We will report the results
of subgroup analyses quoting the Chi ² statistic and P value, and the interaction test I ² value.
Because race is a factor in breastfeeding duration among term infants (39 %
of white US infants are breastfeeding at 6 months compared with 24 %
of African American infants3) and because
of the high incidence
of premature births among African American women, we performed a
subgroup analysis on African American subjects.
In the
subgroup analysis in which we excluded women whose labour was induced by outpatient administration
of prostaglandins, amniotomy or both (118 [4.1 %]
of women in the home - birth group, 344 [7.2 %]
of those who planned a midwife - attended hospital birth and 778 [14.6 %]
of those who planned a physician - attended hospital birth), the relative risks
of obstetric interventions and adverse maternal and neonatal outcomes did not change significantly.
Four
of the 6 studies found no relationship between diphtheria - tetanus - pertussis vaccination and subsequent SIDS, 316, — , 319 and results
of the other 2 studies suggested a temporal relationship but only in specific
subgroup analysis.320, 321 In 2003, the Institute
of Medicine
of the National Academy
of Sciences reviewed available data and concluded that «[t] he evidence favors rejection
of a causal relationship between exposure to multiple vaccinations and SIDS.»
For cessation
of any breastfeeding at up to six months there was no evidence
of a differential effect according to type
of support (test for
subgroup differences: Chi ² = 0.40, df = 2 (P = 0.82), I ² = 0 %;
Analysis 3.1).
There was no evidence
of a difference between the caseload and team
subgroups for any
of the outcomes included in the
subgroup analysis.
Although the results
of the meta - regression showed no evidence
of significant heterogeneity between
subgroups, summary association estimates were slightly different in
subgroup analyses by study design and exposure assessment.
For cessation
of any breastfeeding at up to six months, there is no evidence
of a difference in the effectiveness
of the intervention according to the background breastfeeding rate (test for
subgroup differences: Chi ² = 0.56, df = 2 (P = 0.76), I ² = 0 %;
Analysis 5.1).
Therefore, for the review's four primary outcomes we carried out
subgroup analysis to explore the impact
of interventions involving different types
of supporter (professional versus lay person, or both); types
of support (face - to - face versus telephone support or both); timing
of support (antenatal and postnatal versus postnatal alone); whether the support was proactive (scheduled contacts) or reactive (women needed to request support); and whether support interventions had similar effects in settings with different background breastfeeding initiation rates (low, medium or high background rates).
We had planned to carry out formal
subgroup analysis by whether support was proactive or reactive, but due to the fact that most interventions included at least one scheduled contact (proactive), we did not think that this way
of categorising studies would shed light on types
of interventions that were effective or ineffective.
We conducted the planned
subgroup analyses to investigate the effects
of greater continuity in caseload models, variations in maternal risk status and
of less medicalised environments provided by community settings.
We reported the results
of subgroup analyses quoting the χ2 statistic and P value, and the interaction test I ² value.
For cessation
of exclusive breastfeeding at up to six months face - to - face interventions may be associated with greater effects than other types
of support; however, very high within - group heterogeneity remains in the
analysis, and we advise caution when interpreting this result (test for subgroup differences: Chi ² = 37.55, df = 2 (P <.00001, I ² = 94.7 %; Analys
analysis, and we advise caution when interpreting this result (test for
subgroup differences: Chi ² = 37.55, df = 2 (P <.00001, I ² = 94.7 %;
AnalysisAnalysis 3.2).
All
of these
subgroup analyses investigate potential sources
of heterogeneity, as differences in the type
of intervention, risk profile
of the population or setting may affect the treatment effects.
For cessation
of exclusive breastfeeding at up to six months the treatment effect appears to be greater when the intervention was delivered by non-professionals (lay support) compared with professionals or mixed support (test for
subgroup differences: Chi ² = 7.74, df = 2 (P = 0.02), I ² = 73.1 %;
Analysis 2.2).
(By including people who had vascular disease, the Lancet meta -
analysis overestimated statins» benefits: a
subgroup analysis reveals that among people who did not have vascular disease, statins only reduced the absolute risk
of a cardiovascular event by 0.4 percent per year.)
My goal was that it would be accurate enough to detect poor technique, helping me cross-check my results against a laboratory
analysis of a
subgroup of 20 players (given that it was impossible to film all 200
of them in the lab!).
The
analysis calculated rates
of vaginal birth
of a healthy singleton at term in natural and assisted reproduction conception comparing women in the intervention (lifestyle modification) group and those in the control (prompt treatment) group according to six different
subgroups: these
subgroups were defined by age (over or under 36 years), cycle regularity (ovulatory or anovulatory) and body weight (above or below a BMI
of 35 kg / m2).
The benefits
of gabapentin were also similar on
analysis of different
subgroups, including patient considered at high risk
of nausea and vomiting.
Commenting on the findings, Prof Robert Pirker, programme director for lung cancer at the Vienna General Hospital in Vienna, Austria, not involved in the study, said: «This
subgroup analysis shows that the effect
of necitumumab was slightly greater in patients with EGFR expressing tumours than it was in the entire SQUIRE population.
They also performed comparative and genome - wide
analyses, which provide insights into the evolutionary history
of Asterids, a
subgroup of flowering plants that includes potatoes, tomatoes and coffee.
«By combining the genetic
analysis of a small population
of immune cells from healthy skin with functional experiments we were able to define two
subgroups of memory immune cell and in detail decipher / dissect how these cells behave in healthy and inflamed skin,» explains Liv Eidsmo, researcher at Karolinska Institutet's Department
of Medicine.
«Further
analysis of this cohort revealed that disease - free survival increased by 46 % among the
subgroup of nut consumers who ate tree nuts rather than peanuts,» said Charles S. Fuchs, M.D., M.P.H., director
of Yale Cancer Center and senior author
of the study.
The team's genetic information came from
analyses of two particular genes from 230 species representing all major
subgroups of modern birds.
Even though the primary results
of a related study investigating the effects
of music therapy with children with autism, do not show that music therapy works better than other therapies,
subgroup analysis identified that children with childhood autism or coexisting intellectual disability improve to a greater extend from music therapy than children with another autism diagnosis.
Subgroup analyses did not reveal any differences in the risk estimates based on the type
of studies or the measurement
of the exposure.
A secondary
analysis demonstrated that malaria risk was reduced among the
subgroup of those in the iron group who had iron deficiency and anemia at baseline.»
Using
analysis of variance, t - tests, or chi - square tests as appropriate, we compared the clinical characteristics among exposure groups and among subjects with World Trade Center cough according to prognostic
subgroups: those with predominantly upper - airway symptoms (29 subjects) or lower - airway symptoms (95 subjects) or those with bronchial responsiveness (249 subjects).
Quantitative
analyses of multiple cardiac phenotypes demonstrated essential structural, functional and developmental roles for these genes, including a
subgroup encoding histone H3K4 modifying proteins.
The authors next took 997 tumors in the discovery set, integrated copy number and gene expression data, and performed clustering
analyses to identify
subgroups of tumors with distinct features and clinical outcomes.
«A
subgroup analysis that combines data from children in both the single - and multiple - dose studies demonstrated a mean 5 point increase in muscle function score in children who received at least 9 mg
of ISIS - SMNRx between the ages
of two and 10 who did not have severe scoliosis or baseline HFMSE scores at the extreme low or high ends
of the scale.
There were similar results across prespecified
subgroups of patients including
analyses of PD - L1 expression, which showed hazard ratios in favor
of nivolumab.
Subgroup analyses showed benefit with the combination in all groups, with the exception
of those with brain metastases at baseline, though this included a total
of only 12 patients.
We are continuing this
analysis as the leaders
of the imprinting
subgroup in GTEx, with particular interest in variable imprinting and molecular mechanisms
of allelic silencing.
The results
of a metaanalysis
of observational studies, with
subgroup analysis by sex, presented that shift work was associated with greater diabetes risk in men (107).