Bayesian Survival Analysis

Author: Austin Rochford

Survival analysis studies the distribution of the time to an event. Its applications span many fields across medicine, biology, engineering, and social science. This tutorial shows how to fit and analyze a Bayesian survival model in Python using PyMC3.

We illustrate these concepts by analyzing a mastectomy data set from R‘s HSAUR package.

In [1]:

%matplotlib inline

In [2]:

from matplotlib import pyplot as plt
import numpy as np
import pymc3 as pm
from pymc3.distributions.timeseries import GaussianRandomWalk
import seaborn as sns
from statsmodels import datasets
from theano import tensor as T

Couldn't import dot_parser, loading of dot files will not be possible.

Fortunately, statsmodels.datasets makes it quite easy to load a number of data sets from R.

In [3]:

df = datasets.get_rdataset('mastectomy', 'HSAUR', cache=True).data
df.event = df.event.astype(np.int64)
df.metastized = (df.metastized == 'yes').astype(np.int64)
n_patients = df.shape[0]
patients = np.arange(n_patients)

In [4]:

df.head()

Out[4]:

	time	event	metastized
0	23	1	0
1	47	1	0
2	69	1	0
3	70	0	0
4	100	0	0

In [5]:

n_patients

Out[5]:

44

Each row represents observations from a woman diagnosed with breast cancer that underwent a mastectomy. The column time represents the time (in months) post-surgery that the woman was observed. The column event indicates whether or not the woman died during the observation period. The column metastized represents whether the cancer had metastized prior to surgery.

This tutorial analyzes the relationship between survival time post-mastectomy and whether or not the cancer had metastized.

A crash course in survival analysis

First we introduce a (very little) bit of theory. If the random variable T$T$ is the time to the event we are studying, survival analysis is primarily concerned with the survival function

S(t)=P(T>t)=1−F(t),$$S(t)=P(T>t)=1-F(t),$$

where F$F$ is the CDF of T$T$. It is mathematically convenient to express the survival function in terms of the hazard rate, λ(t)$\lambda (t)$. The hazard rate is the instantaneous probability that the event occurs at time t$t$given that it has not yet occured. That is,

λ(t)=limΔt→0P(t<T<t+Δt | T>t)Δt=limΔt→0P(t<T<t+Δt)Δt⋅P(T>t)=1S(t)⋅limΔt→0S(t+Δt)−S(t)Δt=−S′(t)S(t).$$\begin{array}{rl}\lambda (t)& =\underset{\mathrm{\Delta }t\to 0}{lim}\frac{P(t<T<t+\mathrm{\Delta }t|T>t)}{\mathrm{\Delta }t}\\ & =\underset{\mathrm{\Delta }t\to 0}{lim}\frac{P(t<T<t+\mathrm{\Delta }t)}{\mathrm{\Delta }t\cdot P(T>t)}\\ & =\frac{1}{S(t)}\cdot \underset{\mathrm{\Delta }t\to 0}{lim}\frac{S(t+\mathrm{\Delta }t)-S(t)}{\mathrm{\Delta }t}=-\frac{S^{\prime }(t)}{S(t)}.\end{array}$$

Solving this differential equation for the survival function shows that

S(t)=exp(−∫s0λ(s) ds).$$S(t)=\exp \left(-{\int }_0^s\lambda (s)ds\right).$$

This representation of the survival function shows that the cumulative hazard function

Λ(t)=∫t0λ(s) ds$$\mathrm{\Lambda }(t)={\int }_0^t\lambda (s)ds$$

is an important quantity in survival analysis, since we may consicesly write S(t)=exp(−Λ(t)).$S(t)=\exp (-\mathrm{\Lambda }(t)).$

An important, but subtle, point in survival analysis is censoring. Even though the quantity we are interested in estimating is the time between surgery and death, we do not observe the death of every subject. At the point in time that we perform our analysis, some of our subjects will thankfully still be alive. In the case of our mastectomy study, df.event is one if the subject’s death was observed (the observation is not censored) and is zero if the death was not observed (the observation is censored).

In [6]:

df.event.mean()

Out[6]:

0.59090909090909094

Just over 40% of our observations are censored. We visualize the observed durations and indicate which observations are censored below.

In [7]:

fig, ax = plt.subplots(figsize=(8, 6))

blue, _, red = sns.color_palette()[:3]

ax.hlines(patients[df.event.values == 0], 0, df[df.event.values == 0].time,
          color=blue, label='Censored')

ax.hlines(patients[df.event.values == 1], 0, df[df.event.values == 1].time,
          color=red, label='Uncensored')

ax.scatter(df[df.metastized.values == 1].time, patients[df.metastized.values == 1],
           color='k', zorder=10, label='Metastized')

ax.set_xlim(left=0)
ax.set_xlabel('Months since mastectomy')
ax.set_yticks([])
ax.set_ylabel('Subject')

ax.set_ylim(-0.25, n_patients + 0.25)

ax.legend(loc='center right');

When an observation is censored (df.event is zero), df.time is not the subject’s survival time. All we can conclude from such a censored obsevation is that the subject’s true survival time exceedsdf.time.

This is enough basic surival analysis theory for the purposes of this tutorial; for a more extensive introduction, consult Aalen et al. [1]_

[1]	Aalen, Odd, Ornulf Borgan, and Hakon Gjessing. Survival and event history analysis: a process point of view. Springer Science & Business Media, 2008.

Bayesian proportional hazards model

The two most basic estimators in survial analysis are the Kaplan-Meier estimator of the survival function and the Nelson-Aalen estimator of the cumulative hazard function. However, since we want to understand the impact of metastization on survival time, a risk regression model is more appropriate. Perhaps the most commonly used risk regression model is Cox’s proportional hazards model. In this model, if we have covariates x$\mathbf{x}$ and regression coefficients β$\beta$, the hazard rate is modeled as

λ(t)=λ0(t)exp(xβ).$$\lambda (t)={\lambda }_0(t)\exp (x\beta ).$$

Here λ0(t)${\lambda }_0(t)$ is the baseline hazard, which is independent of the covariates x$\mathbf{x}$. In this example, the covariates are the one-dimensonal vector df.metastized.

Unlike in many regression situations, x$\mathbf{x}$ should not include a constant term corresponding to an intercept. If x$\mathbf{x}$ includes a constant term corresponding to an intercept, the model becomesunidentifiable. To illustrate this unidentifiability, suppose that

λ(t)=λ0(t)exp(β0+xβ)=λ0(t)exp(β0)exp(xβ).$$\lambda (t)={\lambda }_0(t)\exp ({\beta }_0+x\beta )={\lambda }_0(t)\exp ({\beta }_0)\exp (x\beta ).$$

If β~0=β0+δ${\tilde{\beta }}_0={\beta }_0+\delta$ and λ~0(t)=λ0(t)exp(−δ)${\tilde{\lambda }}_0(t)={\lambda }_0(t)\exp (-\delta )$, then λ(t)=λ~0(t)exp(β~0+xβ)$\lambda (t)={\tilde{\lambda }}_0(t)\exp ({\tilde{\beta }}_0+\mathbf{x}\beta )$ as well, making the model with β0${\beta }_0$ unidentifiable.

In order to perform Bayesian inference with the Cox model, we must specify priors on β$\beta$ and λ0(t)${\lambda }_0(t)$. We place a normal prior on β$\beta$, β∼N(μβ,σ2β),$\beta \sim N({\mu }_{\beta },{\sigma }_{\beta }^2),$ where μβ∼N(0,102)${\mu }_{\beta }\sim N(0,{10}^2)$ and σβ∼U(0,10)${\sigma }_{\beta }\sim U(0,10)$.

A suitable prior on λ0(t)${\lambda }_0(t)$ is less obvious. We choose a semiparametric prior, where λ0(t)${\lambda }_0(t)$ is a piecewise constant function. This prior requires us to partition the time range in question into intervals with endpoints 0≤s1<s2<⋯<sN$0\le s_1<s_2<\cdots <s_N$. With this partition, λ0(t)=λj${\lambda }_0(t)={\lambda }_j$ if sj≤t<sj+1$s_j\le t<s_{j+1}$. With λ0(t)${\lambda }_0(t)$ constrained to have this form, all we need to do is choose priors for the N−1$N-1$ values λj${\lambda }_j$. We use independent vague priors λj∼Gamma(10−2,10−2).${\lambda }_j\sim \mathrm{Gamma}({10}^{-2},{10}^{-2}).$ For our mastectomy example, we make each interval three months long.

In [8]:

interval_length = 3
interval_bounds = np.arange(0, df.time.max() + interval_length + 1, interval_length)
n_intervals = interval_bounds.size - 1
intervals = np.arange(n_intervals)

We see how deaths and censored observations are distributed in these intervals.

In [9]:

fig, ax = plt.subplots(figsize=(8, 6))

ax.hist(df[df.event == 1].time.values, bins=interval_bounds,
        color=red, alpha=0.5, lw=0,
        label='Uncensored');
ax.hist(df[df.event == 0].time.values, bins=interval_bounds,
        color=blue, alpha=0.5, lw=0,
        label='Censored');

ax.set_xlim(0, interval_bounds[-1]);
ax.set_xlabel('Months since mastectomy');

ax.set_yticks([0, 1, 2, 3]);
ax.set_ylabel('Number of observations');

ax.legend();

With the prior distributions on β$\beta$ and λ0(t)${\lambda }_0(t)$ chosen, we now show how the model may be fit using MCMC simulation with pymc3. The key observation is that the piecewise-constant proportional hazard model is closely related to a Poisson regression model. (The models are not identical, but their likelihoods differ by a factor that depends only on the observed data and not the parameters β$\beta$and λj${\lambda }_j$. For details, see Germán Rodríguez’s WWS 509 course notes.)

We define indicator variables based on whether or the i$i$-th suject died in the j$j$-th interval,

d_{i, j} =

{10if subject i died in interval jotherwise$$\{\begin{array}{ll}1& \text{if subject }i\text{ died in interval }j\\ 0& \text{otherwise}\end{array}$$

.

In [10]:

last_period = np.floor((df.time - 0.01) / interval_length).astype(int)

death = np.zeros((n_patients, n_intervals))
death[patients, last_period] = df.event

We also define ti,j$t_{i,j}$ to be the amount of time the i$i$-th subject was at risk in the j$j$-th interval.

In [11]:

exposure = np.greater_equal.outer(df.time, interval_bounds[:-1]) * interval_length
exposure[patients, last_period] = df.time - interval_bounds[last_period]

Finally, denote the risk incurred by the i$i$-th subject in the j$j$-th interval as λi,j=λjexp(xiβ)${\lambda }_{i,j}={\lambda }_j\exp ({\mathbf{x}}_i\beta )$.

We may approximate di,j$d_{i,j}$ with a Possion random variable with mean ti,j λi,j$t_{i,j}{\lambda }_{i,j}$. This approximation leads to the following pymc3 model.

In [12]:

SEED = 5078864 # from random.org

In [13]:

with pm.Model() as model:

    lambda0 = pm.Gamma('lambda0', 0.01, 0.01, shape=n_intervals)

    beta = pm.Normal('beta', 0, sd=1000)

    lambda_ = pm.Deterministic('lambda_', T.outer(T.exp(beta * df.metastized), lambda0))
    mu = pm.Deterministic('mu', exposure * lambda_)

    obs = pm.Poisson('obs', mu, observed=death)

Applied log-transform to lambda0 and added transformed lambda0_log to model.

We now sample from the model.

In [14]:

n_samples = 1000

In [15]:

with model:
    trace_ = pm.sample(n_samples,random_seed=SEED)

Assigned NUTS to lambda0_log
Assigned NUTS to beta
 [-----------------100%-----------------] 1001 of 1000 complete in 246.7 sec

In [17]:

trace = trace_[100:]

We see that the hazard rate for subjects whose cancer has metastized is about double the rate of those whose cancer has not metastized.

In [34]:

np.exp(trace['beta'].mean())

Out[34]:

2.1839971003209597

In [18]:

pm.plot_posterior(trace, varnames=['beta'], color='#87ceeb');

In [19]:

pm.autocorrplot(trace, varnames=['beta']);

We now examine the effect of metastization on both the cumulative hazard and on the survival function.

In [20]:

base_hazard = trace['lambda0']
met_hazard = trace['lambda0'] * np.exp(np.atleast_2d(trace['beta']).T)

In [21]:

def cum_hazard(hazard):
    return (interval_length * hazard).cumsum(axis=-1)

def survival(hazard):
    return np.exp(-cum_hazard(hazard))

In [22]:

def plot_with_hpd(x, hazard, f, ax, color=None, label=None, alpha=0.05):
    mean = f(hazard.mean(axis=0))

    percentiles = 100 * np.array([alpha / 2., 1. - alpha / 2.])
    hpd = np.percentile(f(hazard), percentiles, axis=0)

    ax.fill_between(x, hpd[0], hpd[1], color=color, alpha=0.25)
    ax.step(x, mean, color=color, label=label);

In [23]:

fig, (hazard_ax, surv_ax) = plt.subplots(ncols=2, sharex=True, sharey=False, figsize=(16, 6))

plot_with_hpd(interval_bounds[:-1], base_hazard, cum_hazard,
              hazard_ax, color=blue, label='Had not metastized')
plot_with_hpd(interval_bounds[:-1], met_hazard, cum_hazard,
              hazard_ax, color=red, label='Metastized')

hazard_ax.set_xlim(0, df.time.max());
hazard_ax.set_xlabel('Months since mastectomy');

hazard_ax.set_ylabel(r'Cumulative hazard $\Lambda(t)$');

hazard_ax.legend(loc=2);

plot_with_hpd(interval_bounds[:-1], base_hazard, survival,
              surv_ax, color=blue)
plot_with_hpd(interval_bounds[:-1], met_hazard, survival,
              surv_ax, color=red)

surv_ax.set_xlim(0, df.time.max());
surv_ax.set_xlabel('Months since mastectomy');

surv_ax.set_ylabel('Survival function $S(t)$');

fig.suptitle('Bayesian survival model');

We see that the cumulative hazard for metastized subjects increases more rapidly initially (through about seventy months), after which it increases roughly in parallel with the baseline cumulative hazard.

These plots also show the pointwise 95% high posterior density interval for each function. One of the distinct advantages of the Bayesian model fit with pymc3 is the inherent quantification of uncertainty in our estimates.

Time varying effects

Another of the advantages of the model we have built is its flexibility. From the plots above, we may reasonable believe that the additional hazard due to metastization varies over time; it seems plausible that cancer that has metastized increases the hazard rate immediately after the mastectomy, but that the risk due to metastization decreases over time. We can accomodate this mechanism in our model by allowing the regression coefficients to vary over time. In the time-varying coefficent model, if sj≤t<sj+1$s_j\le t<s_{j+1}$, we let λ(t)=λjexp(xβj).$\lambda (t)={\lambda }_j\exp (\mathbf{x}{\beta }_j).$ The sequence of regression coefficients β1,β2,…,βN−1${\beta }_1,{\beta }_2,\dots ,{\beta }_{N-1}$ form a normal random walk with β1∼N(0,1)${\beta }_1\sim N(0,1)$, βj | βj−1∼N(βj−1,1)${\beta }_j|{\beta }_{j-1}\sim N({\beta }_{j-1},1)$.

We implement this model in pymc3 as follows.

In [24]:

with pm.Model() as time_varying_model:

    lambda0 = pm.Gamma('lambda0', 0.01, 0.01, shape=n_intervals)
    beta = GaussianRandomWalk('beta', tau=1., shape=n_intervals)

    lambda_ = pm.Deterministic('h', lambda0 * T.exp(T.outer(T.constant(df.metastized), beta)))
    mu = pm.Deterministic('mu', exposure * lambda_)

    obs = pm.Poisson('obs', mu, observed=death)

Applied log-transform to lambda0 and added transformed lambda0_log to model.

We proceed to sample from this model.

In [25]:

with time_varying_model:
    time_varying_trace_ = pm.sample(n_samples, random_seed=SEED)

Assigned NUTS to lambda0_log
Assigned NUTS to beta
 [-----------------100%-----------------] 1001 of 1000 complete in 949.5 sec

In [26]:

time_varying_trace = time_varying_trace_[100:]

In [27]:

pm.forestplot(time_varying_trace, varnames=['beta']);

We see from the plot of βj${\beta }_j$ over time below that initially βj>0${\beta }_j>0$, indicating an elevated hazard rate due to metastization, but that this risk declines as βj<0${\beta }_j<0$ eventually.

In [28]:

fig, ax = plt.subplots(figsize=(8, 6))

beta_hpd = np.percentile(time_varying_trace['beta'], [2.5, 97.5], axis=0)
beta_low = beta_hpd[0]
beta_high = beta_hpd[1]
ax.fill_between(interval_bounds[:-1], beta_low, beta_high,
                color=blue, alpha=0.25);
beta_hat = time_varying_trace['beta'].mean(axis=0)
ax.step(interval_bounds[:-1], beta_hat, color=blue);
ax.scatter(interval_bounds[last_period[(df.event.values == 1) & (df.metastized == 1)]],
           beta_hat[last_period[(df.event.values == 1) & (df.metastized == 1)]],
           c=red, zorder=10, label='Died, cancer metastized');
ax.scatter(interval_bounds[last_period[(df.event.values == 0) & (df.metastized == 1)]],
           beta_hat[last_period[(df.event.values == 0) & (df.metastized == 1)]],
           c=blue, zorder=10, label='Censored, cancer metastized');

ax.set_xlim(0, df.time.max());
ax.set_xlabel('Months since mastectomy');

ax.set_ylabel(r'$\beta_j$');

ax.legend();

The coefficients βj${\beta }_j$ begin declining rapidly around one hundred months post-mastectomy, which seems reasonable, given that only three of twelve subjects whose cancer had metastized lived past this point died during the study.

The change in our estimate of the cumulative hazard and survival functions due to time-varying effects is also quite apparent in the following plots.

In [29]:

tv_base_hazard = time_varying_trace['lambda0']
tv_met_hazard = time_varying_trace['lambda0'] * np.exp(np.atleast_2d(time_varying_trace['beta']))

In [30]:

fig, ax = plt.subplots(figsize=(8, 6))

ax.step(interval_bounds[:-1], cum_hazard(base_hazard.mean(axis=0)),
        color=blue, label='Had not metastized');
ax.step(interval_bounds[:-1], cum_hazard(met_hazard.mean(axis=0)),
        color=red, label='Metastized');

ax.step(interval_bounds[:-1], cum_hazard(tv_base_hazard.mean(axis=0)),
        color=blue, linestyle='--', label='Had not metastized (time varying effect)');
ax.step(interval_bounds[:-1], cum_hazard(tv_met_hazard.mean(axis=0)),
        color=red, linestyle='--', label='Metastized (time varying effect)');

ax.set_xlim(0, df.time.max() - 4);
ax.set_xlabel('Months since mastectomy');

ax.set_ylim(0, 2);
ax.set_ylabel(r'Cumulative hazard $\Lambda(t)$');

ax.legend(loc=2);

In [31]:

fig, (hazard_ax, surv_ax) = plt.subplots(ncols=2, sharex=True, sharey=False, figsize=(16, 6))

plot_with_hpd(interval_bounds[:-1], tv_base_hazard, cum_hazard,
              hazard_ax, color=blue, label='Had not metastized')
plot_with_hpd(interval_bounds[:-1], tv_met_hazard, cum_hazard,
              hazard_ax, color=red, label='Metastized')

hazard_ax.set_xlim(0, df.time.max());
hazard_ax.set_xlabel('Months since mastectomy');

hazard_ax.set_ylim(0, 2);
hazard_ax.set_ylabel(r'Cumulative hazard $\Lambda(t)$');

hazard_ax.legend(loc=2);

plot_with_hpd(interval_bounds[:-1], tv_base_hazard, survival,
              surv_ax, color=blue)
plot_with_hpd(interval_bounds[:-1], tv_met_hazard, survival,
              surv_ax, color=red)

surv_ax.set_xlim(0, df.time.max());
surv_ax.set_xlabel('Months since mastectomy');

surv_ax.set_ylabel('Survival function $S(t)$');

fig.suptitle('Bayesian survival model with time varying effects');

We have really only scratched the surface of both survival analysis and the Bayesian approach to survival analysis. More information on Bayesian survival analysis is available in Ibrahim et al. [1]_ (For example, we may want to account for individual frailty in either or original or time-varying models.)

This tutorial is available as an IPython notebook here. It is adapted from a blog post that first appeared