CMMM Computational and Mathematical Methods in Medicine 1748-6718 1748-670X Hindawi 10.1155/2017/9270450 9270450 Research Article Pairwise Multiple Comparison Adjustment Procedure for Survival Functions with Right-Censored Data http://orcid.org/0000-0003-3251-1043 Colak Ertugrul 1 http://orcid.org/0000-0003-4144-3732 Ozen Hulya 1 http://orcid.org/0000-0003-4694-1319 Emir Busra 1 http://orcid.org/0000-0002-4759-4913 Oner Setenay 1 Yanover Chen Department of Biostatistics Faculty of Medicine Eskisehir Osmangazi University Eskisehir Turkey ogu.edu.tr 2017 12102017 2017 17 05 2017 31 08 2017 14 09 2017 12102017 2017 Copyright © 2017 Ertugrul Colak et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The aim of this study is to propose a new pairwise multiple comparison adjustment procedure based on Genz’s numerical computation of probabilities from a multivariate normal distribution. This method is applied to the results of two-sample log-rank and weighted log-rank statistics where the survival data contained right-censored observations. We conducted Monte Carlo simulation studies not only to evaluate the familywise error rate and power of the proposed procedure but also to compare the procedure with conventional methods. The proposed method is also applied to the data set consisting of 815 patients on a liver transplant waiting list from 1990 to 1999. It was found that the proposed method can control the type I error rate, and it yielded similar power as Tukey’s and high power with respect to the other adjustment procedures. In addition to having a straightforward formula, it is easy to implement.

1. Introduction

Survival analysis is based on making inferences from the time-to-event data. It provides many statistical procedures for studying the data, including the time from a correctly identified origin until the occurrence of a certain event . One of the main interests in survival analysis is evaluating the equality of survival functions for different groups. Many tests such as log-rank and weighed log-rank have been proposed . Although these tests made important contributions to survival analysis, they can only provide overall or two-sample comparison results. Researchers will fail if they use these tests to compare one with another in a multigroup study design because the probability of making at least one type I error will be increased above the critical level. To prevent this mistake, pairwise multiple comparison procedures are needed. In case of the inequality of more than two groups, it is necessary to correctly decide which groups are different from the others. The appropriate way to control the type I error is to consider the familywise error (FWE) rate, which is the probability of making at least one type I error when making all pairwise comparisons .

Adjustment methods such as Bonferroni, Holm, and Sidak methods are commonly used in the literature. However, in survival analysis this topic has only recently been studied. Adjustment methods are applied to the results of two-sample log-rank and weighted log-rank tests. Bonferroni is the most preferred method among the others. In a two-sided test, Bonferroni assumes the significance level as (α/2)×m, where m is the number of pairwise comparisons in the study, but it fails when controlling the familywise error rate. In spite of its simplicity, it has been determined to be a conservative method in survival analysis [9, 10]. Logan et al. proposed two different adjustment methods that consider the correlation among the pairwise tests . One of the methods was derived from multivariate normal distribution, while the other was obtained from a simulated martingales approach. Koziol and Reid used the Sidak adjustment method to calculate the pairwise comparisons results of weighted log-rank tests. Although it generates more consistent results than Bonferroni’s, it was also found to be conservative . Not only were pairwise multiple comparisons proposed, but comparisons against a single control group were also proposed for survival functions with right-censored data in the statistical literature. Chakraborti and Desu developed linear rank tests, and Chen proposed a generalized Steel’s test and an alternative method to the generalization of Steel’s test .

The aim of this study is to propose a new pairwise multiple comparison adjustment procedure based on Genz’s numerical computation of probabilities from a multivariate normal distribution [16, 17]. This method is applied to the results of two-sample log-rank and weighted log-rank statistics where the survival data contained right-censored observations. In Section 2, some notations are given, and the construction of the simulation study is detailed. In the simulation studies SAS PROC LIFETEST and R package with mvtnorm library were used. Moreover, all adjustment methods are applied to a real life-time data set and they are compared with each other. The results and discussion about other studies are evaluated in Section 3. Finally, conclusions are mentioned in Section 4.

2. Materials and Methods 2.1. Notation and Background

Let Skt be the survival function of the kth group for  k=1,,K, where K is the number of groups. The null and alternative hypotheses for the survival functions are(1)H0: S1t==SKtH1:  at least one of the Skt’s is different for some tτ,where τ is the largest observed time.

Let (Ti,δi,Xi,wi), for i=1,,n, indicates that an independent sample for right-censored survival data where Ti is right-censored time, δi is the indicator variable for censoring (δi=0 if Ti is censored; δi=1 if Ti is an event time), Xi is the group indicator of 1,,K, and wi is a weight function. Let t1<t2<<tDj=1,,D be distinct event times in the sample. At time tj, for the kth group, let Yjk=i:TtjI(Xi=k) and djk=i:Ti=tjδiI(Xi=k) denote the number of individuals at risk and the number of events, respectively. Let Yj=k=1KYjk and dj=k=1Kdjk denote the number individuals at risk and the number of events, respectively. The weighted number of individuals at risk in the kth group is Yjkw=i:TitjwiI(Xi=k), while the weighted number of events in the kth group is djkw=i:Ti=tjwiδiI(Xi=k). Let Yjw=k=1KYjkw and djw=k=1Kdjkw indicate the weighted number of individuals at risk and the weighted number of events, respectively.

For testing the null hypothesis, the test statistics have the form of a K-vector R=r1,r2,,rk, where (2)rk=j=1Ddjkw-YjkwdjwYjw.Variance of rk and covariance for rk and rh are as follows, respectively:(3)vkk=j=1DdjYj-djYjYj-1i=1YjYjkwYjw2wi2IXik+Yjw-YjkwYjw2wi2IXi=k(4)vkh=j=1DdjYj-djYjYj-1i=1YjYjkwYjhwYjw2wi2IXik,h-Yjw-YjkwYjhwYjw2wi2IXi=k-Yjw-YjhwYjkwYjw2wi2IXi=h.Because the sum of rk is equal to 0, they are linearly dependent. Accordingly, the general test statistic is constructed by selecting any K-1 of rk’s. The test statistic, r1,r2,,rK-1VK-1×K-1-1r1,r2,,rK-1, follows a Chi-square distribution with K-1 degrees of freedom, where V is the variance-covariance matrix.

Let m be the number of all pairwise comparisons where  m=K(K-1)/2. The two-sided test statistic, Zkh, compares the groups k and h and follows a standard normal distribution.(5)Zkh=rk-rhvkk+vhh-2vkh.The unadjusted p value is  pkh=P(χ12>Zkh2). The multiple comparison procedures that are used to adjust the p values in this study are shown below:

Bonferroni: pkh=min1,m×Pχ12>Zkh2.

Scheffé: pkh=PχK-12>Zkh2.

Sidak: pkh=1-1-Pχ12>Zkh2m.

Studentized Maximum Modulus: pkh=1-2×ΦZkhm.

Tukey: pkh=1--KϕuΦu-Φu-2ZkhK-1du,

where ϕ and Φ are standard normal and cumulative standard normal functions, respectively.

Z = Z 12 , , Z 1 K , Z 23 , , Z 2 K , Z K - 1 , K m has a multivariate normal distribution with a mean of zero and a variance-covariance matrix Σ. Under the null hypothesis, the elements of Σ follow a rule which is CovZkh,Zkh=0.5, CovZkh,Zkh=0, and CovZkh,Zhk=-0.5, where 1khkhK [9, 14, 15].

The function of a,b,Σ is(6)ΦZa,b,Σ=1Σ2πma1b1a2b2ambme-1/2ZΣ-1Zdz,-aibi,  i=1,,m.For the integration shown above, we used “mvtnorm” library, released February 2, 2016, for numerical computation in R program. There are three algorithms available for evaluating normal probabilities: The default is the randomized Quasi-Monte-Carlo procedure by Genz (1992, 1993). We used this approach because it is easy to use and calculate with R program.

The proposed multiple adjustment procedures for the pairwise comparison of the kth and hth groups are obtained using Φ and shown below:(7)pkh=min1,2×P1-ΦZa,b,Σ,where a=-,,-m and b=Zkh,,Zkhm.

Additionally, the critical value for the pairwise comparison can be evaluated with(8)Zc=Φ-11-α2.

2.3. Simulation Study

We performed Monte Carlo simulation studies to examine the proposed and conventional adjustment procedures. The FWE rate and power of the adjustment procedures were obtained through the simulation results. In this study, the number of groups was determined as K=  4; Xi=1,,4. The sample sizes were considered equal for each group as n=50, 150, and 250 to estimate the FWE rate, while it was just 250 in the power study. The right-censored survival times Ti were derived from the exponential Ti~exp(λXi) and lognormal distribution Ti~lognormal(μXi,σ2). The censoring rate was considered to be 30%. Therefore, the censoring variable was generated from a Bernoulli distribution δi~Bernoulli(p=0.70). Note that the censoring rate was fixed for each group in the FWE rate and power study. To obtain the adjusted p values, Bonferroni, Scheffé, Sidak, SMM, Tukey, and the proposed adjustment procedure were applied to the pairwise comparison results of log-rank and weighted log-rank tests. For each scenario 1000 data sets were simulated independently.

To compare the FWE rates of the adjustment procedures, the survival times for each group were generated from the standard exponential distribution with λk=1 and the lognormal distribution with a mean of μk=0 and scale parameter  σk=0.5. The estimated FWE rates of the adjustment procedures were evaluated with respect to the critical value α=0.05. In the power study, we used exponential distributions with various parameters λk and lognormal distributions with σk=0.5 but different values of μk. For power calculation, we calculated the probability of making a correct decision only for unequal pairs. Note that the exponential distribution provides a proportional hazards model while the lognormal distribution corresponds to location shifts in log survival times. The lognormal distributions with various means were used because they have different hazards at early times .

2.4. Application Data

The data set was obtained from the free data sets used in the R package, “survival” [18, 19]. It consisted of 815 patients on a liver transplant waiting list from 1990 to 1999 with six variables: age at the addition to the waiting list, sex, blood type, year in which a patient entered the waiting list, and time from the entry to end point. The final disposition of the patients was categorized as received a transplant, died while waiting, withdrew from the list, or censored. Blood type is a crucial factor which affects the waiting time for transplantation. Although the liver donation from subjects with blood type O can be used by patients with all blood types, a patient with blood type O can only receive donation from the subjects with blood type O. Thus, patients with O blood type on the waiting list have a disadvantage. These data is of historical interest and provides a useful example of competing risks, but it has little relevance to current practice. We used these data as an example to demonstrate the comparison of the proposed and conventional adjustment techniques on a real data set. We considered that the event is receiving a transplant, while the other categories of final disposition are censored.

3. Results and Discussion

Table 1 shows the simulation results for the estimated FWE rates of the proposed and conventional adjustment procedures for exponential survival distribution with different sample sizes. Under the null hypothesis, FWE rates are expected to be 0.05. As the sample size increases, estimates get closer to the targeted value in all adjustment procedures. It is obvious that the Scheffé method is the most inefficient among the others. The proposed adjustment procedure and Tukey’s present similar results. It can be seen that both adjustment procedures can control the type I error even for small samples. Their performance is followed by Sidak, SMM, and Bonferroni procedures. In Table 2, the estimates of the FWE rates for the survival times from the lognormal distribution with the parameters μk=0 and σk=0.5 are given. Unlike the previous simulation results, not all procedures give estimates that are close to the targeted value. The proposed adjustment procedure and Tukey’s provide the most efficient results. The decrease in the performance of the adjustment procedures could depend on the type of distributions. Because an exponential distribution provides a more appropriate proportional hazard model than a lognormal distribution, this affects the performance of the log-rank and the weighted log-rank tests. Therefore, the adjustment procedures tend to cause errors.

FWE rates of the proposed and conventional adjustment procedures for K=4 and α=0.05 and exponential survival distribution with λk=1.

Sample size Tests Proposed and conventional adjustment techniques
Unadjusted Bonferroni Scheffé Sidak SMM Tukey Proposed
50 Fleming 0.194 0.039 0.031 0.040 0.040 0.053 0.053
Log-rank 0.187 0.040 0.024 0.040 0.040 0.046 0.046
ModPeto 0.196 0.040 0.031 0.043 0.043 0.054 0.054
Peto 0.194 0.039 0.031 0.042 0.042 0.054 0.054
Tarone 0.193 0.041 0.027 0.041 0.041 0.053 0.053
Wilcoxon 0.204 0.043 0.029 0.044 0.044 0.052 0.052

150 Fleming 0.206 0.034 0.022 0.035 0.035 0.039 0.039
Log-rank 0.185 0.036 0.019 0.037 0.037 0.044 0.044
ModPeto 0.204 0.033 0.023 0.034 0.034 0.038 0.038
Peto 0.206 0.034 0.022 0.034 0.034 0.038 0.038
Tarone 0.198 0.035 0.020 0.035 0.035 0.045 0.045
Wilcoxon 0.211 0.038 0.023 0.038 0.038 0.045 0.045

250 Fleming 0.214 0.045 0.032 0.046 0.046 0.057 0.057
Log-rank 0.209 0.043 0.030 0.044 0.044 0.049 0.049
ModPeto 0.214 0.045 0.032 0.046 0.046 0.057 0.057
Peto 0.214 0.045 0.032 0.046 0.046 0.057 0.057
Tarone 0.210 0.047 0.033 0.047 0.047 0.054 0.054
Wilcoxon 0.209 0.044 0.029 0.045 0.045 0.056 0.056

FWE rates of the proposed and conventional adjustment procedures for K=4 and α=0.05 and log-normal survival distribution with μk=0 and σk=0.5.

Sample size Tests Proposed and conventional adjustment techniques
Unadjusted Bonferroni Scheffé Sidak SMM Tukey Proposed
50 Fleming 0.188 0.035 0.023 0.035 0.035 0.041 0.041
Log-rank 0.199 0.038 0.021 0.038 0.038 0.043 0.043
ModPeto 0.187 0.036 0.023 0.036 0.036 0.041 0.041
Peto 0.189 0.035 0.023 0.036 0.036 0.041 0.041
Tarone 0.182 0.032 0.022 0.033 0.033 0.041 0.041
Wilcoxon 0.182 0.038 0.019 0.038 0.038 0.047 0.047

150 Fleming 0.202 0.046 0.025 0.046 0.046 0.051 0.051
Log-rank 0.220 0.043 0.030 0.043 0.043 0.050 0.050
ModPeto 0.200 0.045 0.025 0.046 0.046 0.051 0.051
Peto 0.201 0.045 0.025 0.046 0.046 0.051 0.051
Tarone 0.210 0.041 0.029 0.043 0.043 0.051 0.051
Wilcoxon 0.196 0.044 0.023 0.044 0.044 0.051 0.051

250 Fleming 0.196 0.040 0.024 0.041 0.041 0.049 0.049
Log-rank 0.201 0.037 0.023 0.037 0.037 0.044 0.044
ModPeto 0.197 0.040 0.024 0.040 0.040 0.049 0.049
Peto 0.196 0.040 0.024 0.041 0.041 0.049 0.049
Tarone 0.195 0.045 0.023 0.046 0.046 0.049 0.049
Wilcoxon 0.202 0.032 0.021 0.034 0.034 0.042 0.042

Next, the simulation results are calculated for the power of the proposed and conventional adjustment procedures for the exponential survival distribution. Under a variety of hypothesis configurations denoted by λk, the estimated power results are given in Table 3. As the values of λk become different from each other, the power of all of the adjustment procedures decreases rapidly. The proposed adjustment procedure and Tukey’s provide similar results with the highest power. We also conducted additional simulations where the survival times were derived from a lognormal distribution. The estimates of power under alternative configurations of μk are given in Table 4. Inefficient power results are only seen when all of the μk values are different. Moreover, the performance of all of the adjustment procedures gives very similar results. In all the simulation results, it can be seen that there is no notable difference between the log-rank and weighted log-rank tests.

Power of the proposed and conventional adjustment procedures for K=4 and α=0.05,and exponential survival distribution with different λk.

Parameters Tests Proposed and conventional adjustment techniques
λ 1 , λ 2 , λ 3 , λ 4 Unadjusted Bonferroni Scheffé Sidak SMM Tukey Proposed
(2.25, 1.50, 1.50, 1.50) Fleming 0.765 0.594 0.523 0.597 0.597 0.618 0.618
Log-rank 0.858 0.785 0.726 0.784 0.784 0.802 0.802
ModPeto 0.763 0.593 0.521 0.593 0.593 0.618 0.618
Peto 0.764 0.593 0.522 0.597 0.597 0.618 0.618
Tarone 0.789 0.654 0.595 0.657 0.657 0.676 0.676
Wilcoxon 0.725 0.506 0.428 0.507 0.507 0.535 0.535

(2.25, 2.25, 1.50, 1.50) Fleming 0.757 0.514 0.437 0.516 0.516 0.541 0.541
Log-rank 0.823 0.627 0.555 0.629 0.629 0.665 0.665
ModPeto 0.756 0.512 0.436 0.516 0.516 0.539 0.539
Peto 0.757 0.513 0.436 0.516 0.516 0.541 0.541
Tarone 0.789 0.558 0.490 0.560 0.560 0.588 0.588
Wilcoxon 0.713 0.442 0.358 0.447 0.447 0.468 0.468

(2.25, 1.75, 1.75, 1.25) Fleming 0.243 0.032 0.017 0.033 0.033 0.045 0.045
Log-rank 0.368 0.063 0.034 0.064 0.064 0.080 0.080
ModPeto 0.243 0.032 0.017 0.033 0.033 0.044 0.044
Peto 0.243 0.032 0.017 0.033 0.033 0.044 0.044
Tarone 0.290 0.046 0.024 0.047 0.047 0.055 0.055
Wilcoxon 0.186 0.026 0.009 0.026 0.026 0.031 0.031

(2.50, 2.00, 1.50, 1.00) Fleming 0.168 0.010 0.002 0.010 0.010 0.014 0.014
Log-rank 0.269 0.023 0.006 0.024 0.024 0.035 0.035
ModPeto 0.167 0.009 0.002 0.010 0.010 0.013 0.013
Peto 0.167 0.010 0.002 0.010 0.010 0.014 0.014
Tarone 0.204 0.012 0.006 0.013 0.013 0.018 0.018
Wilcoxon 0.121 0.005 0.000 0.005 0.005 0.005 0.005

Power of the proposed and conventional adjustment procedures for K=4 and α=0.05 and log-normal survival distribution with different μk and σk=0.5.

Parameters Tests Proposed and conventional adjustment techniques
μ 1 , μ 2 , μ 3 , μ 4 Unadjusted Bonferroni Scheffé Sidak SMM Tukey Proposed
(0.5, 0, 0, 0) Fleming 0.871 0.978 0.985 0.978 0.978 0.976 0.976
Log-rank 0.901 0.995 0.996 0.995 0.995 0.991 0.991
ModPeto 0.871 0.978 0.985 0.978 0.978 0.976 0.976
Peto 0.871 0.978 0.985 0.978 0.978 0.976 0.976
Tarone 0.880 0.979 0.989 0.978 0.978 0.973 0.973
Wilcoxon 0.869 0.978 0.985 0.977 0.977 0.973 0.973

(0.5, 0.5, 0, 0) Fleming 0.922 0.987 0.992 0.987 0.987 0.984 0.984
Log-rank 0.923 0.988 0.995 0.988 0.988 0.985 0.985
ModPeto 0.923 0.987 0.992 0.987 0.987 0.984 0.984
Peto 0.923 0.987 0.992 0.987 0.987 0.984 0.984
Tarone 0.928 0.987 0.994 0.987 0.987 0.986 0.986
Wilcoxon 0.931 0.986 0.992 0.986 0.986 0.983 0.983

(0.3, 0, 0, −0.3) Fleming 0.962 0.982 0.980 0.982 0.982 0.984 0.984
Log-rank 0.947 0.940 0.906 0.940 0.940 0.949 0.949
ModPeto 0.962 0.982 0.980 0.982 0.982 0.984 0.984
Peto 0.962 0.982 0.980 0.982 0.982 0.984 0.984
Tarone 0.958 0.979 0.974 0.979 0.979 0.980 0.980
Wilcoxon 0.962 0.985 0.979 0.985 0.985 0.984 0.984

(0.5, 0.3, −0.3, −0.5) Fleming 0.716 0.332 0.245 0.336 0.336 0.373 0.373
Log-rank 0.551 0.260 0.199 0.262 0.262 0.293 0.293
ModPeto 0.716 0.329 0.244 0.337 0.337 0.371 0.371
Peto 0.716 0.332 0.245 0.337 0.337 0.371 0.371
Tarone 0.712 0.345 0.278 0.347 0.347 0.395 0.395
Wilcoxon 0.698 0.266 0.191 0.271 0.271 0.304 0.304

Descriptive statistics of the application data set are given in Table 5 and the survival functions of the groups are shown in Figure 1. The overall comparison of blood type groups is conducted with log-rank test. The result is found to be highly significant (χ2=45.5, df = 3, and p<0.001). Thus, pairwise comparisons followed by multiple adjustment procedures were conducted, and the results are given in Table 6. All of the adjustment procedures had the same conclusions and present results that are similar to those that we observed in the simulation studies. The comparison results show that, with the exception of the pair of B and AB, all of the blood types are highly different from each other. The p values obtained for each comparative test for the application data showed significant differences (p<0.001) between the survival times of the blood groups except for the comparison of AB and B groups (p>0.05). The results can be seen in Kaplan-Meier curves represented in Figure 1. The survival curves show a proportional structure until the middle of the 0–500-day interval. Also, the survival curves of AB and B blood groups are closer to each other compared to the other groups.

Descriptive statistics for the liver transplant waiting list data.

Blood groups LTX Censored Total Percent censored Median follow-up (days) 95% confidence interval
Lower Upper
A 269 56 325 0.172 100 95 108
AB 33 8 41 0.195 84 52 202
B 78 25 103 0.243 173 116 212
0 256 90 346 0.260 223 193 276
Total 636 179 815 0.219

Test statistics and the adjusted p values of the proposed and conventional adjustment techniques for the liver transplant waiting list data.

Tests Blood groups Test statistics Proposed and conventional adjustment techniques
Z Unadjusted Bonferroni Scheffé Sidak SMM Tukey Proposed
Fleming A AB 5.106 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A B 5.236 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A 0 7.570 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
AB B 1.639 0.1012 0.6070 0.4424 0.4727 0.4727 0.3565 0.4297
AB 0 7.039 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
B 0 4.826 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

Log-rank A AB 4.543 <0.0001 <0.0001 0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A B 4.519 <0.0001 <0.0001 0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A 0 6.483 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
AB B 1.258 0.2084 1.0000 0.6634 0.7539 0.7539 0.5898 0.7716
AB 0 5.924 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
B 0 4.088 <0.0001 0.0003 0.0008 0.0003 0.0003 0.0003 <0.001

ModPeto A AB 5.103 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A B 5.233 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A 0 7.570 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
AB B 1.638 0.1015 0.6089 0.4433 0.4738 0.4738 0.3573 0.4303
AB 0 7.042 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
B 0 4.829 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

Peto A AB 5.102 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A B 5.231 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A 0 7.567 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
AB B 1.637 0.1016 0.6094 0.4435 0.4741 0.4741 0.3576 0.4312
AB 0 7.040 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
B 0 4.827 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

Tarone A AB 5.131 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A B 5.153 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A 0 7.480 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
AB B 1.480 0.1388 0.8330 0.5338 0.5921 0.5921 0.4495 0.5594
AB 0 6.887 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
B 0 4.776 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

Wilcoxon A AB 5.100 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A B 5.264 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
A 0 7.598 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
AB B 1.688 0.0915 0.5488 0.4156 0.4376 0.4376 0.3301 0.3944
AB 0 7.084 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
B 0 4.840 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

Kaplan-Meier estimates of not receiving a transplant for each blood type group.

A statistician can use this method in usual data analysis procedure as follows. For example, to calculate the adjusted p value for the comparison of the groups k and h,

calculate Zkh and Σ defined in Section 2.1,

use pvnorm command in mvtnorm library in R as follows:

l=rep(-Inf,m).

u=rep(Zkh,m).

a=pmvnorm(lower=l,  upper=u,  mean=0,  corr=sigma).

p=2(1-(a+attributes(a)\$error)),

where m is the number of all comparisons, and sigma is Σ.

4. Conclusions

In this study, we proposed a multiple adjustment procedure for the pairwise comparisons of survival functions with right-censored data. We conducted Monte Carlo simulation studies not only to evaluate the FWE rate and power of the proposed procedure but also to compare the procedure with conventional methods. It was found that the proposed method can control the type I error rate, and it yielded similar power as Tukey’s and high power with respect to the other adjustment procedures. In addition to having a straightforward formula, it is easy to implement.

This study has some limitations. The main issue was that the simulations were performed by using proposed and conventional methods. However, comparisons can be extended including the methods such as that of Logan et al. (2005) in the comparison. Logan et al. proposed two different adjustment methods that consider the correlation among the pairwise tests. One of the methods was derived from multivariate normal distribution, while the other was obtained from a simulated martingales approach. These models may work well for the data with proportional hazard structure. Future researches should take into account the models for comparisons.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Collett D. Modelling Survival Data in Medical Research 2015 Boca Raton, Fla, USA CRC press Gehan E. A. A generalized Wilcoxon test for comparing arbitrarily singly-censored samples Biometrika 1965 52 203 223 10.1093/biomet/52.1-2.203 MR0207130 Zbl0133.41901 2-s2.0-84951601829 10.2307/2333825 Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration Cancer Chemotherapy Reports 1966 50 3 163 170 2-s2.0-0013886333 Breslow N. A generalized Kruskal-Wallis test for comparing k samples subject to unequal patterns of censorship Biometrika 1970 57 3 579 594 2-s2.0-77956891394 10.1093/biomet/57.3.579 10.2307/2334776 Zbl0215.26403 Peto R. Peto J. Asymptotically efficient rank invariant test procedures Journal of the Royal Statistical Society. Series A (General) 1972 135 2 185 10.2307/2344317 Prentice R. L. Linear rank tests with right censored data Biometrika 1978 65 1 167 179 10.1093/biomet/65.1.167 MR497517 Zbl0377.62024 2-s2.0-0017836371 10.2307/2335292 Tarone R. E. Ware J. On distribution-free tests for equality of survival distributions Biometrika 1977 64 1 156 160 10.1093/biomet/64.1.156 MR0461830 Zbl0352.62046 2-s2.0-0017336302 Harrington D. P. Fleming T. R. A class of rank test procedures for censored survival data Biometrika 1982 69 3 553 566 10.1093/biomet/69.3.553 MR695202 Zbl0532.62026 2-s2.0-33748208652 Logan B. R. Wang H. Zhang M.-J. Pairwise multiple comparison adjustment in survival analysis Statistics in Medicine 2005 24 16 2509 2523 10.1002/sim.2125 MR2112379 2-s2.0-23244466624 Tressler A. Chow A. Multiple pairwise comparison procedures based on the Lin and Wang test for right censored survival data Statistica Neerlandica 2013 67 1 112 120 10.1111/j.1467-9574.2012.00535.x MR3015786 Koziol J. A. Reid N. On multiple comparisons among K samples subject to unequal patterns of censorship Communications in Statistics—Theory and Methods 1977 6 12 1149 1164 10.1080/03610927708827558 2-s2.0-0040430176 Steel R. G. A multiple comparison rank sum test: treatments versus control Biometrics 1959 15 560 572 10.2307/2527654 MR0108869 Chakraborti S. Desu M. M. Linear rank tests for comparing treatments with a control when data are subject to unequal patterns of censorship Statistica Neerlandica 1991 45 3 227 254 10.1111/j.1467-9574.1991.tb01307.x MR1142082 Chen Y.-I. A generalized steel procedure for comparing several treatments with a control under random right-censorship Communications in Statistics—Simulation and Computation 1994 23 1 1 16 10.1080/03610919408813152 2-s2.0-0039837533 Chen Y.-I. Multiple comparisons in carcinogenesis study with right-censored survival data Statistics in Medicine 2000 19 3 353 367 2-s2.0-0034651970 10.1002/(SICI)1097-0258(20000215)19:3<353::AID-SIM333>3.0.CO;2-B 10.1002/(SICI)1097-0258(20000215)19:3<353::AID-SIM333>3.0.CO;2-B Genz A. Numerical computation of multivariate normal probabilities Journal of Computational and Graphical Statistics 1992 1 2 141 149 10.1080/10618600.1992.10477010 2-s2.0-0001341675 Genz A. Comparison of methods for the computation of multivariate normal probabilities Computing Scienc e and Statistics 1993 25 400 4005 Kim W. R. Therneau T. M. Benson J. T. Kremers W. K. Rosen C. B. Gores G. J. Dickson E. R. Deaths on the liver transplant waiting list: an analysis of competing risks Hepatology 2006 43 2 345 351 10.1002/hep.21025 2-s2.0-33644522082 Therneau T. M. Lumley T. Package ‘survival’, Ed., Verze, 2016